diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,264021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9943685455195057,
+  "eval_steps": 500,
+  "global_step": 44000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.9909,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.9824,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 0.9979,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.9784,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.0243,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 0.7993,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4e-05,
+      "loss": 0.8254,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.8486,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-05,
+      "loss": 0.8454,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 0.6792,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999974593536e-05,
+      "loss": 0.7772,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999898374144e-05,
+      "loss": 0.7329,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999771341824e-05,
+      "loss": 0.6691,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999999593496574e-05,
+      "loss": 0.668,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999993648384e-05,
+      "loss": 0.6422,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999990853673e-05,
+      "loss": 0.6852,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999998755083275e-05,
+      "loss": 0.6932,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999837398633e-05,
+      "loss": 0.6604,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999794207646e-05,
+      "loss": 0.6675,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999745935368e-05,
+      "loss": 0.6175,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999996925817976e-05,
+      "loss": 0.6721,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999634146936e-05,
+      "loss": 0.7605,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999995706307837e-05,
+      "loss": 0.7594,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999995020333407e-05,
+      "loss": 0.6646,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999994283546073e-05,
+      "loss": 0.6854,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999349594584e-05,
+      "loss": 0.5974,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999265753271e-05,
+      "loss": 0.5894,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999991768306687e-05,
+      "loss": 0.6329,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999999082826778e-05,
+      "loss": 0.6758,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999998983741599e-05,
+      "loss": 0.6067,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999988795751325e-05,
+      "loss": 0.6483,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999987703273786e-05,
+      "loss": 0.6174,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999998655998338e-05,
+      "loss": 0.6921,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999985365880117e-05,
+      "loss": 0.628,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999984120964e-05,
+      "loss": 0.5899,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999998282523503e-05,
+      "loss": 0.621,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999981478693226e-05,
+      "loss": 0.7353,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999998008133858e-05,
+      "loss": 0.6697,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999997863317111e-05,
+      "loss": 0.6902,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999997713419082e-05,
+      "loss": 0.6301,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999975584397717e-05,
+      "loss": 0.6609,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999973983791808e-05,
+      "loss": 0.6304,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999972332373104e-05,
+      "loss": 0.6619,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999970630141616e-05,
+      "loss": 0.7103,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999968877097342e-05,
+      "loss": 0.6727,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999670732403e-05,
+      "loss": 0.6565,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999652185705e-05,
+      "loss": 0.6822,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999996331308794e-05,
+      "loss": 0.6721,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999961356792647e-05,
+      "loss": 0.5848,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999959349684616e-05,
+      "loss": 0.6848,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999957291763864e-05,
+      "loss": 0.6204,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99999551830304e-05,
+      "loss": 0.6667,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999995302348424e-05,
+      "loss": 0.7393,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999950813125383e-05,
+      "loss": 0.6167,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999994855195385e-05,
+      "loss": 0.6454,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999994623996965e-05,
+      "loss": 0.6647,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999943877172798e-05,
+      "loss": 0.6818,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999941463563297e-05,
+      "loss": 0.66,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999938999141168e-05,
+      "loss": 0.6213,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999936483906416e-05,
+      "loss": 0.6715,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999933917859066e-05,
+      "loss": 0.7579,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999931300999117e-05,
+      "loss": 0.6687,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999992863332659e-05,
+      "loss": 0.635,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999925914841495e-05,
+      "loss": 0.6594,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999923145543853e-05,
+      "loss": 0.6305,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999992032543367e-05,
+      "loss": 0.6036,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999917454510962e-05,
+      "loss": 0.6631,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999914532775744e-05,
+      "loss": 0.6624,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999911560228033e-05,
+      "loss": 0.6977,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999908536867846e-05,
+      "loss": 0.5775,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999990546269519e-05,
+      "loss": 0.6821,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999902337710088e-05,
+      "loss": 0.6577,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999899161912552e-05,
+      "loss": 0.6624,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99998959353026e-05,
+      "loss": 0.6595,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999892657880254e-05,
+      "loss": 0.6295,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999988932964552e-05,
+      "loss": 0.6837,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999988595059842e-05,
+      "loss": 0.6006,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999988252073897e-05,
+      "loss": 0.6364,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999987904006719e-05,
+      "loss": 0.6456,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999875508583094e-05,
+      "loss": 0.7056,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999871926286702e-05,
+      "loss": 0.6716,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999868293178033e-05,
+      "loss": 0.6725,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999864609257104e-05,
+      "loss": 0.7592,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999986087452394e-05,
+      "loss": 0.6453,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999857088978548e-05,
+      "loss": 0.6452,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999853252620952e-05,
+      "loss": 0.6337,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999849365451178e-05,
+      "loss": 0.6689,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999984542746924e-05,
+      "loss": 0.5882,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999841438675155e-05,
+      "loss": 0.6507,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999983739906895e-05,
+      "loss": 0.6544,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999833308650642e-05,
+      "loss": 0.6458,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999829167420254e-05,
+      "loss": 0.6677,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999824975377805e-05,
+      "loss": 0.6789,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999820732523318e-05,
+      "loss": 0.6714,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999981643885681e-05,
+      "loss": 0.6836,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999812094378302e-05,
+      "loss": 0.7371,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999980769908783e-05,
+      "loss": 0.6447,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99998032529854e-05,
+      "loss": 0.6473,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999798756071044e-05,
+      "loss": 0.6628,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999794208344777e-05,
+      "loss": 0.6186,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999789609806633e-05,
+      "loss": 0.642,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999978496045663e-05,
+      "loss": 0.6727,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999780260294786e-05,
+      "loss": 0.655,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999775509321134e-05,
+      "loss": 0.6585,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999977070753569e-05,
+      "loss": 0.669,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999765854938484e-05,
+      "loss": 0.6407,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999760951529544e-05,
+      "loss": 0.6889,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999755997308882e-05,
+      "loss": 0.6441,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999750992276537e-05,
+      "loss": 0.6605,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999974593643253e-05,
+      "loss": 0.6136,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999974082977688e-05,
+      "loss": 0.6536,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999973567230962e-05,
+      "loss": 0.6944,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999730464030774e-05,
+      "loss": 0.6415,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999972520494037e-05,
+      "loss": 0.6477,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999719895038434e-05,
+      "loss": 0.653,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999971453432499e-05,
+      "loss": 0.6666,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999709122800074e-05,
+      "loss": 0.6148,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99997036604637e-05,
+      "loss": 0.6832,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999698147315908e-05,
+      "loss": 0.6611,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999969258335672e-05,
+      "loss": 0.6612,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999686968586166e-05,
+      "loss": 0.7165,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999681303004274e-05,
+      "loss": 0.6566,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999967558661107e-05,
+      "loss": 0.6613,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999669819406588e-05,
+      "loss": 0.6075,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999664001390858e-05,
+      "loss": 0.6626,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99996581325639e-05,
+      "loss": 0.6032,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999652212925756e-05,
+      "loss": 0.6149,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999646242476446e-05,
+      "loss": 0.6232,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999964022121601e-05,
+      "loss": 0.6603,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999963414914447e-05,
+      "loss": 0.6051,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999962802626186e-05,
+      "loss": 0.6183,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999621852568218e-05,
+      "loss": 0.7117,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999961562806356e-05,
+      "loss": 0.6306,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999960935274793e-05,
+      "loss": 0.6286,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999603026621357e-05,
+      "loss": 0.6584,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999596649683872e-05,
+      "loss": 0.6351,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999590221935507e-05,
+      "loss": 0.6882,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99995837433763e-05,
+      "loss": 0.6423,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999577214006277e-05,
+      "loss": 0.6378,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999570633825473e-05,
+      "loss": 0.6231,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999956400283392e-05,
+      "loss": 0.6098,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999557321031657e-05,
+      "loss": 0.6274,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999955058841871e-05,
+      "loss": 0.666,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999543804995124e-05,
+      "loss": 0.5958,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999536970760922e-05,
+      "loss": 0.6623,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999953008571615e-05,
+      "loss": 0.6809,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999952314986083e-05,
+      "loss": 0.6326,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999516163195013e-05,
+      "loss": 0.6444,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999950912571872e-05,
+      "loss": 0.6047,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999950203743199e-05,
+      "loss": 0.6825,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999494898334867e-05,
+      "loss": 0.6323,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999948770842738e-05,
+      "loss": 0.6542,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999948046770957e-05,
+      "loss": 0.6996,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999473176181467e-05,
+      "loss": 0.6612,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999465833843118e-05,
+      "loss": 0.6072,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999945844069455e-05,
+      "loss": 0.6298,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999450996735806e-05,
+      "loss": 0.6273,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999443501966925e-05,
+      "loss": 0.6236,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999943595638794e-05,
+      "loss": 0.6336,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999428359998893e-05,
+      "loss": 0.6489,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999420712799823e-05,
+      "loss": 0.6568,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999413014790767e-05,
+      "loss": 0.6712,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999405265971768e-05,
+      "loss": 0.6558,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999939746634286e-05,
+      "loss": 0.6958,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999389615904087e-05,
+      "loss": 0.6978,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999381714655482e-05,
+      "loss": 0.5999,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999373762597094e-05,
+      "loss": 0.5917,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999365759728958e-05,
+      "loss": 0.6307,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999357706051116e-05,
+      "loss": 0.6465,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999934960156361e-05,
+      "loss": 0.7325,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999341446266483e-05,
+      "loss": 0.674,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999333240159767e-05,
+      "loss": 0.6234,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999324983243515e-05,
+      "loss": 0.6656,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999316675517765e-05,
+      "loss": 0.7162,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999308316982557e-05,
+      "loss": 0.701,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999299907637935e-05,
+      "loss": 0.689,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999929144748394e-05,
+      "loss": 0.724,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999928293652062e-05,
+      "loss": 0.623,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999274374748015e-05,
+      "loss": 0.6698,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999265762166166e-05,
+      "loss": 0.7361,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999257098775117e-05,
+      "loss": 0.6837,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999924838457492e-05,
+      "loss": 0.6037,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999923961956561e-05,
+      "loss": 0.6448,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999230803747235e-05,
+      "loss": 0.6281,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999221937119842e-05,
+      "loss": 0.6935,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999213019683476e-05,
+      "loss": 0.6443,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999204051438174e-05,
+      "loss": 0.625,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999919503238399e-05,
+      "loss": 0.6192,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999918596252097e-05,
+      "loss": 0.5939,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999176841849154e-05,
+      "loss": 0.6487,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999167670368596e-05,
+      "loss": 0.6334,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999158448079336e-05,
+      "loss": 0.6681,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999149174981425e-05,
+      "loss": 0.6713,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999139851074907e-05,
+      "loss": 0.5888,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999130476359835e-05,
+      "loss": 0.6544,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999121050836248e-05,
+      "loss": 0.6688,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99991115745042e-05,
+      "loss": 0.6771,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999910204736374e-05,
+      "loss": 0.662,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999909246941491e-05,
+      "loss": 0.5866,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999082840657768e-05,
+      "loss": 0.6396,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999073161092354e-05,
+      "loss": 0.681,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999063430718722e-05,
+      "loss": 0.6618,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999905364953692e-05,
+      "loss": 0.6509,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999043817546996e-05,
+      "loss": 0.6245,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999033934749006e-05,
+      "loss": 0.6241,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999902400114299e-05,
+      "loss": 0.6415,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999014016729014e-05,
+      "loss": 0.5839,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9999003981507113e-05,
+      "loss": 0.6992,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998993895477347e-05,
+      "loss": 0.6688,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999898375863976e-05,
+      "loss": 0.6496,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998973570994412e-05,
+      "loss": 0.706,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998963332541352e-05,
+      "loss": 0.6619,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999895304328063e-05,
+      "loss": 0.6873,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99989427032123e-05,
+      "loss": 0.6134,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998932312336413e-05,
+      "loss": 0.6395,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999892187065302e-05,
+      "loss": 0.6551,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998911378162182e-05,
+      "loss": 0.6564,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998900834863943e-05,
+      "loss": 0.6896,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9998890240758363e-05,
+      "loss": 0.6632,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.999887959584549e-05,
+      "loss": 0.6651,
+      "step": 220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999886890012538e-05,
+      "loss": 0.6662,
+      "step": 221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998858153598096e-05,
+      "loss": 0.6428,
+      "step": 222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999884735626368e-05,
+      "loss": 0.6622,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998836508122194e-05,
+      "loss": 0.6659,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998825609173692e-05,
+      "loss": 0.6325,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999881465941823e-05,
+      "loss": 0.6451,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998803658855858e-05,
+      "loss": 0.6107,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999879260748664e-05,
+      "loss": 0.6509,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998781505310627e-05,
+      "loss": 0.6852,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998770352327878e-05,
+      "loss": 0.6676,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998759148538446e-05,
+      "loss": 0.6534,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999874789394239e-05,
+      "loss": 0.6343,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998736588539773e-05,
+      "loss": 0.6446,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999872523233064e-05,
+      "loss": 0.5957,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998713825315064e-05,
+      "loss": 0.7266,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999870236749309e-05,
+      "loss": 0.625,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999869085886478e-05,
+      "loss": 0.6254,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998679299430197e-05,
+      "loss": 0.6287,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998667689189393e-05,
+      "loss": 0.6241,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998656028142434e-05,
+      "loss": 0.663,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999864431628937e-05,
+      "loss": 0.6191,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999863255363027e-05,
+      "loss": 0.688,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999862074016519e-05,
+      "loss": 0.6991,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999860887589419e-05,
+      "loss": 0.6796,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998596960817327e-05,
+      "loss": 0.6705,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998584994934666e-05,
+      "loss": 0.5842,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998572978246267e-05,
+      "loss": 0.6656,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999856091075219e-05,
+      "loss": 0.6205,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998548792452495e-05,
+      "loss": 0.6468,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999853662334725e-05,
+      "loss": 0.7028,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998524403436508e-05,
+      "loss": 0.6399,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998512132720337e-05,
+      "loss": 0.5736,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99984998111988e-05,
+      "loss": 0.6656,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999848743887195e-05,
+      "loss": 0.6422,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998475015739864e-05,
+      "loss": 0.6438,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998462541802596e-05,
+      "loss": 0.6537,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998450017060214e-05,
+      "loss": 0.6778,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998437441512776e-05,
+      "loss": 0.6478,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998424815160348e-05,
+      "loss": 0.6698,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998412138003e-05,
+      "loss": 0.62,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998399410040788e-05,
+      "loss": 0.6875,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998386631273784e-05,
+      "loss": 0.63,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998373801702045e-05,
+      "loss": 0.6006,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998360921325643e-05,
+      "loss": 0.6473,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998347990144642e-05,
+      "loss": 0.6011,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998335008159103e-05,
+      "loss": 0.6108,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99983219753691e-05,
+      "loss": 0.6687,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998308891774694e-05,
+      "loss": 0.6512,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999829575737595e-05,
+      "loss": 0.6372,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998282572172937e-05,
+      "loss": 0.6031,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998269336165726e-05,
+      "loss": 0.6235,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998256049354378e-05,
+      "loss": 0.6273,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998242711738962e-05,
+      "loss": 0.614,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998229323319547e-05,
+      "loss": 0.6538,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.99982158840962e-05,
+      "loss": 0.5883,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998202394068994e-05,
+      "loss": 0.681,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999818885323799e-05,
+      "loss": 0.5738,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998175261603262e-05,
+      "loss": 0.645,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998161619164877e-05,
+      "loss": 0.6557,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998147925922905e-05,
+      "loss": 0.6425,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998134181877412e-05,
+      "loss": 0.6688,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998120387028475e-05,
+      "loss": 0.6805,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999810654137616e-05,
+      "loss": 0.6208,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998092644920534e-05,
+      "loss": 0.6202,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998078697661673e-05,
+      "loss": 0.6883,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999806469959965e-05,
+      "loss": 0.6711,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999805065073453e-05,
+      "loss": 0.6726,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9998036551066384e-05,
+      "loss": 0.6185,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999802240059529e-05,
+      "loss": 0.6425,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999800819932131e-05,
+      "loss": 0.6589,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997993947244526e-05,
+      "loss": 0.6387,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997979644365007e-05,
+      "loss": 0.6368,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997965290682827e-05,
+      "loss": 0.5894,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997950886198054e-05,
+      "loss": 0.7008,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997936430910765e-05,
+      "loss": 0.617,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997921924821036e-05,
+      "loss": 0.6027,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997907367928937e-05,
+      "loss": 0.6394,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999789276023454e-05,
+      "loss": 0.6651,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999787810173792e-05,
+      "loss": 0.6503,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999786339243916e-05,
+      "loss": 0.6098,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999784863233832e-05,
+      "loss": 0.6413,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999783382143549e-05,
+      "loss": 0.6298,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997818959730738e-05,
+      "loss": 0.6786,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997804047224138e-05,
+      "loss": 0.6549,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997789083915768e-05,
+      "loss": 0.6207,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997774069805702e-05,
+      "loss": 0.6488,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999775900489402e-05,
+      "loss": 0.6594,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997743889180793e-05,
+      "loss": 0.5959,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997728722666105e-05,
+      "loss": 0.6469,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999771350535003e-05,
+      "loss": 0.5684,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997698237232642e-05,
+      "loss": 0.6342,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997682918314023e-05,
+      "loss": 0.5682,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997667548594247e-05,
+      "loss": 0.6826,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997652128073396e-05,
+      "loss": 0.654,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997636656751548e-05,
+      "loss": 0.6771,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999762113462878e-05,
+      "loss": 0.596,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997605561705166e-05,
+      "loss": 0.6812,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997589937980796e-05,
+      "loss": 0.6021,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997574263455743e-05,
+      "loss": 0.6771,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997558538130086e-05,
+      "loss": 0.6317,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997542762003902e-05,
+      "loss": 0.6465,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999752693507728e-05,
+      "loss": 0.6814,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997511057350297e-05,
+      "loss": 0.5902,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999749512882303e-05,
+      "loss": 0.6993,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999747914949556e-05,
+      "loss": 0.6325,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997463119367974e-05,
+      "loss": 0.6898,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999744703844035e-05,
+      "loss": 0.6215,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999743090671277e-05,
+      "loss": 0.6236,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997414724185314e-05,
+      "loss": 0.6925,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997398490858067e-05,
+      "loss": 0.629,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999738220673111e-05,
+      "loss": 0.6586,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997365871804524e-05,
+      "loss": 0.6405,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997349486078398e-05,
+      "loss": 0.6028,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999733304955281e-05,
+      "loss": 0.6839,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997316562227846e-05,
+      "loss": 0.6271,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997300024103582e-05,
+      "loss": 0.6833,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997283435180117e-05,
+      "loss": 0.5999,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997266795457524e-05,
+      "loss": 0.638,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999725010493589e-05,
+      "loss": 0.6748,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997233363615302e-05,
+      "loss": 0.6021,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997216571495843e-05,
+      "loss": 0.6714,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997199728577598e-05,
+      "loss": 0.7153,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999718283486065e-05,
+      "loss": 0.6317,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997165890345093e-05,
+      "loss": 0.6315,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997148895031008e-05,
+      "loss": 0.6215,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999713184891848e-05,
+      "loss": 0.6337,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997114752007598e-05,
+      "loss": 0.6624,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999709760429845e-05,
+      "loss": 0.6002,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997080405791117e-05,
+      "loss": 0.6626,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997063156485696e-05,
+      "loss": 0.6544,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997045856382265e-05,
+      "loss": 0.6328,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9997028505480916e-05,
+      "loss": 0.6695,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999701110378174e-05,
+      "loss": 0.6682,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999699365128482e-05,
+      "loss": 0.6789,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999697614799025e-05,
+      "loss": 0.6086,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996958593898117e-05,
+      "loss": 0.5869,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996940989008508e-05,
+      "loss": 0.6711,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996923333321513e-05,
+      "loss": 0.6208,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996905626837227e-05,
+      "loss": 0.6633,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996887869555732e-05,
+      "loss": 0.6732,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996870061477126e-05,
+      "loss": 0.624,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996852202601492e-05,
+      "loss": 0.7069,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996834292928924e-05,
+      "loss": 0.6824,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996816332459515e-05,
+      "loss": 0.6258,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9996798321193355e-05,
+      "loss": 0.6157,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999678025913053e-05,
+      "loss": 0.6255,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.999676214627114e-05,
+      "loss": 0.6449,
+      "step": 367
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996743982615277e-05,
+      "loss": 0.6647,
+      "step": 368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999672576816303e-05,
+      "loss": 0.6275,
+      "step": 369
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996707502914486e-05,
+      "loss": 0.6149,
+      "step": 370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996689186869747e-05,
+      "loss": 0.6729,
+      "step": 371
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99966708200289e-05,
+      "loss": 0.6243,
+      "step": 372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996652402392044e-05,
+      "loss": 0.6592,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999663393395927e-05,
+      "loss": 0.6524,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999661541473067e-05,
+      "loss": 0.6709,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999659684470634e-05,
+      "loss": 0.6517,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996578223886374e-05,
+      "loss": 0.7183,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999655955227086e-05,
+      "loss": 0.6213,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996540829859907e-05,
+      "loss": 0.6153,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996522056653603e-05,
+      "loss": 0.5972,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999650323265204e-05,
+      "loss": 0.6764,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996484357855313e-05,
+      "loss": 0.6838,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996465432263527e-05,
+      "loss": 0.6765,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999644645587677e-05,
+      "loss": 0.6499,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996427428695143e-05,
+      "loss": 0.6642,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996408350718738e-05,
+      "loss": 0.6429,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996389221947657e-05,
+      "loss": 0.5567,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996370042381993e-05,
+      "loss": 0.7087,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996350812021844e-05,
+      "loss": 0.6255,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996331530867314e-05,
+      "loss": 0.6942,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999631219891849e-05,
+      "loss": 0.6917,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999629281617548e-05,
+      "loss": 0.6043,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999627338263838e-05,
+      "loss": 0.5973,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996253898307282e-05,
+      "loss": 0.6572,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996234363182293e-05,
+      "loss": 0.6953,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999621477726351e-05,
+      "loss": 0.6393,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996195140551033e-05,
+      "loss": 0.6532,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999617545304496e-05,
+      "loss": 0.6629,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999615571474539e-05,
+      "loss": 0.6797,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999613592565243e-05,
+      "loss": 0.633,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996116085766172e-05,
+      "loss": 0.6609,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999609619508672e-05,
+      "loss": 0.5938,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999607625361418e-05,
+      "loss": 0.7118,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996056261348648e-05,
+      "loss": 0.647,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999603621829022e-05,
+      "loss": 0.6594,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9996016124439013e-05,
+      "loss": 0.6768,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999599597979512e-05,
+      "loss": 0.5979,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999597578435864e-05,
+      "loss": 0.6077,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995955538129677e-05,
+      "loss": 0.6638,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995935241108344e-05,
+      "loss": 0.7035,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995914893294734e-05,
+      "loss": 0.5994,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999589449468895e-05,
+      "loss": 0.6121,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99958740452911e-05,
+      "loss": 0.5833,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995853545101288e-05,
+      "loss": 0.685,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995832994119616e-05,
+      "loss": 0.5667,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999581239234619e-05,
+      "loss": 0.6506,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995791739781115e-05,
+      "loss": 0.6468,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999577103642449e-05,
+      "loss": 0.6154,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995750282276426e-05,
+      "loss": 0.6223,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999572947733703e-05,
+      "loss": 0.6226,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999570862160641e-05,
+      "loss": 0.674,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999568771508466e-05,
+      "loss": 0.5705,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995666757771893e-05,
+      "loss": 0.6766,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995645749668216e-05,
+      "loss": 0.59,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995624690773737e-05,
+      "loss": 0.6355,
+      "step": 425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995603581088563e-05,
+      "loss": 0.6204,
+      "step": 426
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995582420612797e-05,
+      "loss": 0.618,
+      "step": 427
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999556120934655e-05,
+      "loss": 0.666,
+      "step": 428
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995539947289934e-05,
+      "loss": 0.6939,
+      "step": 429
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995518634443044e-05,
+      "loss": 0.6447,
+      "step": 430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995497270806005e-05,
+      "loss": 0.6449,
+      "step": 431
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999547585637891e-05,
+      "loss": 0.611,
+      "step": 432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995454391161878e-05,
+      "loss": 0.6775,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995432875155013e-05,
+      "loss": 0.626,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999541130835843e-05,
+      "loss": 0.648,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995389690772233e-05,
+      "loss": 0.6782,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995368022396535e-05,
+      "loss": 0.6407,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995346303231443e-05,
+      "loss": 0.7262,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995324533277067e-05,
+      "loss": 0.6442,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995302712533524e-05,
+      "loss": 0.5992,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995280841000917e-05,
+      "loss": 0.6378,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995258918679365e-05,
+      "loss": 0.6697,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999523694556897e-05,
+      "loss": 0.6524,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999521492166985e-05,
+      "loss": 0.6433,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999519284698212e-05,
+      "loss": 0.6595,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995170721505886e-05,
+      "loss": 0.7477,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999514854524126e-05,
+      "loss": 0.6148,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999512631818836e-05,
+      "loss": 0.6099,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999510404034729e-05,
+      "loss": 0.6677,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999508171171818e-05,
+      "loss": 0.6748,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995059332301124e-05,
+      "loss": 0.6147,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9995036902096248e-05,
+      "loss": 0.5752,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999501442110366e-05,
+      "loss": 0.6183,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999499188932348e-05,
+      "loss": 0.6268,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994969306755814e-05,
+      "loss": 0.6651,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994946673400785e-05,
+      "loss": 0.6249,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994923989258504e-05,
+      "loss": 0.6805,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999490125432909e-05,
+      "loss": 0.6347,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999487846861265e-05,
+      "loss": 0.6974,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999485563210931e-05,
+      "loss": 0.618,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999483274481918e-05,
+      "loss": 0.5781,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994809806742377e-05,
+      "loss": 0.5794,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994786817879017e-05,
+      "loss": 0.6289,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999476377822922e-05,
+      "loss": 0.6195,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99947406877931e-05,
+      "loss": 0.663,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994717546570777e-05,
+      "loss": 0.6534,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994694354562368e-05,
+      "loss": 0.6376,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994671111767987e-05,
+      "loss": 0.6924,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994647818187757e-05,
+      "loss": 0.6817,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994624473821793e-05,
+      "loss": 0.614,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994601078670217e-05,
+      "loss": 0.6487,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994577632733144e-05,
+      "loss": 0.5966,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994554136010697e-05,
+      "loss": 0.6415,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999453058850299e-05,
+      "loss": 0.6189,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999450699021015e-05,
+      "loss": 0.6318,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999448334113229e-05,
+      "loss": 0.691,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994459641269533e-05,
+      "loss": 0.6349,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994435890622e-05,
+      "loss": 0.6368,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994412089189814e-05,
+      "loss": 0.6954,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999438823697309e-05,
+      "loss": 0.6673,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994364333971953e-05,
+      "loss": 0.5896,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994340380186522e-05,
+      "loss": 0.6708,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994316375616923e-05,
+      "loss": 0.6625,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994292320263277e-05,
+      "loss": 0.6428,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994268214125702e-05,
+      "loss": 0.6802,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994244057204325e-05,
+      "loss": 0.6557,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994219849499264e-05,
+      "loss": 0.6456,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999419559101065e-05,
+      "loss": 0.659,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994171281738598e-05,
+      "loss": 0.656,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994146921683236e-05,
+      "loss": 0.6762,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994122510844685e-05,
+      "loss": 0.6381,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994098049223073e-05,
+      "loss": 0.6554,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994073536818523e-05,
+      "loss": 0.654,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994048973631152e-05,
+      "loss": 0.6435,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9994024359661093e-05,
+      "loss": 0.6828,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993999694908475e-05,
+      "loss": 0.6184,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993974979373413e-05,
+      "loss": 0.6936,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999395021305604e-05,
+      "loss": 0.6557,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993925395956478e-05,
+      "loss": 0.6466,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993900528074853e-05,
+      "loss": 0.6571,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999387560941129e-05,
+      "loss": 0.683,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993850639965923e-05,
+      "loss": 0.6155,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993825619738873e-05,
+      "loss": 0.6163,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999380054873027e-05,
+      "loss": 0.6493,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993775426940235e-05,
+      "loss": 0.6677,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99937502543689e-05,
+      "loss": 0.6147,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99937250310164e-05,
+      "loss": 0.7208,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993699756882853e-05,
+      "loss": 0.7127,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999367443196839e-05,
+      "loss": 0.6159,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999364905627314e-05,
+      "loss": 0.6139,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999362362979723e-05,
+      "loss": 0.6213,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993598152540797e-05,
+      "loss": 0.6513,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999357262450396e-05,
+      "loss": 0.6187,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9993547045686856e-05,
+      "loss": 0.62,
+      "step": 514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993521416089613e-05,
+      "loss": 0.617,
+      "step": 515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999349573571236e-05,
+      "loss": 0.6066,
+      "step": 516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999347000455523e-05,
+      "loss": 0.6559,
+      "step": 517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993444222618352e-05,
+      "loss": 0.6038,
+      "step": 518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993418389901856e-05,
+      "loss": 0.6069,
+      "step": 519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993392506405877e-05,
+      "loss": 0.6491,
+      "step": 520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999336657213054e-05,
+      "loss": 0.5939,
+      "step": 521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999334058707598e-05,
+      "loss": 0.6304,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993314551242332e-05,
+      "loss": 0.5815,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999328846462973e-05,
+      "loss": 0.58,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993262327238297e-05,
+      "loss": 0.6336,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993236139068175e-05,
+      "loss": 0.6232,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999320990011949e-05,
+      "loss": 0.6413,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993183610392383e-05,
+      "loss": 0.6847,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999315726988698e-05,
+      "loss": 0.5802,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993130878603423e-05,
+      "loss": 0.5627,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993104436541836e-05,
+      "loss": 0.6385,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993077943702358e-05,
+      "loss": 0.6718,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999305140008513e-05,
+      "loss": 0.6333,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993024805690278e-05,
+      "loss": 0.6078,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999299816051794e-05,
+      "loss": 0.6512,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992971464568255e-05,
+      "loss": 0.6474,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999294471784135e-05,
+      "loss": 0.6553,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999291792033737e-05,
+      "loss": 0.6383,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992891072056447e-05,
+      "loss": 0.598,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999286417299872e-05,
+      "loss": 0.6423,
+      "step": 540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999283722316432e-05,
+      "loss": 0.6286,
+      "step": 541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992810222553387e-05,
+      "loss": 0.6951,
+      "step": 542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999278317116606e-05,
+      "loss": 0.6301,
+      "step": 543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992756069002474e-05,
+      "loss": 0.6891,
+      "step": 544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999272891606277e-05,
+      "loss": 0.614,
+      "step": 545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992701712347084e-05,
+      "loss": 0.6505,
+      "step": 546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992674457855553e-05,
+      "loss": 0.7414,
+      "step": 547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999264715258832e-05,
+      "loss": 0.6541,
+      "step": 548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992619796545516e-05,
+      "loss": 0.6632,
+      "step": 549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992592389727286e-05,
+      "loss": 0.6348,
+      "step": 550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999256493213377e-05,
+      "loss": 0.6254,
+      "step": 551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99925374237651e-05,
+      "loss": 0.6537,
+      "step": 552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992509864621424e-05,
+      "loss": 0.7016,
+      "step": 553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992482254702878e-05,
+      "loss": 0.6523,
+      "step": 554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992454594009603e-05,
+      "loss": 0.623,
+      "step": 555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999242688254174e-05,
+      "loss": 0.6514,
+      "step": 556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999239912029943e-05,
+      "loss": 0.6464,
+      "step": 557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992371307282814e-05,
+      "loss": 0.6152,
+      "step": 558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992343443492034e-05,
+      "loss": 0.6573,
+      "step": 559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999231552892723e-05,
+      "loss": 0.635,
+      "step": 560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992287563588543e-05,
+      "loss": 0.6474,
+      "step": 561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992259547476116e-05,
+      "loss": 0.6151,
+      "step": 562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992231480590095e-05,
+      "loss": 0.7176,
+      "step": 563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992203362930616e-05,
+      "loss": 0.6962,
+      "step": 564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992175194497826e-05,
+      "loss": 0.6326,
+      "step": 565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999214697529187e-05,
+      "loss": 0.6604,
+      "step": 566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992118705312884e-05,
+      "loss": 0.6622,
+      "step": 567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999209038456102e-05,
+      "loss": 0.6275,
+      "step": 568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999206201303642e-05,
+      "loss": 0.6921,
+      "step": 569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992033590739224e-05,
+      "loss": 0.6889,
+      "step": 570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992005117669578e-05,
+      "loss": 0.6801,
+      "step": 571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991976593827632e-05,
+      "loss": 0.6521,
+      "step": 572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999194801921352e-05,
+      "loss": 0.6277,
+      "step": 573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991919393827402e-05,
+      "loss": 0.6621,
+      "step": 574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999189071766941e-05,
+      "loss": 0.6333,
+      "step": 575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991861990739697e-05,
+      "loss": 0.5877,
+      "step": 576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991833213038406e-05,
+      "loss": 0.671,
+      "step": 577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991804384565685e-05,
+      "loss": 0.6982,
+      "step": 578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999177550532168e-05,
+      "loss": 0.6486,
+      "step": 579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991746575306536e-05,
+      "loss": 0.6035,
+      "step": 580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99917175945204e-05,
+      "loss": 0.6582,
+      "step": 581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991688562963424e-05,
+      "loss": 0.6458,
+      "step": 582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999165948063575e-05,
+      "loss": 0.637,
+      "step": 583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999163034753753e-05,
+      "loss": 0.6849,
+      "step": 584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991601163668906e-05,
+      "loss": 0.6242,
+      "step": 585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991571929030037e-05,
+      "loss": 0.6421,
+      "step": 586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999154264362106e-05,
+      "loss": 0.6055,
+      "step": 587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999151330744213e-05,
+      "loss": 0.6156,
+      "step": 588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991483920493394e-05,
+      "loss": 0.6199,
+      "step": 589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991454482775e-05,
+      "loss": 0.5817,
+      "step": 590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991424994287106e-05,
+      "loss": 0.663,
+      "step": 591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991395455029853e-05,
+      "loss": 0.626,
+      "step": 592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999136586500339e-05,
+      "loss": 0.6568,
+      "step": 593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999133622420787e-05,
+      "loss": 0.6287,
+      "step": 594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999130653264345e-05,
+      "loss": 0.6201,
+      "step": 595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991276790310272e-05,
+      "loss": 0.688,
+      "step": 596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991246997208492e-05,
+      "loss": 0.6105,
+      "step": 597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991217153338262e-05,
+      "loss": 0.6606,
+      "step": 598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999118725869973e-05,
+      "loss": 0.6583,
+      "step": 599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991157313293047e-05,
+      "loss": 0.6399,
+      "step": 600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991127317118368e-05,
+      "loss": 0.6057,
+      "step": 601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991097270175847e-05,
+      "loss": 0.6271,
+      "step": 602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991067172465634e-05,
+      "loss": 0.6348,
+      "step": 603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999103702398788e-05,
+      "loss": 0.6271,
+      "step": 604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991006824742745e-05,
+      "loss": 0.6494,
+      "step": 605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990976574730378e-05,
+      "loss": 0.6391,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990946273950934e-05,
+      "loss": 0.6317,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990915922404564e-05,
+      "loss": 0.6477,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990885520091424e-05,
+      "loss": 0.6683,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999085506701167e-05,
+      "loss": 0.6228,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999082456316545e-05,
+      "loss": 0.7337,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999079400855293e-05,
+      "loss": 0.6935,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999076340317426e-05,
+      "loss": 0.637,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990732747029594e-05,
+      "loss": 0.6253,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999070204011909e-05,
+      "loss": 0.5963,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990671282442898e-05,
+      "loss": 0.6413,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990640474001183e-05,
+      "loss": 0.6252,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99906096147941e-05,
+      "loss": 0.6336,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990578704821802e-05,
+      "loss": 0.6319,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990547744084447e-05,
+      "loss": 0.6227,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990516732582192e-05,
+      "loss": 0.6382,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99904856703152e-05,
+      "loss": 0.6832,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990454557283622e-05,
+      "loss": 0.6065,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990423393487615e-05,
+      "loss": 0.6408,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990392178927344e-05,
+      "loss": 0.6174,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990360913602966e-05,
+      "loss": 0.6415,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990329597514636e-05,
+      "loss": 0.6685,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990298230662513e-05,
+      "loss": 0.7042,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990266813046764e-05,
+      "loss": 0.6616,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990235344667545e-05,
+      "loss": 0.6344,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990203825525007e-05,
+      "loss": 0.6332,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999017225561932e-05,
+      "loss": 0.6508,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999014063495065e-05,
+      "loss": 0.6013,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999010896351914e-05,
+      "loss": 0.6091,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990077241324966e-05,
+      "loss": 0.6602,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999004546836828e-05,
+      "loss": 0.7262,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999001364464925e-05,
+      "loss": 0.606,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989981770168036e-05,
+      "loss": 0.6455,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989949844924795e-05,
+      "loss": 0.6174,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989917868919696e-05,
+      "loss": 0.5955,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989885842152898e-05,
+      "loss": 0.6535,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989853764624564e-05,
+      "loss": 0.6653,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989821636334857e-05,
+      "loss": 0.6291,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998978945728394e-05,
+      "loss": 0.685,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998975722747198e-05,
+      "loss": 0.6441,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989724946899133e-05,
+      "loss": 0.6573,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989692615565573e-05,
+      "loss": 0.6198,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998966023347146e-05,
+      "loss": 0.6188,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989627800616954e-05,
+      "loss": 0.6308,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989595317002225e-05,
+      "loss": 0.6414,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989562782627434e-05,
+      "loss": 0.6043,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998953019749275e-05,
+      "loss": 0.6711,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989497561598342e-05,
+      "loss": 0.6472,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989464874944367e-05,
+      "loss": 0.6056,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989432137530994e-05,
+      "loss": 0.65,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989399349358396e-05,
+      "loss": 0.6175,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998936651042673e-05,
+      "loss": 0.6724,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989333620736168e-05,
+      "loss": 0.6128,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989300680286877e-05,
+      "loss": 0.6893,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989267689079024e-05,
+      "loss": 0.6176,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989234647112776e-05,
+      "loss": 0.6323,
+      "step": 661
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9989201554388303e-05,
+      "loss": 0.6277,
+      "step": 662
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9989168410905767e-05,
+      "loss": 0.6903,
+      "step": 663
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9989135216665347e-05,
+      "loss": 0.6933,
+      "step": 664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99891019716672e-05,
+      "loss": 0.6399,
+      "step": 665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9989068675911506e-05,
+      "loss": 0.6558,
+      "step": 666
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998903532939842e-05,
+      "loss": 0.6717,
+      "step": 667
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998900193212813e-05,
+      "loss": 0.604,
+      "step": 668
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998896848410079e-05,
+      "loss": 0.6005,
+      "step": 669
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988934985316576e-05,
+      "loss": 0.6113,
+      "step": 670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998890143577566e-05,
+      "loss": 0.5884,
+      "step": 671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988867835478206e-05,
+      "loss": 0.624,
+      "step": 672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988834184424396e-05,
+      "loss": 0.6516,
+      "step": 673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988800482614387e-05,
+      "loss": 0.6477,
+      "step": 674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988766730048364e-05,
+      "loss": 0.6397,
+      "step": 675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988732926726487e-05,
+      "loss": 0.6087,
+      "step": 676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988699072648936e-05,
+      "loss": 0.5752,
+      "step": 677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998866516781588e-05,
+      "loss": 0.6667,
+      "step": 678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988631212227488e-05,
+      "loss": 0.6115,
+      "step": 679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988597205883933e-05,
+      "loss": 0.6133,
+      "step": 680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99885631487854e-05,
+      "loss": 0.7039,
+      "step": 681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988529040932043e-05,
+      "loss": 0.6431,
+      "step": 682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988494882324047e-05,
+      "loss": 0.6254,
+      "step": 683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988460672961585e-05,
+      "loss": 0.6097,
+      "step": 684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998842641284483e-05,
+      "loss": 0.6595,
+      "step": 685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998839210197395e-05,
+      "loss": 0.6225,
+      "step": 686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998835774034913e-05,
+      "loss": 0.6642,
+      "step": 687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988323327970536e-05,
+      "loss": 0.6276,
+      "step": 688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998828886483835e-05,
+      "loss": 0.6485,
+      "step": 689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988254350952737e-05,
+      "loss": 0.6082,
+      "step": 690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998821978631388e-05,
+      "loss": 0.6062,
+      "step": 691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988185170921957e-05,
+      "loss": 0.6016,
+      "step": 692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988150504777138e-05,
+      "loss": 0.6808,
+      "step": 693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99881157878796e-05,
+      "loss": 0.6644,
+      "step": 694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998808102022952e-05,
+      "loss": 0.6317,
+      "step": 695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998804620182708e-05,
+      "loss": 0.6118,
+      "step": 696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988011332672447e-05,
+      "loss": 0.6745,
+      "step": 697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987976412765803e-05,
+      "loss": 0.6408,
+      "step": 698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987941442107327e-05,
+      "loss": 0.6048,
+      "step": 699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987906420697198e-05,
+      "loss": 0.6461,
+      "step": 700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998787134853559e-05,
+      "loss": 0.6714,
+      "step": 701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987836225622683e-05,
+      "loss": 0.621,
+      "step": 702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987801051958652e-05,
+      "loss": 0.5829,
+      "step": 703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998776582754368e-05,
+      "loss": 0.6395,
+      "step": 704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987730552377944e-05,
+      "loss": 0.6206,
+      "step": 705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998769522646163e-05,
+      "loss": 0.6105,
+      "step": 706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99876598497949e-05,
+      "loss": 0.6463,
+      "step": 707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987624422377957e-05,
+      "loss": 0.5797,
+      "step": 708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987588944210966e-05,
+      "loss": 0.6857,
+      "step": 709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987553415294108e-05,
+      "loss": 0.5929,
+      "step": 710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987517835627565e-05,
+      "loss": 0.6852,
+      "step": 711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998748220521152e-05,
+      "loss": 0.6392,
+      "step": 712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987446524046156e-05,
+      "loss": 0.6168,
+      "step": 713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987410792131648e-05,
+      "loss": 0.6223,
+      "step": 714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987375009468188e-05,
+      "loss": 0.5375,
+      "step": 715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998733917605594e-05,
+      "loss": 0.6826,
+      "step": 716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987303291895105e-05,
+      "loss": 0.6586,
+      "step": 717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987267356985855e-05,
+      "loss": 0.6093,
+      "step": 718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998723137132837e-05,
+      "loss": 0.6434,
+      "step": 719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998719533492284e-05,
+      "loss": 0.6453,
+      "step": 720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987159247769448e-05,
+      "loss": 0.6157,
+      "step": 721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987123109868373e-05,
+      "loss": 0.6461,
+      "step": 722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99870869212198e-05,
+      "loss": 0.619,
+      "step": 723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987050681823915e-05,
+      "loss": 0.643,
+      "step": 724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99870143916809e-05,
+      "loss": 0.6817,
+      "step": 725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998697805079094e-05,
+      "loss": 0.5924,
+      "step": 726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998694165915422e-05,
+      "loss": 0.6266,
+      "step": 727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986905216770927e-05,
+      "loss": 0.6174,
+      "step": 728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998686872364124e-05,
+      "loss": 0.6532,
+      "step": 729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998683217976535e-05,
+      "loss": 0.6455,
+      "step": 730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986795585143438e-05,
+      "loss": 0.6114,
+      "step": 731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986758939775698e-05,
+      "loss": 0.652,
+      "step": 732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986722243662305e-05,
+      "loss": 0.609,
+      "step": 733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986685496803457e-05,
+      "loss": 0.6738,
+      "step": 734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986648699199332e-05,
+      "loss": 0.5903,
+      "step": 735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998661185085012e-05,
+      "loss": 0.6334,
+      "step": 736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986574951756013e-05,
+      "loss": 0.6099,
+      "step": 737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998653800191719e-05,
+      "loss": 0.6599,
+      "step": 738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986501001333845e-05,
+      "loss": 0.6808,
+      "step": 739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986463950006165e-05,
+      "loss": 0.646,
+      "step": 740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998642684793433e-05,
+      "loss": 0.647,
+      "step": 741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986389695118542e-05,
+      "loss": 0.6847,
+      "step": 742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998635249155898e-05,
+      "loss": 0.663,
+      "step": 743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998631523725584e-05,
+      "loss": 0.7056,
+      "step": 744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986277932209302e-05,
+      "loss": 0.6394,
+      "step": 745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986240576419566e-05,
+      "loss": 0.6903,
+      "step": 746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986203169886817e-05,
+      "loss": 0.6222,
+      "step": 747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998616571261124e-05,
+      "loss": 0.6882,
+      "step": 748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986128204593034e-05,
+      "loss": 0.6969,
+      "step": 749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986090645832384e-05,
+      "loss": 0.5842,
+      "step": 750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986053036329487e-05,
+      "loss": 0.6248,
+      "step": 751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998601537608453e-05,
+      "loss": 0.5251,
+      "step": 752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985977665097702e-05,
+      "loss": 0.6377,
+      "step": 753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985939903369195e-05,
+      "loss": 0.6026,
+      "step": 754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985902090899205e-05,
+      "loss": 0.6179,
+      "step": 755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998586422768792e-05,
+      "loss": 0.6004,
+      "step": 756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998582631373554e-05,
+      "loss": 0.6001,
+      "step": 757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998578834904225e-05,
+      "loss": 0.6,
+      "step": 758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998575033360824e-05,
+      "loss": 0.6175,
+      "step": 759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998571226743371e-05,
+      "loss": 0.5732,
+      "step": 760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985674150518856e-05,
+      "loss": 0.6506,
+      "step": 761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985635982863862e-05,
+      "loss": 0.6468,
+      "step": 762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985597764468928e-05,
+      "loss": 0.6152,
+      "step": 763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998555949533425e-05,
+      "loss": 0.5821,
+      "step": 764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985521175460017e-05,
+      "loss": 0.6543,
+      "step": 765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998548280484643e-05,
+      "loss": 0.7131,
+      "step": 766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985444383493676e-05,
+      "loss": 0.6398,
+      "step": 767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998540591140196e-05,
+      "loss": 0.6415,
+      "step": 768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998536738857147e-05,
+      "loss": 0.648,
+      "step": 769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99853288150024e-05,
+      "loss": 0.601,
+      "step": 770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985290190694956e-05,
+      "loss": 0.5923,
+      "step": 771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998525151564932e-05,
+      "loss": 0.5906,
+      "step": 772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985212789865703e-05,
+      "loss": 0.6987,
+      "step": 773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985174013344295e-05,
+      "loss": 0.6293,
+      "step": 774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985135186085294e-05,
+      "loss": 0.6516,
+      "step": 775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985096308088897e-05,
+      "loss": 0.5895,
+      "step": 776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985057379355298e-05,
+      "loss": 0.6373,
+      "step": 777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99850183998847e-05,
+      "loss": 0.6819,
+      "step": 778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984979369677303e-05,
+      "loss": 0.6296,
+      "step": 779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984940288733296e-05,
+      "loss": 0.6451,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984901157052883e-05,
+      "loss": 0.6557,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984861974636268e-05,
+      "loss": 0.6338,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984822741483642e-05,
+      "loss": 0.5714,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998478345759521e-05,
+      "loss": 0.6564,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984744122971168e-05,
+      "loss": 0.5895,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984704737611716e-05,
+      "loss": 0.6633,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984665301517054e-05,
+      "loss": 0.6165,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984625814687385e-05,
+      "loss": 0.6223,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984586277122905e-05,
+      "loss": 0.6181,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984546688823822e-05,
+      "loss": 0.6244,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984507049790332e-05,
+      "loss": 0.6199,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984467360022638e-05,
+      "loss": 0.657,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998442761952094e-05,
+      "loss": 0.6649,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984387828285444e-05,
+      "loss": 0.6222,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984347986316348e-05,
+      "loss": 0.6178,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984308093613854e-05,
+      "loss": 0.6466,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998426815017817e-05,
+      "loss": 0.602,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998422815600949e-05,
+      "loss": 0.6732,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984188111108027e-05,
+      "loss": 0.701,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998414801547398e-05,
+      "loss": 0.5346,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998410786910755e-05,
+      "loss": 0.6391,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984067672008947e-05,
+      "loss": 0.5881,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984027424178373e-05,
+      "loss": 0.7131,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983987125616028e-05,
+      "loss": 0.692,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998394677632212e-05,
+      "loss": 0.6328,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983906376296857e-05,
+      "loss": 0.6038,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998386592554044e-05,
+      "loss": 0.5721,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983825424053084e-05,
+      "loss": 0.5697,
+      "step": 808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998378487183498e-05,
+      "loss": 0.6136,
+      "step": 809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998374426888634e-05,
+      "loss": 0.6461,
+      "step": 810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983703615207376e-05,
+      "loss": 0.602,
+      "step": 811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983662910798288e-05,
+      "loss": 0.6946,
+      "step": 812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983622155659282e-05,
+      "loss": 0.5998,
+      "step": 813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998358134979057e-05,
+      "loss": 0.6871,
+      "step": 814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998354049319236e-05,
+      "loss": 0.6378,
+      "step": 815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983499585864853e-05,
+      "loss": 0.6593,
+      "step": 816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983458627808265e-05,
+      "loss": 0.6342,
+      "step": 817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983417619022797e-05,
+      "loss": 0.6134,
+      "step": 818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998337655950866e-05,
+      "loss": 0.5926,
+      "step": 819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983335449266063e-05,
+      "loss": 0.6326,
+      "step": 820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983294288295218e-05,
+      "loss": 0.661,
+      "step": 821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983253076596326e-05,
+      "loss": 0.6348,
+      "step": 822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983211814169606e-05,
+      "loss": 0.6021,
+      "step": 823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983170501015262e-05,
+      "loss": 0.679,
+      "step": 824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983129137133504e-05,
+      "loss": 0.6833,
+      "step": 825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983087722524545e-05,
+      "loss": 0.6834,
+      "step": 826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998304625718859e-05,
+      "loss": 0.6261,
+      "step": 827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983004741125852e-05,
+      "loss": 0.6044,
+      "step": 828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998296317433655e-05,
+      "loss": 0.6128,
+      "step": 829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982921556820885e-05,
+      "loss": 0.6545,
+      "step": 830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982879888579074e-05,
+      "loss": 0.6985,
+      "step": 831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998283816961132e-05,
+      "loss": 0.6185,
+      "step": 832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982796399917848e-05,
+      "loss": 0.5955,
+      "step": 833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982754579498863e-05,
+      "loss": 0.6254,
+      "step": 834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982712708354576e-05,
+      "loss": 0.6948,
+      "step": 835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982670786485205e-05,
+      "loss": 0.6276,
+      "step": 836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982628813890958e-05,
+      "loss": 0.6789,
+      "step": 837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998258679057205e-05,
+      "loss": 0.6587,
+      "step": 838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982544716528695e-05,
+      "loss": 0.6783,
+      "step": 839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998250259176111e-05,
+      "loss": 0.6133,
+      "step": 840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982460416269505e-05,
+      "loss": 0.6504,
+      "step": 841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998241819005409e-05,
+      "loss": 0.6386,
+      "step": 842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982375913115092e-05,
+      "loss": 0.6666,
+      "step": 843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982333585452714e-05,
+      "loss": 0.5729,
+      "step": 844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982291207067174e-05,
+      "loss": 0.645,
+      "step": 845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982248777958696e-05,
+      "loss": 0.6443,
+      "step": 846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982206298127482e-05,
+      "loss": 0.6103,
+      "step": 847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982163767573758e-05,
+      "loss": 0.6407,
+      "step": 848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982121186297735e-05,
+      "loss": 0.6411,
+      "step": 849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982078554299627e-05,
+      "loss": 0.6478,
+      "step": 850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982035871579662e-05,
+      "loss": 0.6412,
+      "step": 851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981993138138047e-05,
+      "loss": 0.595,
+      "step": 852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981950353974997e-05,
+      "loss": 0.672,
+      "step": 853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981907519090737e-05,
+      "loss": 0.6361,
+      "step": 854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981864633485483e-05,
+      "loss": 0.5984,
+      "step": 855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998182169715945e-05,
+      "loss": 0.6976,
+      "step": 856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998177871011286e-05,
+      "loss": 0.6599,
+      "step": 857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981735672345922e-05,
+      "loss": 0.6674,
+      "step": 858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998169258385887e-05,
+      "loss": 0.6298,
+      "step": 859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998164944465191e-05,
+      "loss": 0.6555,
+      "step": 860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998160625472527e-05,
+      "loss": 0.5556,
+      "step": 861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998156301407916e-05,
+      "loss": 0.6361,
+      "step": 862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998151972271381e-05,
+      "loss": 0.6507,
+      "step": 863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998147638062943e-05,
+      "loss": 0.6315,
+      "step": 864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981432987826248e-05,
+      "loss": 0.6395,
+      "step": 865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998138954430448e-05,
+      "loss": 0.637,
+      "step": 866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998134605006435e-05,
+      "loss": 0.5941,
+      "step": 867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981302505106082e-05,
+      "loss": 0.6187,
+      "step": 868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981258909429887e-05,
+      "loss": 0.6238,
+      "step": 869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981215263035994e-05,
+      "loss": 0.6051,
+      "step": 870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981171565924625e-05,
+      "loss": 0.6323,
+      "step": 871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981127818096e-05,
+      "loss": 0.6115,
+      "step": 872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981084019550337e-05,
+      "loss": 0.6411,
+      "step": 873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981040170287866e-05,
+      "loss": 0.6422,
+      "step": 874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980996270308806e-05,
+      "loss": 0.634,
+      "step": 875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998095231961338e-05,
+      "loss": 0.6218,
+      "step": 876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980908318201817e-05,
+      "loss": 0.5828,
+      "step": 877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998086426607433e-05,
+      "loss": 0.6314,
+      "step": 878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998082016323115e-05,
+      "loss": 0.6337,
+      "step": 879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99807760096725e-05,
+      "loss": 0.6225,
+      "step": 880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980731805398606e-05,
+      "loss": 0.6613,
+      "step": 881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998068755040969e-05,
+      "loss": 0.6479,
+      "step": 882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998064324470598e-05,
+      "loss": 0.622,
+      "step": 883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980598888287695e-05,
+      "loss": 0.6583,
+      "step": 884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980554481155064e-05,
+      "loss": 0.6417,
+      "step": 885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980510023308314e-05,
+      "loss": 0.7261,
+      "step": 886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998046551474767e-05,
+      "loss": 0.635,
+      "step": 887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998042095547336e-05,
+      "loss": 0.6075,
+      "step": 888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99803763454856e-05,
+      "loss": 0.634,
+      "step": 889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980331684784634e-05,
+      "loss": 0.5898,
+      "step": 890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980286973370673e-05,
+      "loss": 0.6347,
+      "step": 891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980242211243956e-05,
+      "loss": 0.7084,
+      "step": 892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980197398404705e-05,
+      "loss": 0.5902,
+      "step": 893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980152534853145e-05,
+      "loss": 0.6466,
+      "step": 894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998010762058951e-05,
+      "loss": 0.6518,
+      "step": 895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980062655614026e-05,
+      "loss": 0.6097,
+      "step": 896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998001763992692e-05,
+      "loss": 0.6362,
+      "step": 897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979972573528416e-05,
+      "loss": 0.6266,
+      "step": 898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979927456418756e-05,
+      "loss": 0.6753,
+      "step": 899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979882288598158e-05,
+      "loss": 0.6163,
+      "step": 900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997983707006686e-05,
+      "loss": 0.6207,
+      "step": 901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997979180082508e-05,
+      "loss": 0.6546,
+      "step": 902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997974648087306e-05,
+      "loss": 0.6444,
+      "step": 903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997970111021102e-05,
+      "loss": 0.615,
+      "step": 904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979655688839202e-05,
+      "loss": 0.6509,
+      "step": 905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979610216757833e-05,
+      "loss": 0.655,
+      "step": 906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979564693967134e-05,
+      "loss": 0.6471,
+      "step": 907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997951912046735e-05,
+      "loss": 0.6352,
+      "step": 908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979473496258704e-05,
+      "loss": 0.6493,
+      "step": 909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997942782134143e-05,
+      "loss": 0.5688,
+      "step": 910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997938209571576e-05,
+      "loss": 0.6171,
+      "step": 911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979336319381934e-05,
+      "loss": 0.6119,
+      "step": 912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997929049234017e-05,
+      "loss": 0.6532,
+      "step": 913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979244614590712e-05,
+      "loss": 0.6013,
+      "step": 914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997919868613379e-05,
+      "loss": 0.6491,
+      "step": 915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979152706969634e-05,
+      "loss": 0.6555,
+      "step": 916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997910667709848e-05,
+      "loss": 0.6455,
+      "step": 917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979060596520565e-05,
+      "loss": 0.6207,
+      "step": 918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997901446523612e-05,
+      "loss": 0.5855,
+      "step": 919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997896828324538e-05,
+      "loss": 0.6473,
+      "step": 920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978922050548578e-05,
+      "loss": 0.6112,
+      "step": 921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978875767145954e-05,
+      "loss": 0.5978,
+      "step": 922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978829433037736e-05,
+      "loss": 0.5978,
+      "step": 923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978783048224167e-05,
+      "loss": 0.6514,
+      "step": 924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978736612705478e-05,
+      "loss": 0.6608,
+      "step": 925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978690126481902e-05,
+      "loss": 0.5893,
+      "step": 926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978643589553684e-05,
+      "loss": 0.6129,
+      "step": 927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978597001921054e-05,
+      "loss": 0.6465,
+      "step": 928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997855036358425e-05,
+      "loss": 0.5694,
+      "step": 929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978503674543507e-05,
+      "loss": 0.6536,
+      "step": 930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978456934799067e-05,
+      "loss": 0.6418,
+      "step": 931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978410144351164e-05,
+      "loss": 0.6751,
+      "step": 932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978363303200038e-05,
+      "loss": 0.5926,
+      "step": 933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978316411345927e-05,
+      "loss": 0.6089,
+      "step": 934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978269468789063e-05,
+      "loss": 0.676,
+      "step": 935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978222475529695e-05,
+      "loss": 0.628,
+      "step": 936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978175431568053e-05,
+      "loss": 0.6318,
+      "step": 937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978128336904384e-05,
+      "loss": 0.6122,
+      "step": 938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997808119153892e-05,
+      "loss": 0.6253,
+      "step": 939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978033995471898e-05,
+      "loss": 0.6119,
+      "step": 940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977986748703568e-05,
+      "loss": 0.645,
+      "step": 941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977939451234164e-05,
+      "loss": 0.6374,
+      "step": 942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997789210306393e-05,
+      "loss": 0.597,
+      "step": 943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99778447041931e-05,
+      "loss": 0.7084,
+      "step": 944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997779725462192e-05,
+      "loss": 0.7058,
+      "step": 945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977749754350634e-05,
+      "loss": 0.6472,
+      "step": 946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977702203379477e-05,
+      "loss": 0.6615,
+      "step": 947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997765460170869e-05,
+      "loss": 0.6453,
+      "step": 948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997760694933852e-05,
+      "loss": 0.6189,
+      "step": 949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977559246269206e-05,
+      "loss": 0.6118,
+      "step": 950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977511492500995e-05,
+      "loss": 0.6504,
+      "step": 951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977463688034125e-05,
+      "loss": 0.605,
+      "step": 952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997741583286884e-05,
+      "loss": 0.6189,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997736792700538e-05,
+      "loss": 0.6808,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977319970443992e-05,
+      "loss": 0.5877,
+      "step": 955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9977271963184922e-05,
+      "loss": 0.6461,
+      "step": 956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997722390522841e-05,
+      "loss": 0.6453,
+      "step": 957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99771757965747e-05,
+      "loss": 0.583,
+      "step": 958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997712763722404e-05,
+      "loss": 0.6894,
+      "step": 959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997707942717667e-05,
+      "loss": 0.6904,
+      "step": 960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997703116643284e-05,
+      "loss": 0.669,
+      "step": 961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976982854992793e-05,
+      "loss": 0.6361,
+      "step": 962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997693449285677e-05,
+      "loss": 0.6144,
+      "step": 963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976886080025027e-05,
+      "loss": 0.5947,
+      "step": 964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99768376164978e-05,
+      "loss": 0.5872,
+      "step": 965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997678910227534e-05,
+      "loss": 0.6245,
+      "step": 966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976740537357892e-05,
+      "loss": 0.6397,
+      "step": 967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976691921745705e-05,
+      "loss": 0.6801,
+      "step": 968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976643255439027e-05,
+      "loss": 0.6446,
+      "step": 969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976594538438097e-05,
+      "loss": 0.638,
+      "step": 970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997654577074317e-05,
+      "loss": 0.6083,
+      "step": 971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976496952354494e-05,
+      "loss": 0.6884,
+      "step": 972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976448083272316e-05,
+      "loss": 0.6108,
+      "step": 973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976399163496883e-05,
+      "loss": 0.5768,
+      "step": 974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997635019302844e-05,
+      "loss": 0.6811,
+      "step": 975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976301171867244e-05,
+      "loss": 0.6198,
+      "step": 976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997625210001354e-05,
+      "loss": 0.6348,
+      "step": 977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976202977467577e-05,
+      "loss": 0.5682,
+      "step": 978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976153804229603e-05,
+      "loss": 0.6514,
+      "step": 979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976104580299874e-05,
+      "loss": 0.6446,
+      "step": 980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9976055305678633e-05,
+      "loss": 0.6725,
+      "step": 981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997600598036613e-05,
+      "loss": 0.6072,
+      "step": 982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975956604362625e-05,
+      "loss": 0.602,
+      "step": 983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997590717766836e-05,
+      "loss": 0.6541,
+      "step": 984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997585770028359e-05,
+      "loss": 0.6443,
+      "step": 985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975808172208566e-05,
+      "loss": 0.6014,
+      "step": 986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997575859344354e-05,
+      "loss": 0.6984,
+      "step": 987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997570896398876e-05,
+      "loss": 0.6413,
+      "step": 988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975659283844485e-05,
+      "loss": 0.6728,
+      "step": 989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997560955301096e-05,
+      "loss": 0.6341,
+      "step": 990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997555977148844e-05,
+      "loss": 0.5745,
+      "step": 991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975509939277183e-05,
+      "loss": 0.5746,
+      "step": 992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975460056377438e-05,
+      "loss": 0.6168,
+      "step": 993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997541012278946e-05,
+      "loss": 0.5904,
+      "step": 994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975360138513498e-05,
+      "loss": 0.6598,
+      "step": 995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975310103549814e-05,
+      "loss": 0.6541,
+      "step": 996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975260017898655e-05,
+      "loss": 0.583,
+      "step": 997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997520988156028e-05,
+      "loss": 0.6255,
+      "step": 998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997515969453494e-05,
+      "loss": 0.6149,
+      "step": 999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975109456822896e-05,
+      "loss": 0.6384,
+      "step": 1000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975059168424396e-05,
+      "loss": 0.6201,
+      "step": 1001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9975008829339702e-05,
+      "loss": 0.6382,
+      "step": 1002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974958439569062e-05,
+      "loss": 0.6275,
+      "step": 1003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974907999112742e-05,
+      "loss": 0.5847,
+      "step": 1004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997485750797099e-05,
+      "loss": 0.6916,
+      "step": 1005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974806966144066e-05,
+      "loss": 0.6262,
+      "step": 1006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997475637363223e-05,
+      "loss": 0.6853,
+      "step": 1007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974705730435732e-05,
+      "loss": 0.6124,
+      "step": 1008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974655036554836e-05,
+      "loss": 0.6465,
+      "step": 1009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974604291989795e-05,
+      "loss": 0.6265,
+      "step": 1010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974553496740868e-05,
+      "loss": 0.6325,
+      "step": 1011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997450265080831e-05,
+      "loss": 0.6422,
+      "step": 1012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997445175419239e-05,
+      "loss": 0.5887,
+      "step": 1013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974400806893353e-05,
+      "loss": 0.6301,
+      "step": 1014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974349808911467e-05,
+      "loss": 0.6546,
+      "step": 1015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974298760246992e-05,
+      "loss": 0.6203,
+      "step": 1016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997424766090018e-05,
+      "loss": 0.6847,
+      "step": 1017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974196510871295e-05,
+      "loss": 0.688,
+      "step": 1018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974145310160598e-05,
+      "loss": 0.6288,
+      "step": 1019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974094058768345e-05,
+      "loss": 0.6411,
+      "step": 1020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9974042756694797e-05,
+      "loss": 0.6447,
+      "step": 1021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973991403940222e-05,
+      "loss": 0.5622,
+      "step": 1022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973940000504874e-05,
+      "loss": 0.6106,
+      "step": 1023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973888546389017e-05,
+      "loss": 0.6618,
+      "step": 1024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997383704159291e-05,
+      "loss": 0.6758,
+      "step": 1025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973785486116814e-05,
+      "loss": 0.6087,
+      "step": 1026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973733879960996e-05,
+      "loss": 0.5992,
+      "step": 1027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973682223125713e-05,
+      "loss": 0.5597,
+      "step": 1028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973630515611232e-05,
+      "loss": 0.6625,
+      "step": 1029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997357875741781e-05,
+      "loss": 0.6306,
+      "step": 1030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973526948545715e-05,
+      "loss": 0.6378,
+      "step": 1031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997347508899521e-05,
+      "loss": 0.634,
+      "step": 1032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973423178766558e-05,
+      "loss": 0.5836,
+      "step": 1033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997337121786002e-05,
+      "loss": 0.672,
+      "step": 1034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973319206275863e-05,
+      "loss": 0.662,
+      "step": 1035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997326714401435e-05,
+      "loss": 0.6386,
+      "step": 1036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973215031075746e-05,
+      "loss": 0.5958,
+      "step": 1037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973162867460316e-05,
+      "loss": 0.6284,
+      "step": 1038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973110653168323e-05,
+      "loss": 0.6697,
+      "step": 1039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973058388200033e-05,
+      "loss": 0.6538,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9973006072555717e-05,
+      "loss": 0.6595,
+      "step": 1041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972953706235632e-05,
+      "loss": 0.6563,
+      "step": 1042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997290128924005e-05,
+      "loss": 0.6063,
+      "step": 1043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972848821569237e-05,
+      "loss": 0.6452,
+      "step": 1044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997279630322346e-05,
+      "loss": 0.6252,
+      "step": 1045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997274373420298e-05,
+      "loss": 0.5664,
+      "step": 1046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997269111450807e-05,
+      "loss": 0.6895,
+      "step": 1047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972638444138995e-05,
+      "loss": 0.6395,
+      "step": 1048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972585723096026e-05,
+      "loss": 0.6298,
+      "step": 1049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972532951379427e-05,
+      "loss": 0.6832,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972480128989468e-05,
+      "loss": 0.5916,
+      "step": 1051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972427255926417e-05,
+      "loss": 0.5873,
+      "step": 1052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972374332190544e-05,
+      "loss": 0.6275,
+      "step": 1053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972321357782115e-05,
+      "loss": 0.5494,
+      "step": 1054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972268332701397e-05,
+      "loss": 0.6589,
+      "step": 1055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972215256948667e-05,
+      "loss": 0.6142,
+      "step": 1056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997216213052419e-05,
+      "loss": 0.6673,
+      "step": 1057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972108953428234e-05,
+      "loss": 0.6646,
+      "step": 1058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9972055725661076e-05,
+      "loss": 0.5857,
+      "step": 1059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997200244722298e-05,
+      "loss": 0.6425,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971949118114223e-05,
+      "loss": 0.6078,
+      "step": 1061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971895738335066e-05,
+      "loss": 0.6265,
+      "step": 1062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971842307885793e-05,
+      "loss": 0.6424,
+      "step": 1063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971788826766665e-05,
+      "loss": 0.6682,
+      "step": 1064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997173529497796e-05,
+      "loss": 0.6206,
+      "step": 1065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971681712519943e-05,
+      "loss": 0.6135,
+      "step": 1066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971628079392896e-05,
+      "loss": 0.637,
+      "step": 1067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971574395597084e-05,
+      "loss": 0.6205,
+      "step": 1068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971520661132778e-05,
+      "loss": 0.6033,
+      "step": 1069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997146687600026e-05,
+      "loss": 0.5887,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99714130401998e-05,
+      "loss": 0.6371,
+      "step": 1071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997135915373167e-05,
+      "loss": 0.6635,
+      "step": 1072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971305216596143e-05,
+      "loss": 0.6604,
+      "step": 1073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997125122879349e-05,
+      "loss": 0.6423,
+      "step": 1074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997119719032399e-05,
+      "loss": 0.5988,
+      "step": 1075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971143101187918e-05,
+      "loss": 0.6553,
+      "step": 1076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9971088961385553e-05,
+      "loss": 0.6168,
+      "step": 1077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997103477091716e-05,
+      "loss": 0.6039,
+      "step": 1078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997098052978302e-05,
+      "loss": 0.6255,
+      "step": 1079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970926237983407e-05,
+      "loss": 0.6112,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970871895518597e-05,
+      "loss": 0.6605,
+      "step": 1081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997081750238887e-05,
+      "loss": 0.6415,
+      "step": 1082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970763058594498e-05,
+      "loss": 0.6161,
+      "step": 1083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970708564135758e-05,
+      "loss": 0.6064,
+      "step": 1084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970654019012924e-05,
+      "loss": 0.6239,
+      "step": 1085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970599423226282e-05,
+      "loss": 0.6169,
+      "step": 1086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970544776776103e-05,
+      "loss": 0.6234,
+      "step": 1087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970490079662667e-05,
+      "loss": 0.5561,
+      "step": 1088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970435331886252e-05,
+      "loss": 0.6082,
+      "step": 1089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.997038053344713e-05,
+      "loss": 0.5567,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970325684345585e-05,
+      "loss": 0.6381,
+      "step": 1091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99702707845819e-05,
+      "loss": 0.6351,
+      "step": 1092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970215834156347e-05,
+      "loss": 0.6612,
+      "step": 1093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970160833069207e-05,
+      "loss": 0.6288,
+      "step": 1094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970105781320762e-05,
+      "loss": 0.6767,
+      "step": 1095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9970050678911287e-05,
+      "loss": 0.6519,
+      "step": 1096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969995525841066e-05,
+      "loss": 0.617,
+      "step": 1097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969940322110376e-05,
+      "loss": 0.685,
+      "step": 1098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969885067719504e-05,
+      "loss": 0.6083,
+      "step": 1099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.996982976266872e-05,
+      "loss": 0.6312,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969774406958316e-05,
+      "loss": 0.6504,
+      "step": 1101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9969719000588564e-05,
+      "loss": 0.6162,
+      "step": 1102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969663543559753e-05,
+      "loss": 0.6311,
+      "step": 1103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996960803587216e-05,
+      "loss": 0.6261,
+      "step": 1104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969552477526074e-05,
+      "loss": 0.6518,
+      "step": 1105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969496868521768e-05,
+      "loss": 0.6532,
+      "step": 1106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996944120885953e-05,
+      "loss": 0.6634,
+      "step": 1107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996938549853964e-05,
+      "loss": 0.6097,
+      "step": 1108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969329737562383e-05,
+      "loss": 0.7059,
+      "step": 1109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996927392592804e-05,
+      "loss": 0.654,
+      "step": 1110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969218063636902e-05,
+      "loss": 0.6617,
+      "step": 1111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969162150689243e-05,
+      "loss": 0.6511,
+      "step": 1112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996910618708535e-05,
+      "loss": 0.5583,
+      "step": 1113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9969050172825514e-05,
+      "loss": 0.6303,
+      "step": 1114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996899410791001e-05,
+      "loss": 0.6485,
+      "step": 1115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996893799233913e-05,
+      "loss": 0.6298,
+      "step": 1116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968881826113157e-05,
+      "loss": 0.6351,
+      "step": 1117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968825609232374e-05,
+      "loss": 0.6675,
+      "step": 1118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968769341697068e-05,
+      "loss": 0.7015,
+      "step": 1119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996871302350753e-05,
+      "loss": 0.671,
+      "step": 1120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968656654664037e-05,
+      "loss": 0.641,
+      "step": 1121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996860023516688e-05,
+      "loss": 0.5785,
+      "step": 1122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968543765016347e-05,
+      "loss": 0.5679,
+      "step": 1123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968487244212726e-05,
+      "loss": 0.5878,
+      "step": 1124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.99684306727563e-05,
+      "loss": 0.6431,
+      "step": 1125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968374050647358e-05,
+      "loss": 0.6827,
+      "step": 1126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968317377886192e-05,
+      "loss": 0.5994,
+      "step": 1127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996826065447308e-05,
+      "loss": 0.6519,
+      "step": 1128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968203880408317e-05,
+      "loss": 0.6416,
+      "step": 1129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968147055692194e-05,
+      "loss": 0.6969,
+      "step": 1130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968090180324995e-05,
+      "loss": 0.6315,
+      "step": 1131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9968033254307012e-05,
+      "loss": 0.6299,
+      "step": 1132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996797627763853e-05,
+      "loss": 0.5891,
+      "step": 1133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996791925031984e-05,
+      "loss": 0.6243,
+      "step": 1134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967862172351237e-05,
+      "loss": 0.7011,
+      "step": 1135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967805043733005e-05,
+      "loss": 0.5708,
+      "step": 1136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967747864465438e-05,
+      "loss": 0.6312,
+      "step": 1137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967690634548825e-05,
+      "loss": 0.6683,
+      "step": 1138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967633353983456e-05,
+      "loss": 0.6586,
+      "step": 1139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996757602276962e-05,
+      "loss": 0.6419,
+      "step": 1140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967518640907616e-05,
+      "loss": 0.6529,
+      "step": 1141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967461208397727e-05,
+      "loss": 0.6108,
+      "step": 1142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996740372524025e-05,
+      "loss": 0.6917,
+      "step": 1143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967346191435474e-05,
+      "loss": 0.6335,
+      "step": 1144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967288606983696e-05,
+      "loss": 0.6844,
+      "step": 1145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967230971885203e-05,
+      "loss": 0.6061,
+      "step": 1146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996717328614029e-05,
+      "loss": 0.5947,
+      "step": 1147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967115549749253e-05,
+      "loss": 0.5905,
+      "step": 1148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9967057762712382e-05,
+      "loss": 0.6518,
+      "step": 1149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996699992502997e-05,
+      "loss": 0.6373,
+      "step": 1150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966942036702312e-05,
+      "loss": 0.6491,
+      "step": 1151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966884097729706e-05,
+      "loss": 0.6072,
+      "step": 1152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996682610811244e-05,
+      "loss": 0.6283,
+      "step": 1153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966768067850814e-05,
+      "loss": 0.606,
+      "step": 1154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996670997694512e-05,
+      "loss": 0.6271,
+      "step": 1155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966651835395652e-05,
+      "loss": 0.6194,
+      "step": 1156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996659364320271e-05,
+      "loss": 0.5944,
+      "step": 1157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966535400366584e-05,
+      "loss": 0.6141,
+      "step": 1158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966477106887572e-05,
+      "loss": 0.647,
+      "step": 1159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996641876276597e-05,
+      "loss": 0.6022,
+      "step": 1160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996636036800208e-05,
+      "loss": 0.6461,
+      "step": 1161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966301922596192e-05,
+      "loss": 0.6554,
+      "step": 1162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966243426548607e-05,
+      "loss": 0.5926,
+      "step": 1163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996618487985962e-05,
+      "loss": 0.6561,
+      "step": 1164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966126282529526e-05,
+      "loss": 0.6199,
+      "step": 1165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966067634558627e-05,
+      "loss": 0.6408,
+      "step": 1166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9966008935947222e-05,
+      "loss": 0.6209,
+      "step": 1167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965950186695604e-05,
+      "loss": 0.5876,
+      "step": 1168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965891386804074e-05,
+      "loss": 0.6218,
+      "step": 1169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965832536272932e-05,
+      "loss": 0.5939,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965773635102475e-05,
+      "loss": 0.6147,
+      "step": 1171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965714683293007e-05,
+      "loss": 0.5786,
+      "step": 1172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965655680844823e-05,
+      "loss": 0.6537,
+      "step": 1173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965596627758225e-05,
+      "loss": 0.7342,
+      "step": 1174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996553752403351e-05,
+      "loss": 0.6795,
+      "step": 1175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965478369670983e-05,
+      "loss": 0.6119,
+      "step": 1176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965419164670936e-05,
+      "loss": 0.5987,
+      "step": 1177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996535990903368e-05,
+      "loss": 0.6568,
+      "step": 1178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996530060275951e-05,
+      "loss": 0.6253,
+      "step": 1179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965241245848733e-05,
+      "loss": 0.6448,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996518183830164e-05,
+      "loss": 0.5753,
+      "step": 1181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965122380118547e-05,
+      "loss": 0.6173,
+      "step": 1182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9965062871299746e-05,
+      "loss": 0.5855,
+      "step": 1183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996500331184554e-05,
+      "loss": 0.6884,
+      "step": 1184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964943701756236e-05,
+      "loss": 0.5758,
+      "step": 1185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964884041032135e-05,
+      "loss": 0.6186,
+      "step": 1186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996482432967354e-05,
+      "loss": 0.6317,
+      "step": 1187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996476456768075e-05,
+      "loss": 0.6464,
+      "step": 1188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964704755054076e-05,
+      "loss": 0.5954,
+      "step": 1189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996464489179382e-05,
+      "loss": 0.5893,
+      "step": 1190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964584977900284e-05,
+      "loss": 0.6329,
+      "step": 1191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964525013373773e-05,
+      "loss": 0.6258,
+      "step": 1192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964464998214593e-05,
+      "loss": 0.6166,
+      "step": 1193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964404932423046e-05,
+      "loss": 0.7209,
+      "step": 1194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964344815999444e-05,
+      "loss": 0.5365,
+      "step": 1195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964284648944085e-05,
+      "loss": 0.5867,
+      "step": 1196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964224431257277e-05,
+      "loss": 0.6099,
+      "step": 1197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964164162939326e-05,
+      "loss": 0.6135,
+      "step": 1198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996410384399054e-05,
+      "loss": 0.6384,
+      "step": 1199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9964043474411222e-05,
+      "loss": 0.6402,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963983054201687e-05,
+      "loss": 0.608,
+      "step": 1201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996392258336223e-05,
+      "loss": 0.6058,
+      "step": 1202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963862061893165e-05,
+      "loss": 0.6567,
+      "step": 1203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.99638014897948e-05,
+      "loss": 0.6511,
+      "step": 1204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996374086706744e-05,
+      "loss": 0.7084,
+      "step": 1205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.99636801937114e-05,
+      "loss": 0.6702,
+      "step": 1206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963619469726974e-05,
+      "loss": 0.6055,
+      "step": 1207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963558695114486e-05,
+      "loss": 0.6448,
+      "step": 1208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996349786987424e-05,
+      "loss": 0.6248,
+      "step": 1209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996343699400654e-05,
+      "loss": 0.6104,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963376067511698e-05,
+      "loss": 0.605,
+      "step": 1211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963315090390023e-05,
+      "loss": 0.6347,
+      "step": 1212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996325406264183e-05,
+      "loss": 0.6844,
+      "step": 1213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963192984267423e-05,
+      "loss": 0.6699,
+      "step": 1214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963131855267117e-05,
+      "loss": 0.6665,
+      "step": 1215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9963070675641217e-05,
+      "loss": 0.5819,
+      "step": 1216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996300944539004e-05,
+      "loss": 0.6939,
+      "step": 1217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996294816451389e-05,
+      "loss": 0.6303,
+      "step": 1218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962886833013085e-05,
+      "loss": 0.6275,
+      "step": 1219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962825450887936e-05,
+      "loss": 0.6272,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962764018138753e-05,
+      "loss": 0.6404,
+      "step": 1221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962702534765845e-05,
+      "loss": 0.5946,
+      "step": 1222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962641000769533e-05,
+      "loss": 0.6931,
+      "step": 1223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996257941615012e-05,
+      "loss": 0.599,
+      "step": 1224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962517780907925e-05,
+      "loss": 0.6308,
+      "step": 1225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962456095043257e-05,
+      "loss": 0.6471,
+      "step": 1226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962394358556433e-05,
+      "loss": 0.5908,
+      "step": 1227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962332571447767e-05,
+      "loss": 0.5929,
+      "step": 1228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996227073371757e-05,
+      "loss": 0.648,
+      "step": 1229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996220884536616e-05,
+      "loss": 0.6499,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9962146906393845e-05,
+      "loss": 0.6532,
+      "step": 1231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996208491680095e-05,
+      "loss": 0.642,
+      "step": 1232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996202287658778e-05,
+      "loss": 0.569,
+      "step": 1233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961960785754657e-05,
+      "loss": 0.6195,
+      "step": 1234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996189864430189e-05,
+      "loss": 0.5843,
+      "step": 1235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961836452229804e-05,
+      "loss": 0.6514,
+      "step": 1236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961774209538704e-05,
+      "loss": 0.6333,
+      "step": 1237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961711916228914e-05,
+      "loss": 0.6235,
+      "step": 1238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961649572300748e-05,
+      "loss": 0.5989,
+      "step": 1239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961587177754526e-05,
+      "loss": 0.6528,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961524732590557e-05,
+      "loss": 0.6321,
+      "step": 1241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961462236809166e-05,
+      "loss": 0.6091,
+      "step": 1242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996139969041067e-05,
+      "loss": 0.5852,
+      "step": 1243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961337093395384e-05,
+      "loss": 0.6065,
+      "step": 1244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961274445763625e-05,
+      "loss": 0.6427,
+      "step": 1245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961211747515712e-05,
+      "loss": 0.6753,
+      "step": 1246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961148998651966e-05,
+      "loss": 0.6952,
+      "step": 1247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.996108619917271e-05,
+      "loss": 0.6402,
+      "step": 1248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9961023349078253e-05,
+      "loss": 0.6634,
+      "step": 1249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960960448368916e-05,
+      "loss": 0.5673,
+      "step": 1250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960897497045028e-05,
+      "loss": 0.7047,
+      "step": 1251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.99608344951069e-05,
+      "loss": 0.6281,
+      "step": 1252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960771442554855e-05,
+      "loss": 0.6936,
+      "step": 1253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.996070833938921e-05,
+      "loss": 0.6384,
+      "step": 1254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960645185610296e-05,
+      "loss": 0.5872,
+      "step": 1255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960581981218417e-05,
+      "loss": 0.5951,
+      "step": 1256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960518726213915e-05,
+      "loss": 0.636,
+      "step": 1257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.996045542059709e-05,
+      "loss": 0.6278,
+      "step": 1258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960392064368278e-05,
+      "loss": 0.6093,
+      "step": 1259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.99603286575278e-05,
+      "loss": 0.6981,
+      "step": 1260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960265200075972e-05,
+      "loss": 0.6512,
+      "step": 1261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.996020169201312e-05,
+      "loss": 0.6733,
+      "step": 1262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960138133339567e-05,
+      "loss": 0.6445,
+      "step": 1263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960074524055633e-05,
+      "loss": 0.6439,
+      "step": 1264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9960010864161647e-05,
+      "loss": 0.6376,
+      "step": 1265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959947153657926e-05,
+      "loss": 0.6044,
+      "step": 1266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959883392544794e-05,
+      "loss": 0.6132,
+      "step": 1267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995981958082258e-05,
+      "loss": 0.6378,
+      "step": 1268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995975571849161e-05,
+      "loss": 0.5832,
+      "step": 1269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.99596918055522e-05,
+      "loss": 0.6364,
+      "step": 1270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995962784200468e-05,
+      "loss": 0.6223,
+      "step": 1271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959563827849372e-05,
+      "loss": 0.6192,
+      "step": 1272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959499763086607e-05,
+      "loss": 0.696,
+      "step": 1273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959435647716706e-05,
+      "loss": 0.6734,
+      "step": 1274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959371481739995e-05,
+      "loss": 0.5874,
+      "step": 1275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959307265156803e-05,
+      "loss": 0.6257,
+      "step": 1276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995924299796745e-05,
+      "loss": 0.6917,
+      "step": 1277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995917868017227e-05,
+      "loss": 0.6184,
+      "step": 1278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959114311771586e-05,
+      "loss": 0.624,
+      "step": 1279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9959049892765724e-05,
+      "loss": 0.6709,
+      "step": 1280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958985423155013e-05,
+      "loss": 0.6879,
+      "step": 1281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958920902939782e-05,
+      "loss": 0.6303,
+      "step": 1282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958856332120354e-05,
+      "loss": 0.6742,
+      "step": 1283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958791710697067e-05,
+      "loss": 0.6132,
+      "step": 1284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958727038670235e-05,
+      "loss": 0.6518,
+      "step": 1285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.99586623160402e-05,
+      "loss": 0.6599,
+      "step": 1286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958597542807282e-05,
+      "loss": 0.6692,
+      "step": 1287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995853271897181e-05,
+      "loss": 0.6357,
+      "step": 1288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958467844534124e-05,
+      "loss": 0.6866,
+      "step": 1289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958402919494543e-05,
+      "loss": 0.529,
+      "step": 1290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958337943853404e-05,
+      "loss": 0.6042,
+      "step": 1291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958272917611027e-05,
+      "loss": 0.6353,
+      "step": 1292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958207840767753e-05,
+      "loss": 0.621,
+      "step": 1293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995814271332391e-05,
+      "loss": 0.6262,
+      "step": 1294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9958077535279827e-05,
+      "loss": 0.6732,
+      "step": 1295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995801230663583e-05,
+      "loss": 0.6128,
+      "step": 1296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957947027392262e-05,
+      "loss": 0.62,
+      "step": 1297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957881697549446e-05,
+      "loss": 0.6684,
+      "step": 1298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995781631710772e-05,
+      "loss": 0.6406,
+      "step": 1299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957750886067407e-05,
+      "loss": 0.6408,
+      "step": 1300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957685404428853e-05,
+      "loss": 0.599,
+      "step": 1301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995761987219238e-05,
+      "loss": 0.5855,
+      "step": 1302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957554289358324e-05,
+      "loss": 0.7219,
+      "step": 1303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957488655927014e-05,
+      "loss": 0.6141,
+      "step": 1304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957422971898794e-05,
+      "loss": 0.6353,
+      "step": 1305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995735723727399e-05,
+      "loss": 0.6193,
+      "step": 1306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957291452052937e-05,
+      "loss": 0.6244,
+      "step": 1307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995722561623597e-05,
+      "loss": 0.6144,
+      "step": 1308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957159729823425e-05,
+      "loss": 0.6648,
+      "step": 1309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957093792815635e-05,
+      "loss": 0.6029,
+      "step": 1310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9957027805212934e-05,
+      "loss": 0.6558,
+      "step": 1311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956961767015658e-05,
+      "loss": 0.673,
+      "step": 1312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956895678224148e-05,
+      "loss": 0.6218,
+      "step": 1313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995682953883873e-05,
+      "loss": 0.6502,
+      "step": 1314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956763348859746e-05,
+      "loss": 0.6484,
+      "step": 1315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995669710828753e-05,
+      "loss": 0.7129,
+      "step": 1316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956630817122425e-05,
+      "loss": 0.6523,
+      "step": 1317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995656447536476e-05,
+      "loss": 0.5823,
+      "step": 1318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956498083014876e-05,
+      "loss": 0.6673,
+      "step": 1319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956431640073107e-05,
+      "loss": 0.6119,
+      "step": 1320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956365146539795e-05,
+      "loss": 0.6171,
+      "step": 1321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956298602415275e-05,
+      "loss": 0.6392,
+      "step": 1322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956232007699887e-05,
+      "loss": 0.5855,
+      "step": 1323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956165362393968e-05,
+      "loss": 0.6361,
+      "step": 1324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9956098666497857e-05,
+      "loss": 0.64,
+      "step": 1325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995603192001189e-05,
+      "loss": 0.6491,
+      "step": 1326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9955965122936415e-05,
+      "loss": 0.6225,
+      "step": 1327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995589827527176e-05,
+      "loss": 0.6187,
+      "step": 1328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9955831377018272e-05,
+      "loss": 0.5566,
+      "step": 1329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995576442817629e-05,
+      "loss": 0.69,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995569742874615e-05,
+      "loss": 0.6353,
+      "step": 1331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.99556303787282e-05,
+      "loss": 0.6868,
+      "step": 1332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9955563278122775e-05,
+      "loss": 0.6318,
+      "step": 1333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995549612693022e-05,
+      "loss": 0.637,
+      "step": 1334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9955428925150868e-05,
+      "loss": 0.6049,
+      "step": 1335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995536167278507e-05,
+      "loss": 0.6577,
+      "step": 1336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995529436983316e-05,
+      "loss": 0.6509,
+      "step": 1337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9955227016295486e-05,
+      "loss": 0.5681,
+      "step": 1338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995515961217239e-05,
+      "loss": 0.6871,
+      "step": 1339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995509215746421e-05,
+      "loss": 0.6905,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995502465217129e-05,
+      "loss": 0.5895,
+      "step": 1341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954957096293975e-05,
+      "loss": 0.6775,
+      "step": 1342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954889489832612e-05,
+      "loss": 0.5959,
+      "step": 1343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954821832787535e-05,
+      "loss": 0.6422,
+      "step": 1344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954754125159094e-05,
+      "loss": 0.6869,
+      "step": 1345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954686366947635e-05,
+      "loss": 0.6603,
+      "step": 1346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954618558153495e-05,
+      "loss": 0.6961,
+      "step": 1347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954550698777028e-05,
+      "loss": 0.6094,
+      "step": 1348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995448278881857e-05,
+      "loss": 0.6721,
+      "step": 1349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954414828278467e-05,
+      "loss": 0.6713,
+      "step": 1350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995434681715707e-05,
+      "loss": 0.617,
+      "step": 1351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954278755454723e-05,
+      "loss": 0.6296,
+      "step": 1352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954210643171768e-05,
+      "loss": 0.6576,
+      "step": 1353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954142480308553e-05,
+      "loss": 0.6015,
+      "step": 1354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995407426686543e-05,
+      "loss": 0.5826,
+      "step": 1355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9954006002842733e-05,
+      "loss": 0.6415,
+      "step": 1356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953937688240822e-05,
+      "loss": 0.656,
+      "step": 1357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953869323060037e-05,
+      "loss": 0.6403,
+      "step": 1358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953800907300726e-05,
+      "loss": 0.6723,
+      "step": 1359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953732440963237e-05,
+      "loss": 0.6042,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953663924047918e-05,
+      "loss": 0.6304,
+      "step": 1361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995359535655512e-05,
+      "loss": 0.6696,
+      "step": 1362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953526738485186e-05,
+      "loss": 0.6404,
+      "step": 1363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995345806983847e-05,
+      "loss": 0.6444,
+      "step": 1364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953389350615317e-05,
+      "loss": 0.6757,
+      "step": 1365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953320580816078e-05,
+      "loss": 0.6408,
+      "step": 1366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953251760441105e-05,
+      "loss": 0.7028,
+      "step": 1367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995318288949074e-05,
+      "loss": 0.6013,
+      "step": 1368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995311396796534e-05,
+      "loss": 0.6737,
+      "step": 1369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9953044995865253e-05,
+      "loss": 0.6207,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952975973190832e-05,
+      "loss": 0.5978,
+      "step": 1371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952906899942423e-05,
+      "loss": 0.6688,
+      "step": 1372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952837776120377e-05,
+      "loss": 0.5938,
+      "step": 1373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995276860172505e-05,
+      "loss": 0.6227,
+      "step": 1374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995269937675679e-05,
+      "loss": 0.5948,
+      "step": 1375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995263010121595e-05,
+      "loss": 0.6495,
+      "step": 1376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952560775102882e-05,
+      "loss": 0.6365,
+      "step": 1377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995249139841794e-05,
+      "loss": 0.6359,
+      "step": 1378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995242197116147e-05,
+      "loss": 0.6225,
+      "step": 1379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952352493333833e-05,
+      "loss": 0.7284,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952282964935376e-05,
+      "loss": 0.5468,
+      "step": 1381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952213385966453e-05,
+      "loss": 0.6629,
+      "step": 1382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952143756427425e-05,
+      "loss": 0.6151,
+      "step": 1383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952074076318638e-05,
+      "loss": 0.68,
+      "step": 1384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9952004345640446e-05,
+      "loss": 0.6319,
+      "step": 1385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951934564393205e-05,
+      "loss": 0.6238,
+      "step": 1386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995186473257727e-05,
+      "loss": 0.6171,
+      "step": 1387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951794850192998e-05,
+      "loss": 0.6801,
+      "step": 1388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995172491724074e-05,
+      "loss": 0.5269,
+      "step": 1389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951654933720855e-05,
+      "loss": 0.6533,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951584899633697e-05,
+      "loss": 0.6139,
+      "step": 1391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951514814979617e-05,
+      "loss": 0.6191,
+      "step": 1392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.995144467975898e-05,
+      "loss": 0.5725,
+      "step": 1393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951374493972134e-05,
+      "loss": 0.5933,
+      "step": 1394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9951304257619447e-05,
+      "loss": 0.6356,
+      "step": 1395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9951233970701262e-05,
+      "loss": 0.6179,
+      "step": 1396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9951163633217943e-05,
+      "loss": 0.6142,
+      "step": 1397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995109324516985e-05,
+      "loss": 0.5917,
+      "step": 1398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9951022806557342e-05,
+      "loss": 0.622,
+      "step": 1399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950952317380767e-05,
+      "loss": 0.5835,
+      "step": 1400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995088177764049e-05,
+      "loss": 0.6559,
+      "step": 1401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995081118733687e-05,
+      "loss": 0.6228,
+      "step": 1402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950740546470263e-05,
+      "loss": 0.5848,
+      "step": 1403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995066985504103e-05,
+      "loss": 0.6092,
+      "step": 1404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950599113049526e-05,
+      "loss": 0.5945,
+      "step": 1405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950528320496117e-05,
+      "loss": 0.701,
+      "step": 1406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995045747738116e-05,
+      "loss": 0.6253,
+      "step": 1407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.995038658370501e-05,
+      "loss": 0.6708,
+      "step": 1408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950315639468037e-05,
+      "loss": 0.6602,
+      "step": 1409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950244644670594e-05,
+      "loss": 0.5995,
+      "step": 1410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950173599313043e-05,
+      "loss": 0.6208,
+      "step": 1411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950102503395746e-05,
+      "loss": 0.6478,
+      "step": 1412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9950031356919062e-05,
+      "loss": 0.5922,
+      "step": 1413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949960159883362e-05,
+      "loss": 0.5994,
+      "step": 1414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949888912288996e-05,
+      "loss": 0.6685,
+      "step": 1415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994981761413633e-05,
+      "loss": 0.6164,
+      "step": 1416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949746265425728e-05,
+      "loss": 0.6537,
+      "step": 1417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994967486615755e-05,
+      "loss": 0.5923,
+      "step": 1418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949603416332158e-05,
+      "loss": 0.7082,
+      "step": 1419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994953191594992e-05,
+      "loss": 0.6417,
+      "step": 1420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949460365011196e-05,
+      "loss": 0.6265,
+      "step": 1421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949388763516352e-05,
+      "loss": 0.6066,
+      "step": 1422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949317111465747e-05,
+      "loss": 0.6499,
+      "step": 1423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994924540885975e-05,
+      "loss": 0.6317,
+      "step": 1424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949173655698723e-05,
+      "loss": 0.5905,
+      "step": 1425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994910185198303e-05,
+      "loss": 0.6618,
+      "step": 1426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9949029997713035e-05,
+      "loss": 0.6056,
+      "step": 1427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994895809288911e-05,
+      "loss": 0.5747,
+      "step": 1428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994888613751161e-05,
+      "loss": 0.6315,
+      "step": 1429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948814131580906e-05,
+      "loss": 0.597,
+      "step": 1430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948742075097368e-05,
+      "loss": 0.6456,
+      "step": 1431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948669968061354e-05,
+      "loss": 0.6173,
+      "step": 1432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994859781047324e-05,
+      "loss": 0.6408,
+      "step": 1433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994852560233338e-05,
+      "loss": 0.6087,
+      "step": 1434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994845334364215e-05,
+      "loss": 0.6805,
+      "step": 1435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948381034399914e-05,
+      "loss": 0.66,
+      "step": 1436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948308674607043e-05,
+      "loss": 0.6336,
+      "step": 1437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948236264263903e-05,
+      "loss": 0.61,
+      "step": 1438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994816380337086e-05,
+      "loss": 0.646,
+      "step": 1439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994809129192828e-05,
+      "loss": 0.666,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9948018729936536e-05,
+      "loss": 0.582,
+      "step": 1441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947946117396e-05,
+      "loss": 0.6578,
+      "step": 1442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947873454307035e-05,
+      "loss": 0.5841,
+      "step": 1443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947800740670006e-05,
+      "loss": 0.6826,
+      "step": 1444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947727976485293e-05,
+      "loss": 0.6282,
+      "step": 1445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994765516175326e-05,
+      "loss": 0.6453,
+      "step": 1446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994758229647428e-05,
+      "loss": 0.6456,
+      "step": 1447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947509380648715e-05,
+      "loss": 0.6432,
+      "step": 1448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947436414276947e-05,
+      "loss": 0.5862,
+      "step": 1449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947363397359338e-05,
+      "loss": 0.6445,
+      "step": 1450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947290329896264e-05,
+      "loss": 0.6482,
+      "step": 1451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947217211888095e-05,
+      "loss": 0.5913,
+      "step": 1452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947144043335203e-05,
+      "loss": 0.6416,
+      "step": 1453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9947070824237958e-05,
+      "loss": 0.6909,
+      "step": 1454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994699755459673e-05,
+      "loss": 0.6118,
+      "step": 1455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.99469242344119e-05,
+      "loss": 0.672,
+      "step": 1456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994685086368383e-05,
+      "loss": 0.5781,
+      "step": 1457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.99467774424129e-05,
+      "loss": 0.6537,
+      "step": 1458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994670397059948e-05,
+      "loss": 0.6259,
+      "step": 1459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9946630448243946e-05,
+      "loss": 0.6875,
+      "step": 1460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9946556875346666e-05,
+      "loss": 0.6142,
+      "step": 1461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994648325190802e-05,
+      "loss": 0.6426,
+      "step": 1462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994640957792838e-05,
+      "loss": 0.662,
+      "step": 1463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994633585340812e-05,
+      "loss": 0.6626,
+      "step": 1464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9946262078347612e-05,
+      "loss": 0.5907,
+      "step": 1465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9946188252747235e-05,
+      "loss": 0.6177,
+      "step": 1466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9946114376607362e-05,
+      "loss": 0.6386,
+      "step": 1467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994604044992837e-05,
+      "loss": 0.6206,
+      "step": 1468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945966472710635e-05,
+      "loss": 0.6616,
+      "step": 1469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945892444954528e-05,
+      "loss": 0.5933,
+      "step": 1470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945818366660435e-05,
+      "loss": 0.6414,
+      "step": 1471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994574423782872e-05,
+      "loss": 0.6234,
+      "step": 1472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945670058459767e-05,
+      "loss": 0.6268,
+      "step": 1473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945595828553955e-05,
+      "loss": 0.6075,
+      "step": 1474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945521548111654e-05,
+      "loss": 0.6459,
+      "step": 1475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945447217133246e-05,
+      "loss": 0.6837,
+      "step": 1476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945372835619108e-05,
+      "loss": 0.5928,
+      "step": 1477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994529840356962e-05,
+      "loss": 0.6055,
+      "step": 1478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945223920985154e-05,
+      "loss": 0.6552,
+      "step": 1479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9945149387866097e-05,
+      "loss": 0.6509,
+      "step": 1480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994507480421282e-05,
+      "loss": 0.6106,
+      "step": 1481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994500017002571e-05,
+      "loss": 0.6591,
+      "step": 1482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944925485305136e-05,
+      "loss": 0.6278,
+      "step": 1483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994485075005149e-05,
+      "loss": 0.6722,
+      "step": 1484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994477596426514e-05,
+      "loss": 0.6543,
+      "step": 1485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994470112794647e-05,
+      "loss": 0.5966,
+      "step": 1486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944626241095863e-05,
+      "loss": 0.7532,
+      "step": 1487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944551303713697e-05,
+      "loss": 0.6664,
+      "step": 1488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944476315800357e-05,
+      "loss": 0.6379,
+      "step": 1489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944401277356213e-05,
+      "loss": 0.5555,
+      "step": 1490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944326188381658e-05,
+      "loss": 0.6305,
+      "step": 1491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994425104887707e-05,
+      "loss": 0.6469,
+      "step": 1492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944175858842832e-05,
+      "loss": 0.6497,
+      "step": 1493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994410061827932e-05,
+      "loss": 0.6408,
+      "step": 1494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9944025327186926e-05,
+      "loss": 0.6217,
+      "step": 1495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994394998556602e-05,
+      "loss": 0.7058,
+      "step": 1496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943874593416998e-05,
+      "loss": 0.678,
+      "step": 1497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943799150740234e-05,
+      "loss": 0.6127,
+      "step": 1498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994372365753611e-05,
+      "loss": 0.6918,
+      "step": 1499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994364811380502e-05,
+      "loss": 0.6239,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994357251954734e-05,
+      "loss": 0.6481,
+      "step": 1501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943496874763456e-05,
+      "loss": 0.716,
+      "step": 1502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943421179453756e-05,
+      "loss": 0.592,
+      "step": 1503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943345433618615e-05,
+      "loss": 0.5814,
+      "step": 1504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9943269637258426e-05,
+      "loss": 0.6298,
+      "step": 1505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994319379037357e-05,
+      "loss": 0.6043,
+      "step": 1506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994311789296444e-05,
+      "loss": 0.6126,
+      "step": 1507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994304194503141e-05,
+      "loss": 0.6471,
+      "step": 1508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942965946574876e-05,
+      "loss": 0.587,
+      "step": 1509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942889897595215e-05,
+      "loss": 0.6366,
+      "step": 1510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942813798092824e-05,
+      "loss": 0.6656,
+      "step": 1511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942737648068085e-05,
+      "loss": 0.658,
+      "step": 1512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942661447521382e-05,
+      "loss": 0.6108,
+      "step": 1513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.99425851964531e-05,
+      "loss": 0.6589,
+      "step": 1514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942508894863633e-05,
+      "loss": 0.6252,
+      "step": 1515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994243254275337e-05,
+      "loss": 0.6385,
+      "step": 1516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942356140122692e-05,
+      "loss": 0.6523,
+      "step": 1517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9942279686971994e-05,
+      "loss": 0.6476,
+      "step": 1518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994220318330166e-05,
+      "loss": 0.6187,
+      "step": 1519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994212662911208e-05,
+      "loss": 0.6351,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994205002440364e-05,
+      "loss": 0.6024,
+      "step": 1521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941973369176738e-05,
+      "loss": 0.6812,
+      "step": 1522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994189666343175e-05,
+      "loss": 0.6157,
+      "step": 1523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994181990716908e-05,
+      "loss": 0.6469,
+      "step": 1524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994174310038911e-05,
+      "loss": 0.6664,
+      "step": 1525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941666243092228e-05,
+      "loss": 0.5689,
+      "step": 1526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994158933527883e-05,
+      "loss": 0.567,
+      "step": 1527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941512376949304e-05,
+      "loss": 0.6628,
+      "step": 1528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941435368104047e-05,
+      "loss": 0.63,
+      "step": 1529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941358308743442e-05,
+      "loss": 0.6547,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941281198867886e-05,
+      "loss": 0.6954,
+      "step": 1531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941204038477765e-05,
+      "loss": 0.6086,
+      "step": 1532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994112682757348e-05,
+      "loss": 0.6486,
+      "step": 1533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9941049566155414e-05,
+      "loss": 0.6441,
+      "step": 1534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940972254223964e-05,
+      "loss": 0.5946,
+      "step": 1535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940894891779524e-05,
+      "loss": 0.5812,
+      "step": 1536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940817478822487e-05,
+      "loss": 0.621,
+      "step": 1537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940740015353244e-05,
+      "loss": 0.5926,
+      "step": 1538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940662501372188e-05,
+      "loss": 0.6641,
+      "step": 1539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940584936879714e-05,
+      "loss": 0.5988,
+      "step": 1540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.994050732187622e-05,
+      "loss": 0.5804,
+      "step": 1541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9940429656362096e-05,
+      "loss": 0.6011,
+      "step": 1542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9940351940337735e-05,
+      "loss": 0.6798,
+      "step": 1543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.994027417380354e-05,
+      "loss": 0.6368,
+      "step": 1544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9940196356759894e-05,
+      "loss": 0.7251,
+      "step": 1545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9940118489207204e-05,
+      "loss": 0.6225,
+      "step": 1546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.994004057114586e-05,
+      "loss": 0.6346,
+      "step": 1547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993996260257626e-05,
+      "loss": 0.6676,
+      "step": 1548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.99398845834988e-05,
+      "loss": 0.6156,
+      "step": 1549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993980651391387e-05,
+      "loss": 0.648,
+      "step": 1550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939728393821875e-05,
+      "loss": 0.6414,
+      "step": 1551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993965022322321e-05,
+      "loss": 0.6566,
+      "step": 1552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939572002118268e-05,
+      "loss": 0.6401,
+      "step": 1553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939493730507453e-05,
+      "loss": 0.663,
+      "step": 1554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939415408391157e-05,
+      "loss": 0.6398,
+      "step": 1555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939337035769783e-05,
+      "loss": 0.6471,
+      "step": 1556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939258612643727e-05,
+      "loss": 0.6207,
+      "step": 1557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939180139013386e-05,
+      "loss": 0.6401,
+      "step": 1558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939101614879156e-05,
+      "loss": 0.6453,
+      "step": 1559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9939023040241444e-05,
+      "loss": 0.5962,
+      "step": 1560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938944415100646e-05,
+      "loss": 0.6117,
+      "step": 1561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993886573945716e-05,
+      "loss": 0.6262,
+      "step": 1562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938787013311388e-05,
+      "loss": 0.6375,
+      "step": 1563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938708236663724e-05,
+      "loss": 0.6246,
+      "step": 1564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938629409514574e-05,
+      "loss": 0.688,
+      "step": 1565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993855053186434e-05,
+      "loss": 0.6446,
+      "step": 1566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993847160371342e-05,
+      "loss": 0.5727,
+      "step": 1567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993839262506221e-05,
+      "loss": 0.658,
+      "step": 1568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993831359591112e-05,
+      "loss": 0.6124,
+      "step": 1569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993823451626055e-05,
+      "loss": 0.603,
+      "step": 1570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938155386110897e-05,
+      "loss": 0.5949,
+      "step": 1571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9938076205462565e-05,
+      "loss": 0.5927,
+      "step": 1572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993799697431596e-05,
+      "loss": 0.6582,
+      "step": 1573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937917692671483e-05,
+      "loss": 0.6262,
+      "step": 1574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937838360529533e-05,
+      "loss": 0.6375,
+      "step": 1575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937758977890514e-05,
+      "loss": 0.5847,
+      "step": 1576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937679544754837e-05,
+      "loss": 0.6087,
+      "step": 1577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937600061122897e-05,
+      "loss": 0.6162,
+      "step": 1578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.99375205269951e-05,
+      "loss": 0.5886,
+      "step": 1579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993744094237185e-05,
+      "loss": 0.6834,
+      "step": 1580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937361307253554e-05,
+      "loss": 0.6511,
+      "step": 1581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937281621640615e-05,
+      "loss": 0.6282,
+      "step": 1582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993720188553344e-05,
+      "loss": 0.657,
+      "step": 1583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9937122098932428e-05,
+      "loss": 0.5916,
+      "step": 1584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993704226183799e-05,
+      "loss": 0.6145,
+      "step": 1585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993696237425053e-05,
+      "loss": 0.5977,
+      "step": 1586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936882436170454e-05,
+      "loss": 0.6455,
+      "step": 1587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936802447598167e-05,
+      "loss": 0.6863,
+      "step": 1588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936722408534075e-05,
+      "loss": 0.6135,
+      "step": 1589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993664231897859e-05,
+      "loss": 0.6333,
+      "step": 1590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936562178932116e-05,
+      "loss": 0.6053,
+      "step": 1591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936481988395058e-05,
+      "loss": 0.5938,
+      "step": 1592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936401747367822e-05,
+      "loss": 0.6359,
+      "step": 1593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936321455850823e-05,
+      "loss": 0.7243,
+      "step": 1594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936241113844463e-05,
+      "loss": 0.6616,
+      "step": 1595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9936160721349153e-05,
+      "loss": 0.6234,
+      "step": 1596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.99360802783653e-05,
+      "loss": 0.6064,
+      "step": 1597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935999784893313e-05,
+      "loss": 0.6352,
+      "step": 1598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.99359192409336e-05,
+      "loss": 0.6058,
+      "step": 1599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935838646486574e-05,
+      "loss": 0.6004,
+      "step": 1600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993575800155264e-05,
+      "loss": 0.6359,
+      "step": 1601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993567730613221e-05,
+      "loss": 0.6023,
+      "step": 1602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935596560225692e-05,
+      "loss": 0.6513,
+      "step": 1603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935515763833502e-05,
+      "loss": 0.5946,
+      "step": 1604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993543491695604e-05,
+      "loss": 0.669,
+      "step": 1605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935354019593732e-05,
+      "loss": 0.6359,
+      "step": 1606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935273071746975e-05,
+      "loss": 0.6648,
+      "step": 1607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993519207341619e-05,
+      "loss": 0.6198,
+      "step": 1608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935111024601782e-05,
+      "loss": 0.6084,
+      "step": 1609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9935029925304166e-05,
+      "loss": 0.6541,
+      "step": 1610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934948775523752e-05,
+      "loss": 0.6568,
+      "step": 1611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934867575260953e-05,
+      "loss": 0.6497,
+      "step": 1612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934786324516184e-05,
+      "loss": 0.5951,
+      "step": 1613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934705023289856e-05,
+      "loss": 0.6772,
+      "step": 1614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993462367158238e-05,
+      "loss": 0.6184,
+      "step": 1615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934542269394175e-05,
+      "loss": 0.6196,
+      "step": 1616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993446081672565e-05,
+      "loss": 0.6354,
+      "step": 1617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993437931357722e-05,
+      "loss": 0.7381,
+      "step": 1618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.99342977599493e-05,
+      "loss": 0.6224,
+      "step": 1619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934216155842302e-05,
+      "loss": 0.6016,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934134501256645e-05,
+      "loss": 0.6691,
+      "step": 1621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9934052796192737e-05,
+      "loss": 0.5998,
+      "step": 1622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933971040651e-05,
+      "loss": 0.5886,
+      "step": 1623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933889234631845e-05,
+      "loss": 0.6505,
+      "step": 1624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933807378135695e-05,
+      "loss": 0.6108,
+      "step": 1625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933725471162955e-05,
+      "loss": 0.6391,
+      "step": 1626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993364351371405e-05,
+      "loss": 0.5843,
+      "step": 1627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933561505789392e-05,
+      "loss": 0.5779,
+      "step": 1628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933479447389397e-05,
+      "loss": 0.6026,
+      "step": 1629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933397338514484e-05,
+      "loss": 0.619,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933315179165072e-05,
+      "loss": 0.6698,
+      "step": 1631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933232969341576e-05,
+      "loss": 0.631,
+      "step": 1632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933150709044417e-05,
+      "loss": 0.5842,
+      "step": 1633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9933068398274004e-05,
+      "loss": 0.5866,
+      "step": 1634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932986037030765e-05,
+      "loss": 0.5909,
+      "step": 1635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932903625315116e-05,
+      "loss": 0.595,
+      "step": 1636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993282116312747e-05,
+      "loss": 0.674,
+      "step": 1637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932738650468256e-05,
+      "loss": 0.626,
+      "step": 1638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932656087337885e-05,
+      "loss": 0.6648,
+      "step": 1639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932573473736778e-05,
+      "loss": 0.6379,
+      "step": 1640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993249080966536e-05,
+      "loss": 0.6641,
+      "step": 1641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932408095124044e-05,
+      "loss": 0.5972,
+      "step": 1642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932325330113257e-05,
+      "loss": 0.6307,
+      "step": 1643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932242514633414e-05,
+      "loss": 0.6319,
+      "step": 1644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993215964868494e-05,
+      "loss": 0.6152,
+      "step": 1645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9932076732268253e-05,
+      "loss": 0.6454,
+      "step": 1646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931993765383775e-05,
+      "loss": 0.6534,
+      "step": 1647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931910748031928e-05,
+      "loss": 0.5519,
+      "step": 1648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931827680213134e-05,
+      "loss": 0.5838,
+      "step": 1649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931744561927815e-05,
+      "loss": 0.5942,
+      "step": 1650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931661393176395e-05,
+      "loss": 0.5932,
+      "step": 1651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993157817395929e-05,
+      "loss": 0.5787,
+      "step": 1652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931494904276932e-05,
+      "loss": 0.5783,
+      "step": 1653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931411584129737e-05,
+      "loss": 0.5688,
+      "step": 1654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931328213518134e-05,
+      "loss": 0.6326,
+      "step": 1655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993124479244254e-05,
+      "loss": 0.6243,
+      "step": 1656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931161320903385e-05,
+      "loss": 0.5808,
+      "step": 1657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9931077798901093e-05,
+      "loss": 0.6021,
+      "step": 1658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993099422643608e-05,
+      "loss": 0.6424,
+      "step": 1659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930910603508783e-05,
+      "loss": 0.6308,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993082693011962e-05,
+      "loss": 0.6086,
+      "step": 1661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930743206269016e-05,
+      "loss": 0.6421,
+      "step": 1662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930659431957397e-05,
+      "loss": 0.6034,
+      "step": 1663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993057560718519e-05,
+      "loss": 0.6052,
+      "step": 1664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993049173195282e-05,
+      "loss": 0.6348,
+      "step": 1665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930407806260712e-05,
+      "loss": 0.5437,
+      "step": 1666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930323830109294e-05,
+      "loss": 0.6346,
+      "step": 1667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930239803498992e-05,
+      "loss": 0.5934,
+      "step": 1668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9930155726430236e-05,
+      "loss": 0.6612,
+      "step": 1669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.993007159890345e-05,
+      "loss": 0.6872,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.992998742091906e-05,
+      "loss": 0.6413,
+      "step": 1671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929903192477495e-05,
+      "loss": 0.553,
+      "step": 1672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929818913579186e-05,
+      "loss": 0.6276,
+      "step": 1673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.992973458422456e-05,
+      "loss": 0.5717,
+      "step": 1674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.992965020441404e-05,
+      "loss": 0.6186,
+      "step": 1675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929565774148063e-05,
+      "loss": 0.643,
+      "step": 1676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929481293427054e-05,
+      "loss": 0.6477,
+      "step": 1677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929396762251443e-05,
+      "loss": 0.649,
+      "step": 1678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929312180621657e-05,
+      "loss": 0.5868,
+      "step": 1679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.992922754853813e-05,
+      "loss": 0.6568,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929142866001288e-05,
+      "loss": 0.6049,
+      "step": 1681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9929058133011563e-05,
+      "loss": 0.7018,
+      "step": 1682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928973349569387e-05,
+      "loss": 0.6173,
+      "step": 1683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928888515675192e-05,
+      "loss": 0.6203,
+      "step": 1684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928803631329403e-05,
+      "loss": 0.6531,
+      "step": 1685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928718696532455e-05,
+      "loss": 0.6489,
+      "step": 1686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928633711284782e-05,
+      "loss": 0.6354,
+      "step": 1687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.992854867558681e-05,
+      "loss": 0.6591,
+      "step": 1688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9928463589438975e-05,
+      "loss": 0.6129,
+      "step": 1689
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992837845284171e-05,
+      "loss": 0.6489,
+      "step": 1690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992829326579545e-05,
+      "loss": 0.6333,
+      "step": 1691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992820802830062e-05,
+      "loss": 0.6448,
+      "step": 1692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9928122740357655e-05,
+      "loss": 0.571,
+      "step": 1693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9928037401966994e-05,
+      "loss": 0.6264,
+      "step": 1694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992795201312907e-05,
+      "loss": 0.5572,
+      "step": 1695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927866573844313e-05,
+      "loss": 0.5988,
+      "step": 1696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927781084113157e-05,
+      "loss": 0.6336,
+      "step": 1697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927695543936036e-05,
+      "loss": 0.6761,
+      "step": 1698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992760995331339e-05,
+      "loss": 0.6041,
+      "step": 1699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992752431224565e-05,
+      "loss": 0.6409,
+      "step": 1700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927438620733248e-05,
+      "loss": 0.5966,
+      "step": 1701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927352878776625e-05,
+      "loss": 0.624,
+      "step": 1702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927267086376216e-05,
+      "loss": 0.6742,
+      "step": 1703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927181243532458e-05,
+      "loss": 0.6504,
+      "step": 1704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927095350245776e-05,
+      "loss": 0.6576,
+      "step": 1705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9927009406516623e-05,
+      "loss": 0.6887,
+      "step": 1706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926923412345424e-05,
+      "loss": 0.5476,
+      "step": 1707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926837367732623e-05,
+      "loss": 0.6611,
+      "step": 1708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926751272678655e-05,
+      "loss": 0.6305,
+      "step": 1709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926665127183953e-05,
+      "loss": 0.639,
+      "step": 1710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926578931248957e-05,
+      "loss": 0.623,
+      "step": 1711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992649268487411e-05,
+      "loss": 0.5695,
+      "step": 1712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926406388059846e-05,
+      "loss": 0.6514,
+      "step": 1713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926320040806602e-05,
+      "loss": 0.6227,
+      "step": 1714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9926233643114823e-05,
+      "loss": 0.6761,
+      "step": 1715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992614719498494e-05,
+      "loss": 0.6121,
+      "step": 1716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.99260606964174e-05,
+      "loss": 0.5859,
+      "step": 1717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925974147412634e-05,
+      "loss": 0.6398,
+      "step": 1718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992588754797109e-05,
+      "loss": 0.6117,
+      "step": 1719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925800898093203e-05,
+      "loss": 0.5882,
+      "step": 1720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925714197779415e-05,
+      "loss": 0.63,
+      "step": 1721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925627447030164e-05,
+      "loss": 0.6737,
+      "step": 1722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925540645845895e-05,
+      "loss": 0.6557,
+      "step": 1723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992545379422705e-05,
+      "loss": 0.6427,
+      "step": 1724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992536689217406e-05,
+      "loss": 0.6233,
+      "step": 1725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992527993968738e-05,
+      "loss": 0.5877,
+      "step": 1726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925192936767447e-05,
+      "loss": 0.6762,
+      "step": 1727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925105883414697e-05,
+      "loss": 0.5682,
+      "step": 1728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9925018779629578e-05,
+      "loss": 0.5916,
+      "step": 1729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924931625412534e-05,
+      "loss": 0.6332,
+      "step": 1730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924844420764006e-05,
+      "loss": 0.6267,
+      "step": 1731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924757165684437e-05,
+      "loss": 0.5578,
+      "step": 1732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924669860174268e-05,
+      "loss": 0.6473,
+      "step": 1733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924582504233946e-05,
+      "loss": 0.6557,
+      "step": 1734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924495097863915e-05,
+      "loss": 0.59,
+      "step": 1735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924407641064615e-05,
+      "loss": 0.6365,
+      "step": 1736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924320133836498e-05,
+      "loss": 0.6889,
+      "step": 1737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992423257618e-05,
+      "loss": 0.6578,
+      "step": 1738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992414496809557e-05,
+      "loss": 0.615,
+      "step": 1739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9924057309583654e-05,
+      "loss": 0.6308,
+      "step": 1740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923969600644696e-05,
+      "loss": 0.6629,
+      "step": 1741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923881841279146e-05,
+      "loss": 0.5984,
+      "step": 1742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992379403148744e-05,
+      "loss": 0.6764,
+      "step": 1743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992370617127003e-05,
+      "loss": 0.6447,
+      "step": 1744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992361826062737e-05,
+      "loss": 0.653,
+      "step": 1745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923530299559895e-05,
+      "loss": 0.5978,
+      "step": 1746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923442288068053e-05,
+      "loss": 0.6124,
+      "step": 1747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.99233542261523e-05,
+      "loss": 0.7269,
+      "step": 1748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923266113813073e-05,
+      "loss": 0.7523,
+      "step": 1749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992317795105083e-05,
+      "loss": 0.6752,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992308973786601e-05,
+      "loss": 0.6615,
+      "step": 1751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9923001474259065e-05,
+      "loss": 0.6039,
+      "step": 1752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9922913160230445e-05,
+      "loss": 0.5681,
+      "step": 1753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9922824795780594e-05,
+      "loss": 0.5979,
+      "step": 1754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992273638090997e-05,
+      "loss": 0.6184,
+      "step": 1755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992264791561901e-05,
+      "loss": 0.6456,
+      "step": 1756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9922559399908174e-05,
+      "loss": 0.6551,
+      "step": 1757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992247083377791e-05,
+      "loss": 0.6499,
+      "step": 1758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992238221722866e-05,
+      "loss": 0.6036,
+      "step": 1759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992229355026088e-05,
+      "loss": 0.6358,
+      "step": 1760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9922204832875027e-05,
+      "loss": 0.5501,
+      "step": 1761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992211606507154e-05,
+      "loss": 0.5964,
+      "step": 1762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992202724685088e-05,
+      "loss": 0.6484,
+      "step": 1763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992193837821349e-05,
+      "loss": 0.6202,
+      "step": 1764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921849459159827e-05,
+      "loss": 0.6274,
+      "step": 1765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921760489690337e-05,
+      "loss": 0.7056,
+      "step": 1766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921671469805483e-05,
+      "loss": 0.6113,
+      "step": 1767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921582399505705e-05,
+      "loss": 0.6764,
+      "step": 1768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921493278791462e-05,
+      "loss": 0.625,
+      "step": 1769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921404107663207e-05,
+      "loss": 0.6166,
+      "step": 1770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921314886121394e-05,
+      "loss": 0.6215,
+      "step": 1771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921225614166473e-05,
+      "loss": 0.6475,
+      "step": 1772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.99211362917989e-05,
+      "loss": 0.6547,
+      "step": 1773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9921046919019123e-05,
+      "loss": 0.6241,
+      "step": 1774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920957495827605e-05,
+      "loss": 0.5788,
+      "step": 1775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920868022224796e-05,
+      "loss": 0.5842,
+      "step": 1776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920778498211156e-05,
+      "loss": 0.6904,
+      "step": 1777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920688923787127e-05,
+      "loss": 0.5968,
+      "step": 1778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920599298953175e-05,
+      "loss": 0.6486,
+      "step": 1779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920509623709754e-05,
+      "loss": 0.6004,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992041989805732e-05,
+      "loss": 0.6152,
+      "step": 1781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920330121996324e-05,
+      "loss": 0.5942,
+      "step": 1782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920240295527226e-05,
+      "loss": 0.6631,
+      "step": 1783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9920150418650482e-05,
+      "loss": 0.6463,
+      "step": 1784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.992006049136655e-05,
+      "loss": 0.599,
+      "step": 1785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991997051367588e-05,
+      "loss": 0.6488,
+      "step": 1786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991988048557894e-05,
+      "loss": 0.6331,
+      "step": 1787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9919790407076182e-05,
+      "loss": 0.6289,
+      "step": 1788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991970027816806e-05,
+      "loss": 0.5957,
+      "step": 1789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991961009885504e-05,
+      "loss": 0.5912,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9919519869137573e-05,
+      "loss": 0.66,
+      "step": 1791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991942958901612e-05,
+      "loss": 0.6288,
+      "step": 1792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9919339258491143e-05,
+      "loss": 0.6554,
+      "step": 1793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9919248877563096e-05,
+      "loss": 0.6002,
+      "step": 1794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991915844623244e-05,
+      "loss": 0.5991,
+      "step": 1795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991906796449964e-05,
+      "loss": 0.5893,
+      "step": 1796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918977432365146e-05,
+      "loss": 0.6007,
+      "step": 1797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918886849829425e-05,
+      "loss": 0.5953,
+      "step": 1798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991879621689293e-05,
+      "loss": 0.7013,
+      "step": 1799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918705533556133e-05,
+      "loss": 0.6676,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918614799819488e-05,
+      "loss": 0.6077,
+      "step": 1801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991852401568345e-05,
+      "loss": 0.6749,
+      "step": 1802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991843318114849e-05,
+      "loss": 0.6681,
+      "step": 1803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991834229621507e-05,
+      "loss": 0.6785,
+      "step": 1804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918251360883644e-05,
+      "loss": 0.6119,
+      "step": 1805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991816037515468e-05,
+      "loss": 0.6368,
+      "step": 1806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9918069339028637e-05,
+      "loss": 0.6368,
+      "step": 1807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991797825250598e-05,
+      "loss": 0.6132,
+      "step": 1808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917887115587172e-05,
+      "loss": 0.6226,
+      "step": 1809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991779592827267e-05,
+      "loss": 0.6224,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917704690562948e-05,
+      "loss": 0.6827,
+      "step": 1811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991761340245846e-05,
+      "loss": 0.619,
+      "step": 1812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991752206395968e-05,
+      "loss": 0.571,
+      "step": 1813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917430675067063e-05,
+      "loss": 0.579,
+      "step": 1814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917339235781074e-05,
+      "loss": 0.6537,
+      "step": 1815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991724774610218e-05,
+      "loss": 0.5898,
+      "step": 1816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917156206030847e-05,
+      "loss": 0.5552,
+      "step": 1817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9917064615567538e-05,
+      "loss": 0.6207,
+      "step": 1818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991697297471272e-05,
+      "loss": 0.5537,
+      "step": 1819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9916881283466857e-05,
+      "loss": 0.6959,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991678954183042e-05,
+      "loss": 0.6511,
+      "step": 1821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9916697749803865e-05,
+      "loss": 0.6813,
+      "step": 1822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991660590738767e-05,
+      "loss": 0.6132,
+      "step": 1823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9916514014582292e-05,
+      "loss": 0.6547,
+      "step": 1824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9916422071388203e-05,
+      "loss": 0.5785,
+      "step": 1825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991633007780587e-05,
+      "loss": 0.5967,
+      "step": 1826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991623803383576e-05,
+      "loss": 0.6238,
+      "step": 1827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991614593947834e-05,
+      "loss": 0.6262,
+      "step": 1828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991605379473408e-05,
+      "loss": 0.657,
+      "step": 1829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9915961599603446e-05,
+      "loss": 0.6076,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9915869354086906e-05,
+      "loss": 0.6191,
+      "step": 1831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991577705818493e-05,
+      "loss": 0.6314,
+      "step": 1832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9915684711897988e-05,
+      "loss": 0.6317,
+      "step": 1833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9915592315226545e-05,
+      "loss": 0.6577,
+      "step": 1834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9915499868171078e-05,
+      "loss": 0.6049,
+      "step": 1835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.991540737073205e-05,
+      "loss": 0.6237,
+      "step": 1836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9915314822909934e-05,
+      "loss": 0.6287,
+      "step": 1837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9915222224705195e-05,
+      "loss": 0.5676,
+      "step": 1838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991512957611831e-05,
+      "loss": 0.605,
+      "step": 1839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9915036877149753e-05,
+      "loss": 0.6156,
+      "step": 1840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914944127799987e-05,
+      "loss": 0.6696,
+      "step": 1841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991485132806948e-05,
+      "loss": 0.5776,
+      "step": 1842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914758477958716e-05,
+      "loss": 0.6304,
+      "step": 1843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914665577468163e-05,
+      "loss": 0.6099,
+      "step": 1844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991457262659829e-05,
+      "loss": 0.6589,
+      "step": 1845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914479625349564e-05,
+      "loss": 0.6252,
+      "step": 1846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914386573722465e-05,
+      "loss": 0.6204,
+      "step": 1847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914293471717467e-05,
+      "loss": 0.5861,
+      "step": 1848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991420031933504e-05,
+      "loss": 0.5834,
+      "step": 1849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9914107116575656e-05,
+      "loss": 0.6282,
+      "step": 1850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991401386343979e-05,
+      "loss": 0.6067,
+      "step": 1851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991392055992792e-05,
+      "loss": 0.6398,
+      "step": 1852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9913827206040513e-05,
+      "loss": 0.5559,
+      "step": 1853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991373380177805e-05,
+      "loss": 0.5863,
+      "step": 1854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9913640347140998e-05,
+      "loss": 0.6094,
+      "step": 1855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9913546842129837e-05,
+      "loss": 0.6349,
+      "step": 1856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991345328674504e-05,
+      "loss": 0.6124,
+      "step": 1857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9913359680987086e-05,
+      "loss": 0.6513,
+      "step": 1858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9913266024856444e-05,
+      "loss": 0.6136,
+      "step": 1859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99131723183536e-05,
+      "loss": 0.6872,
+      "step": 1860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991307856147902e-05,
+      "loss": 0.5511,
+      "step": 1861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9912984754233187e-05,
+      "loss": 0.5662,
+      "step": 1862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9912890896616575e-05,
+      "loss": 0.6313,
+      "step": 1863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991279698862966e-05,
+      "loss": 0.5886,
+      "step": 1864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9912703030272922e-05,
+      "loss": 0.5679,
+      "step": 1865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9912609021546835e-05,
+      "loss": 0.687,
+      "step": 1866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991251496245188e-05,
+      "loss": 0.5931,
+      "step": 1867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991242085298853e-05,
+      "loss": 0.7095,
+      "step": 1868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991232669315727e-05,
+      "loss": 0.6711,
+      "step": 1869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991223248295858e-05,
+      "loss": 0.6345,
+      "step": 1870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9912138222392924e-05,
+      "loss": 0.6406,
+      "step": 1871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99120439114608e-05,
+      "loss": 0.704,
+      "step": 1872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991194955016267e-05,
+      "loss": 0.6402,
+      "step": 1873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9911855138499026e-05,
+      "loss": 0.5813,
+      "step": 1874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9911760676470343e-05,
+      "loss": 0.5982,
+      "step": 1875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99116661640771e-05,
+      "loss": 0.6528,
+      "step": 1876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991157160131978e-05,
+      "loss": 0.6275,
+      "step": 1877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991147698819886e-05,
+      "loss": 0.5718,
+      "step": 1878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991138232471482e-05,
+      "loss": 0.5818,
+      "step": 1879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991128761086815e-05,
+      "loss": 0.6484,
+      "step": 1880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9911192846659323e-05,
+      "loss": 0.6268,
+      "step": 1881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991109803208882e-05,
+      "loss": 0.6455,
+      "step": 1882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991100316715713e-05,
+      "loss": 0.6233,
+      "step": 1883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910908251864723e-05,
+      "loss": 0.6277,
+      "step": 1884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910813286212095e-05,
+      "loss": 0.6121,
+      "step": 1885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910718270199717e-05,
+      "loss": 0.6337,
+      "step": 1886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910623203828082e-05,
+      "loss": 0.6596,
+      "step": 1887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910528087097666e-05,
+      "loss": 0.6183,
+      "step": 1888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910432920008955e-05,
+      "loss": 0.6924,
+      "step": 1889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991033770256243e-05,
+      "loss": 0.6244,
+      "step": 1890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.991024243475858e-05,
+      "loss": 0.6485,
+      "step": 1891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910147116597883e-05,
+      "loss": 0.6458,
+      "step": 1892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9910051748080827e-05,
+      "loss": 0.5684,
+      "step": 1893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909956329207897e-05,
+      "loss": 0.6259,
+      "step": 1894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909860859979573e-05,
+      "loss": 0.6201,
+      "step": 1895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990976534039635e-05,
+      "loss": 0.6654,
+      "step": 1896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99096697704587e-05,
+      "loss": 0.6979,
+      "step": 1897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909574150167123e-05,
+      "loss": 0.5786,
+      "step": 1898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909478479522092e-05,
+      "loss": 0.6579,
+      "step": 1899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99093827585241e-05,
+      "loss": 0.5684,
+      "step": 1900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990928698717363e-05,
+      "loss": 0.6231,
+      "step": 1901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909191165471174e-05,
+      "loss": 0.6302,
+      "step": 1902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9909095293417216e-05,
+      "loss": 0.6635,
+      "step": 1903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990899937101224e-05,
+      "loss": 0.5961,
+      "step": 1904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908903398256738e-05,
+      "loss": 0.57,
+      "step": 1905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908807375151195e-05,
+      "loss": 0.6572,
+      "step": 1906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908711301696102e-05,
+      "loss": 0.6795,
+      "step": 1907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908615177891944e-05,
+      "loss": 0.6061,
+      "step": 1908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990851900373921e-05,
+      "loss": 0.5651,
+      "step": 1909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990842277923839e-05,
+      "loss": 0.5982,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990832650438997e-05,
+      "loss": 0.5804,
+      "step": 1911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908230179194444e-05,
+      "loss": 0.6269,
+      "step": 1912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9908133803652297e-05,
+      "loss": 0.683,
+      "step": 1913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990803737776402e-05,
+      "loss": 0.6432,
+      "step": 1914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907940901530104e-05,
+      "loss": 0.6084,
+      "step": 1915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907844374951036e-05,
+      "loss": 0.6552,
+      "step": 1916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907747798027315e-05,
+      "loss": 0.6647,
+      "step": 1917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990765117075942e-05,
+      "loss": 0.5954,
+      "step": 1918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990755449314785e-05,
+      "loss": 0.6784,
+      "step": 1919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907457765193096e-05,
+      "loss": 0.6276,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907360986895647e-05,
+      "loss": 0.6363,
+      "step": 1921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907264158255993e-05,
+      "loss": 0.7074,
+      "step": 1922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907167279274628e-05,
+      "loss": 0.6857,
+      "step": 1923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9907070349952048e-05,
+      "loss": 0.6299,
+      "step": 1924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906973370288737e-05,
+      "loss": 0.6696,
+      "step": 1925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99068763402852e-05,
+      "loss": 0.6357,
+      "step": 1926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906779259941916e-05,
+      "loss": 0.6286,
+      "step": 1927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990668212925939e-05,
+      "loss": 0.6652,
+      "step": 1928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906584948238107e-05,
+      "loss": 0.6458,
+      "step": 1929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906487716878567e-05,
+      "loss": 0.65,
+      "step": 1930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990639043518126e-05,
+      "loss": 0.5918,
+      "step": 1931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906293103146685e-05,
+      "loss": 0.6005,
+      "step": 1932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906195720775333e-05,
+      "loss": 0.638,
+      "step": 1933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906098288067696e-05,
+      "loss": 0.5979,
+      "step": 1934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9906000805024276e-05,
+      "loss": 0.6156,
+      "step": 1935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905903271645564e-05,
+      "loss": 0.5978,
+      "step": 1936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905805687932057e-05,
+      "loss": 0.5792,
+      "step": 1937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990570805388425e-05,
+      "loss": 0.5502,
+      "step": 1938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905610369502638e-05,
+      "loss": 0.6724,
+      "step": 1939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905512634787718e-05,
+      "loss": 0.6112,
+      "step": 1940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990541484973999e-05,
+      "loss": 0.5513,
+      "step": 1941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905317014359947e-05,
+      "loss": 0.606,
+      "step": 1942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905219128648085e-05,
+      "loss": 0.6368,
+      "step": 1943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905121192604905e-05,
+      "loss": 0.6471,
+      "step": 1944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9905023206230906e-05,
+      "loss": 0.6113,
+      "step": 1945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990492516952658e-05,
+      "loss": 0.6503,
+      "step": 1946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990482708249243e-05,
+      "loss": 0.6805,
+      "step": 1947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904728945128954e-05,
+      "loss": 0.6666,
+      "step": 1948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990463075743665e-05,
+      "loss": 0.5957,
+      "step": 1949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904532519416012e-05,
+      "loss": 0.5816,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904434231067547e-05,
+      "loss": 0.6291,
+      "step": 1951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990433589239175e-05,
+      "loss": 0.5204,
+      "step": 1952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904237503389122e-05,
+      "loss": 0.6221,
+      "step": 1953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904139064060165e-05,
+      "loss": 0.6786,
+      "step": 1954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9904040574405372e-05,
+      "loss": 0.6383,
+      "step": 1955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990394203442525e-05,
+      "loss": 0.6424,
+      "step": 1956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903843444120298e-05,
+      "loss": 0.6006,
+      "step": 1957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990374480349102e-05,
+      "loss": 0.6192,
+      "step": 1958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990364611253791e-05,
+      "loss": 0.5646,
+      "step": 1959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990354737126148e-05,
+      "loss": 0.6375,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903448579662218e-05,
+      "loss": 0.5775,
+      "step": 1961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903349737740635e-05,
+      "loss": 0.5805,
+      "step": 1962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903250845497235e-05,
+      "loss": 0.6611,
+      "step": 1963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903151902932516e-05,
+      "loss": 0.6191,
+      "step": 1964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9903052910046982e-05,
+      "loss": 0.7136,
+      "step": 1965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902953866841135e-05,
+      "loss": 0.6092,
+      "step": 1966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990285477331548e-05,
+      "loss": 0.6245,
+      "step": 1967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902755629470518e-05,
+      "loss": 0.6352,
+      "step": 1968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902656435306757e-05,
+      "loss": 0.5831,
+      "step": 1969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902557190824696e-05,
+      "loss": 0.594,
+      "step": 1970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902457896024842e-05,
+      "loss": 0.602,
+      "step": 1971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.99023585509077e-05,
+      "loss": 0.5914,
+      "step": 1972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902259155473774e-05,
+      "loss": 0.619,
+      "step": 1973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990215970972357e-05,
+      "loss": 0.6059,
+      "step": 1974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9902060213657594e-05,
+      "loss": 0.6332,
+      "step": 1975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901960667276348e-05,
+      "loss": 0.6236,
+      "step": 1976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901861070580338e-05,
+      "loss": 0.6193,
+      "step": 1977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901761423570077e-05,
+      "loss": 0.6171,
+      "step": 1978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901661726246062e-05,
+      "loss": 0.5855,
+      "step": 1979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901561978608805e-05,
+      "loss": 0.6379,
+      "step": 1980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901462180658814e-05,
+      "loss": 0.6349,
+      "step": 1981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.990136233239659e-05,
+      "loss": 0.6703,
+      "step": 1982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9901262433822647e-05,
+      "loss": 0.6175,
+      "step": 1983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.990116248493749e-05,
+      "loss": 0.6153,
+      "step": 1984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9901062485741626e-05,
+      "loss": 0.639,
+      "step": 1985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900962436235565e-05,
+      "loss": 0.6043,
+      "step": 1986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900862336419813e-05,
+      "loss": 0.6823,
+      "step": 1987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.990076218629488e-05,
+      "loss": 0.6198,
+      "step": 1988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900661985861275e-05,
+      "loss": 0.6254,
+      "step": 1989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900561735119502e-05,
+      "loss": 0.5985,
+      "step": 1990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.990046143407008e-05,
+      "loss": 0.6298,
+      "step": 1991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900361082713516e-05,
+      "loss": 0.6615,
+      "step": 1992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900260681050316e-05,
+      "loss": 0.5829,
+      "step": 1993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900160229080993e-05,
+      "loss": 0.677,
+      "step": 1994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9900059726806054e-05,
+      "loss": 0.5633,
+      "step": 1995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899959174226014e-05,
+      "loss": 0.5628,
+      "step": 1996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989985857134138e-05,
+      "loss": 0.5996,
+      "step": 1997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899757918152667e-05,
+      "loss": 0.612,
+      "step": 1998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899657214660383e-05,
+      "loss": 0.5816,
+      "step": 1999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899556460865044e-05,
+      "loss": 0.5862,
+      "step": 2000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989945565676716e-05,
+      "loss": 0.5941,
+      "step": 2001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899354802367242e-05,
+      "loss": 0.674,
+      "step": 2002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.98992538976658e-05,
+      "loss": 0.6825,
+      "step": 2003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9899152942663353e-05,
+      "loss": 0.6195,
+      "step": 2004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989905193736041e-05,
+      "loss": 0.6352,
+      "step": 2005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898950881757483e-05,
+      "loss": 0.6304,
+      "step": 2006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898849775855092e-05,
+      "loss": 0.6857,
+      "step": 2007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898748619653743e-05,
+      "loss": 0.632,
+      "step": 2008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898647413153952e-05,
+      "loss": 0.6607,
+      "step": 2009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898546156356236e-05,
+      "loss": 0.6035,
+      "step": 2010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989844484926111e-05,
+      "loss": 0.6001,
+      "step": 2011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898343491869085e-05,
+      "loss": 0.6745,
+      "step": 2012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989824208418068e-05,
+      "loss": 0.5758,
+      "step": 2013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9898140626196406e-05,
+      "loss": 0.5291,
+      "step": 2014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989803911791678e-05,
+      "loss": 0.689,
+      "step": 2015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989793755934232e-05,
+      "loss": 0.5811,
+      "step": 2016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989783595047354e-05,
+      "loss": 0.6551,
+      "step": 2017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897734291310955e-05,
+      "loss": 0.6219,
+      "step": 2018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897632581855085e-05,
+      "loss": 0.6404,
+      "step": 2019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897530822106444e-05,
+      "loss": 0.5989,
+      "step": 2020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897429012065554e-05,
+      "loss": 0.6196,
+      "step": 2021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897327151732926e-05,
+      "loss": 0.6353,
+      "step": 2022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989722524110908e-05,
+      "loss": 0.5437,
+      "step": 2023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897123280194534e-05,
+      "loss": 0.6984,
+      "step": 2024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9897021268989807e-05,
+      "loss": 0.6309,
+      "step": 2025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896919207495416e-05,
+      "loss": 0.6339,
+      "step": 2026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989681709571188e-05,
+      "loss": 0.6717,
+      "step": 2027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989671493363972e-05,
+      "loss": 0.6182,
+      "step": 2028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896612721279446e-05,
+      "loss": 0.6718,
+      "step": 2029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989651045863159e-05,
+      "loss": 0.6212,
+      "step": 2030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896408145696667e-05,
+      "loss": 0.6418,
+      "step": 2031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896305782475194e-05,
+      "loss": 0.6567,
+      "step": 2032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896203368967693e-05,
+      "loss": 0.6237,
+      "step": 2033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9896100905174685e-05,
+      "loss": 0.5867,
+      "step": 2034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895998391096693e-05,
+      "loss": 0.5527,
+      "step": 2035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989589582673423e-05,
+      "loss": 0.6892,
+      "step": 2036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895793212087825e-05,
+      "loss": 0.6314,
+      "step": 2037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895690547157994e-05,
+      "loss": 0.6106,
+      "step": 2038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895587831945265e-05,
+      "loss": 0.6299,
+      "step": 2039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989548506645015e-05,
+      "loss": 0.6433,
+      "step": 2040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895382250673184e-05,
+      "loss": 0.5848,
+      "step": 2041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989527938461488e-05,
+      "loss": 0.6055,
+      "step": 2042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895176468275765e-05,
+      "loss": 0.6267,
+      "step": 2043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9895073501656355e-05,
+      "loss": 0.6316,
+      "step": 2044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894970484757182e-05,
+      "loss": 0.6131,
+      "step": 2045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989486741757877e-05,
+      "loss": 0.6262,
+      "step": 2046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894764300121635e-05,
+      "loss": 0.6081,
+      "step": 2047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894661132386307e-05,
+      "loss": 0.6005,
+      "step": 2048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894557914373303e-05,
+      "loss": 0.6274,
+      "step": 2049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894454646083157e-05,
+      "loss": 0.6634,
+      "step": 2050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894351327516392e-05,
+      "loss": 0.6843,
+      "step": 2051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989424795867352e-05,
+      "loss": 0.6203,
+      "step": 2052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894144539555085e-05,
+      "loss": 0.6246,
+      "step": 2053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9894041070161604e-05,
+      "loss": 0.5724,
+      "step": 2054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893937550493598e-05,
+      "loss": 0.5637,
+      "step": 2055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893833980551604e-05,
+      "loss": 0.5756,
+      "step": 2056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893730360336138e-05,
+      "loss": 0.6149,
+      "step": 2057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989362668984773e-05,
+      "loss": 0.6093,
+      "step": 2058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893522969086904e-05,
+      "loss": 0.6207,
+      "step": 2059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893419198054196e-05,
+      "loss": 0.6393,
+      "step": 2060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893315376750127e-05,
+      "loss": 0.6341,
+      "step": 2061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989321150517522e-05,
+      "loss": 0.6246,
+      "step": 2062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9893107583330015e-05,
+      "loss": 0.6502,
+      "step": 2063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989300361121503e-05,
+      "loss": 0.6347,
+      "step": 2064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892899588830797e-05,
+      "loss": 0.5782,
+      "step": 2065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892795516177843e-05,
+      "loss": 0.5995,
+      "step": 2066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892691393256698e-05,
+      "loss": 0.6161,
+      "step": 2067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892587220067892e-05,
+      "loss": 0.6413,
+      "step": 2068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989248299661195e-05,
+      "loss": 0.6875,
+      "step": 2069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989237872288941e-05,
+      "loss": 0.5828,
+      "step": 2070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892274398900793e-05,
+      "loss": 0.6258,
+      "step": 2071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9892170024646633e-05,
+      "loss": 0.6684,
+      "step": 2072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989206560012746e-05,
+      "loss": 0.629,
+      "step": 2073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9891961125343805e-05,
+      "loss": 0.6289,
+      "step": 2074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9891856600296198e-05,
+      "loss": 0.6388,
+      "step": 2075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989175202498517e-05,
+      "loss": 0.6919,
+      "step": 2076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9891647399411257e-05,
+      "loss": 0.6334,
+      "step": 2077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989154272357498e-05,
+      "loss": 0.5846,
+      "step": 2078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989143799747688e-05,
+      "loss": 0.5932,
+      "step": 2079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989133322111749e-05,
+      "loss": 0.6462,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989122839449734e-05,
+      "loss": 0.6489,
+      "step": 2081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9891123517616955e-05,
+      "loss": 0.6115,
+      "step": 2082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9891018590476882e-05,
+      "loss": 0.6253,
+      "step": 2083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989091361307764e-05,
+      "loss": 0.578,
+      "step": 2084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890808585419775e-05,
+      "loss": 0.6268,
+      "step": 2085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890703507503813e-05,
+      "loss": 0.6949,
+      "step": 2086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890598379330285e-05,
+      "loss": 0.6431,
+      "step": 2087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890493200899735e-05,
+      "loss": 0.5927,
+      "step": 2088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890387972212694e-05,
+      "loss": 0.5892,
+      "step": 2089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989028269326969e-05,
+      "loss": 0.6209,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.989017736407127e-05,
+      "loss": 0.5987,
+      "step": 2091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9890071984617954e-05,
+      "loss": 0.606,
+      "step": 2092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889966554910292e-05,
+      "loss": 0.5884,
+      "step": 2093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889861074948813e-05,
+      "loss": 0.6348,
+      "step": 2094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988975554473405e-05,
+      "loss": 0.6572,
+      "step": 2095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988964996426655e-05,
+      "loss": 0.6525,
+      "step": 2096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889544333546837e-05,
+      "loss": 0.5589,
+      "step": 2097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889438652575454e-05,
+      "loss": 0.6436,
+      "step": 2098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889332921352933e-05,
+      "loss": 0.6268,
+      "step": 2099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988922713987982e-05,
+      "loss": 0.6463,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9889121308156644e-05,
+      "loss": 0.6807,
+      "step": 2101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988901542618395e-05,
+      "loss": 0.63,
+      "step": 2102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988890949396227e-05,
+      "loss": 0.6211,
+      "step": 2103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888803511492148e-05,
+      "loss": 0.573,
+      "step": 2104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888697478774115e-05,
+      "loss": 0.5987,
+      "step": 2105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888591395808715e-05,
+      "loss": 0.6128,
+      "step": 2106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988848526259649e-05,
+      "loss": 0.6606,
+      "step": 2107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988837907913797e-05,
+      "loss": 0.6413,
+      "step": 2108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888272845433706e-05,
+      "loss": 0.623,
+      "step": 2109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888166561484226e-05,
+      "loss": 0.6583,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9888060227290077e-05,
+      "loss": 0.6034,
+      "step": 2111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.98879538428518e-05,
+      "loss": 0.5558,
+      "step": 2112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887847408169932e-05,
+      "loss": 0.6337,
+      "step": 2113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887740923245017e-05,
+      "loss": 0.6229,
+      "step": 2114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887634388077593e-05,
+      "loss": 0.634,
+      "step": 2115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887527802668204e-05,
+      "loss": 0.5754,
+      "step": 2116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887421167017394e-05,
+      "loss": 0.6501,
+      "step": 2117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.98873144811257e-05,
+      "loss": 0.6402,
+      "step": 2118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988720774499366e-05,
+      "loss": 0.6085,
+      "step": 2119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9887100958621825e-05,
+      "loss": 0.6157,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886994122010734e-05,
+      "loss": 0.6711,
+      "step": 2121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886887235160933e-05,
+      "loss": 0.6315,
+      "step": 2122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988678029807296e-05,
+      "loss": 0.5701,
+      "step": 2123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886673310747362e-05,
+      "loss": 0.6143,
+      "step": 2124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988656627318468e-05,
+      "loss": 0.6105,
+      "step": 2125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988645918538546e-05,
+      "loss": 0.6337,
+      "step": 2126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886352047350245e-05,
+      "loss": 0.618,
+      "step": 2127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886244859079582e-05,
+      "loss": 0.6702,
+      "step": 2128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9886137620574014e-05,
+      "loss": 0.6194,
+      "step": 2129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.988603033183408e-05,
+      "loss": 0.6067,
+      "step": 2130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9885922992860337e-05,
+      "loss": 0.6063,
+      "step": 2131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988581560365332e-05,
+      "loss": 0.6624,
+      "step": 2132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988570816421358e-05,
+      "loss": 0.6139,
+      "step": 2133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9885600674541662e-05,
+      "loss": 0.6247,
+      "step": 2134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988549313463811e-05,
+      "loss": 0.6287,
+      "step": 2135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9885385544503475e-05,
+      "loss": 0.5812,
+      "step": 2136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98852779041383e-05,
+      "loss": 0.6293,
+      "step": 2137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9885170213543134e-05,
+      "loss": 0.5742,
+      "step": 2138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988506247271852e-05,
+      "loss": 0.6477,
+      "step": 2139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9884954681665012e-05,
+      "loss": 0.6527,
+      "step": 2140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988484684038315e-05,
+      "loss": 0.6468,
+      "step": 2141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988473894887349e-05,
+      "loss": 0.5858,
+      "step": 2142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9884631007136575e-05,
+      "loss": 0.6239,
+      "step": 2143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9884523015172957e-05,
+      "loss": 0.6211,
+      "step": 2144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9884414972983183e-05,
+      "loss": 0.6064,
+      "step": 2145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98843068805678e-05,
+      "loss": 0.5742,
+      "step": 2146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9884198737927355e-05,
+      "loss": 0.6121,
+      "step": 2147
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988409054506241e-05,
+      "loss": 0.6757,
+      "step": 2148
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98839823019735e-05,
+      "loss": 0.6371,
+      "step": 2149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883874008661183e-05,
+      "loss": 0.5844,
+      "step": 2150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883765665126007e-05,
+      "loss": 0.601,
+      "step": 2151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883657271368522e-05,
+      "loss": 0.576,
+      "step": 2152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883548827389282e-05,
+      "loss": 0.6036,
+      "step": 2153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883440333188836e-05,
+      "loss": 0.6361,
+      "step": 2154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883331788767734e-05,
+      "loss": 0.6149,
+      "step": 2155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988322319412653e-05,
+      "loss": 0.5738,
+      "step": 2156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883114549265773e-05,
+      "loss": 0.5867,
+      "step": 2157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9883005854186018e-05,
+      "loss": 0.6075,
+      "step": 2158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9882897108887814e-05,
+      "loss": 0.6108,
+      "step": 2159
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988278831337172e-05,
+      "loss": 0.6351,
+      "step": 2160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988267946763828e-05,
+      "loss": 0.5717,
+      "step": 2161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9882570571688052e-05,
+      "loss": 0.6005,
+      "step": 2162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988246162552159e-05,
+      "loss": 0.5527,
+      "step": 2163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9882352629139444e-05,
+      "loss": 0.6539,
+      "step": 2164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988224358254217e-05,
+      "loss": 0.6158,
+      "step": 2165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9882134485730326e-05,
+      "loss": 0.6029,
+      "step": 2166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988202533870446e-05,
+      "loss": 0.6461,
+      "step": 2167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881916141465128e-05,
+      "loss": 0.5672,
+      "step": 2168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881806894012887e-05,
+      "loss": 0.6359,
+      "step": 2169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988169759634829e-05,
+      "loss": 0.628,
+      "step": 2170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881588248471897e-05,
+      "loss": 0.5827,
+      "step": 2171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988147885038426e-05,
+      "loss": 0.6489,
+      "step": 2172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988136940208593e-05,
+      "loss": 0.6537,
+      "step": 2173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881259903577472e-05,
+      "loss": 0.6248,
+      "step": 2174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881150354859437e-05,
+      "loss": 0.5957,
+      "step": 2175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9881040755932383e-05,
+      "loss": 0.6733,
+      "step": 2176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988093110679687e-05,
+      "loss": 0.5646,
+      "step": 2177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988082140745345e-05,
+      "loss": 0.6837,
+      "step": 2178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9880711657902685e-05,
+      "loss": 0.6425,
+      "step": 2179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988060185814513e-05,
+      "loss": 0.6155,
+      "step": 2180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988049200818134e-05,
+      "loss": 0.6152,
+      "step": 2181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9880382108011877e-05,
+      "loss": 0.59,
+      "step": 2182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98802721576373e-05,
+      "loss": 0.7581,
+      "step": 2183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.988016215705817e-05,
+      "loss": 0.653,
+      "step": 2184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9880052106275037e-05,
+      "loss": 0.6736,
+      "step": 2185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879942005288468e-05,
+      "loss": 0.631,
+      "step": 2186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879831854099022e-05,
+      "loss": 0.6538,
+      "step": 2187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879721652707255e-05,
+      "loss": 0.6065,
+      "step": 2188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879611401113728e-05,
+      "loss": 0.6495,
+      "step": 2189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879501099319002e-05,
+      "loss": 0.6742,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9879390747323638e-05,
+      "loss": 0.6661,
+      "step": 2191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98792803451282e-05,
+      "loss": 0.5955,
+      "step": 2192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987916989273324e-05,
+      "loss": 0.645,
+      "step": 2193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987905939013933e-05,
+      "loss": 0.6381,
+      "step": 2194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878948837347022e-05,
+      "loss": 0.5673,
+      "step": 2195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878838234356884e-05,
+      "loss": 0.6614,
+      "step": 2196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878727581169472e-05,
+      "loss": 0.6625,
+      "step": 2197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878616877785356e-05,
+      "loss": 0.6297,
+      "step": 2198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878506124205092e-05,
+      "loss": 0.6334,
+      "step": 2199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878395320429246e-05,
+      "loss": 0.611,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878284466458383e-05,
+      "loss": 0.6452,
+      "step": 2201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987817356229306e-05,
+      "loss": 0.6689,
+      "step": 2202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9878062607933847e-05,
+      "loss": 0.5929,
+      "step": 2203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877951603381302e-05,
+      "loss": 0.636,
+      "step": 2204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987784054863599e-05,
+      "loss": 0.5923,
+      "step": 2205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877729443698484e-05,
+      "loss": 0.7031,
+      "step": 2206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987761828856934e-05,
+      "loss": 0.6047,
+      "step": 2207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877507083249122e-05,
+      "loss": 0.63,
+      "step": 2208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98773958277384e-05,
+      "loss": 0.6176,
+      "step": 2209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877284522037735e-05,
+      "loss": 0.6167,
+      "step": 2210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877173166147697e-05,
+      "loss": 0.5869,
+      "step": 2211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9877061760068847e-05,
+      "loss": 0.5812,
+      "step": 2212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876950303801753e-05,
+      "loss": 0.6478,
+      "step": 2213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876838797346987e-05,
+      "loss": 0.6189,
+      "step": 2214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876727240705106e-05,
+      "loss": 0.5983,
+      "step": 2215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876615633876682e-05,
+      "loss": 0.6018,
+      "step": 2216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876503976862285e-05,
+      "loss": 0.6413,
+      "step": 2217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876392269662476e-05,
+      "loss": 0.6503,
+      "step": 2218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876280512277822e-05,
+      "loss": 0.566,
+      "step": 2219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.98761687047089e-05,
+      "loss": 0.6185,
+      "step": 2220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9876056846956268e-05,
+      "loss": 0.6485,
+      "step": 2221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9875944939020503e-05,
+      "loss": 0.6191,
+      "step": 2222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9875832980902166e-05,
+      "loss": 0.6838,
+      "step": 2223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9875720972601833e-05,
+      "loss": 0.606,
+      "step": 2224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9875608914120066e-05,
+      "loss": 0.6861,
+      "step": 2225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9875496805457438e-05,
+      "loss": 0.6454,
+      "step": 2226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987538464661452e-05,
+      "loss": 0.5786,
+      "step": 2227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987527243759188e-05,
+      "loss": 0.6405,
+      "step": 2228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987516017839009e-05,
+      "loss": 0.6095,
+      "step": 2229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987504786900972e-05,
+      "loss": 0.616,
+      "step": 2230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874935509451335e-05,
+      "loss": 0.6333,
+      "step": 2231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987482309971551e-05,
+      "loss": 0.6258,
+      "step": 2232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874710639802822e-05,
+      "loss": 0.664,
+      "step": 2233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874598129713835e-05,
+      "loss": 0.6282,
+      "step": 2234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874485569449127e-05,
+      "loss": 0.6683,
+      "step": 2235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987437295900926e-05,
+      "loss": 0.5786,
+      "step": 2236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874260298394814e-05,
+      "loss": 0.6147,
+      "step": 2237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987414758760636e-05,
+      "loss": 0.6501,
+      "step": 2238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9874034826644474e-05,
+      "loss": 0.6311,
+      "step": 2239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987392201550972e-05,
+      "loss": 0.6268,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873809154202677e-05,
+      "loss": 0.6588,
+      "step": 2241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987369624272392e-05,
+      "loss": 0.6174,
+      "step": 2242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873583281074022e-05,
+      "loss": 0.6486,
+      "step": 2243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873470269253553e-05,
+      "loss": 0.5963,
+      "step": 2244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873357207263088e-05,
+      "loss": 0.5935,
+      "step": 2245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873244095103208e-05,
+      "loss": 0.6553,
+      "step": 2246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987313093277448e-05,
+      "loss": 0.6191,
+      "step": 2247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9873017720277486e-05,
+      "loss": 0.6366,
+      "step": 2248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872904457612794e-05,
+      "loss": 0.646,
+      "step": 2249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872791144780986e-05,
+      "loss": 0.5671,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872677781782632e-05,
+      "loss": 0.5948,
+      "step": 2251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987256436861831e-05,
+      "loss": 0.6373,
+      "step": 2252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872450905288604e-05,
+      "loss": 0.5481,
+      "step": 2253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872337391794077e-05,
+      "loss": 0.6743,
+      "step": 2254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9872223828135312e-05,
+      "loss": 0.6359,
+      "step": 2255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987211021431289e-05,
+      "loss": 0.602,
+      "step": 2256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9871996550327383e-05,
+      "loss": 0.6421,
+      "step": 2257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987188283617937e-05,
+      "loss": 0.6031,
+      "step": 2258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9871769071869435e-05,
+      "loss": 0.6271,
+      "step": 2259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9871655257398142e-05,
+      "loss": 0.6252,
+      "step": 2260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987154139276608e-05,
+      "loss": 0.63,
+      "step": 2261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987142747797383e-05,
+      "loss": 0.6172,
+      "step": 2262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987131351302196e-05,
+      "loss": 0.6192,
+      "step": 2263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9871199497911057e-05,
+      "loss": 0.5789,
+      "step": 2264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9871085432641698e-05,
+      "loss": 0.6145,
+      "step": 2265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870971317214463e-05,
+      "loss": 0.6717,
+      "step": 2266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870857151629935e-05,
+      "loss": 0.6845,
+      "step": 2267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870742935888686e-05,
+      "loss": 0.6127,
+      "step": 2268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870628669991307e-05,
+      "loss": 0.5697,
+      "step": 2269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987051435393837e-05,
+      "loss": 0.6087,
+      "step": 2270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870399987730455e-05,
+      "loss": 0.6316,
+      "step": 2271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870285571368152e-05,
+      "loss": 0.6803,
+      "step": 2272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9870171104852036e-05,
+      "loss": 0.6403,
+      "step": 2273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.987005658818269e-05,
+      "loss": 0.5897,
+      "step": 2274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9869942021360696e-05,
+      "loss": 0.5872,
+      "step": 2275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9869827404386636e-05,
+      "loss": 0.6003,
+      "step": 2276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9869712737261096e-05,
+      "loss": 0.6323,
+      "step": 2277
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986959801998465e-05,
+      "loss": 0.6367,
+      "step": 2278
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9869483252557886e-05,
+      "loss": 0.6101,
+      "step": 2279
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9869368434981392e-05,
+      "loss": 0.5773,
+      "step": 2280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986925356725574e-05,
+      "loss": 0.6466,
+      "step": 2281
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9869138649381526e-05,
+      "loss": 0.6347,
+      "step": 2282
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9869023681359325e-05,
+      "loss": 0.6484,
+      "step": 2283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868908663189725e-05,
+      "loss": 0.618,
+      "step": 2284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868793594873312e-05,
+      "loss": 0.6298,
+      "step": 2285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868678476410664e-05,
+      "loss": 0.6058,
+      "step": 2286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868563307802375e-05,
+      "loss": 0.6138,
+      "step": 2287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868448089049023e-05,
+      "loss": 0.5504,
+      "step": 2288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868332820151196e-05,
+      "loss": 0.6325,
+      "step": 2289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868217501109483e-05,
+      "loss": 0.5973,
+      "step": 2290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9868102131924463e-05,
+      "loss": 0.5731,
+      "step": 2291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867986712596728e-05,
+      "loss": 0.6432,
+      "step": 2292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867871243126863e-05,
+      "loss": 0.6552,
+      "step": 2293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867755723515455e-05,
+      "loss": 0.6412,
+      "step": 2294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986764015376309e-05,
+      "loss": 0.5899,
+      "step": 2295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867524533870353e-05,
+      "loss": 0.5696,
+      "step": 2296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867408863837834e-05,
+      "loss": 0.7069,
+      "step": 2297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986729314366612e-05,
+      "loss": 0.5519,
+      "step": 2298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9867177373355805e-05,
+      "loss": 0.5728,
+      "step": 2299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986706155290747e-05,
+      "loss": 0.6466,
+      "step": 2300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9866945682321703e-05,
+      "loss": 0.6365,
+      "step": 2301
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9866829761599097e-05,
+      "loss": 0.5748,
+      "step": 2302
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986671379074024e-05,
+      "loss": 0.5927,
+      "step": 2303
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986659776974572e-05,
+      "loss": 0.6791,
+      "step": 2304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9866481698616124e-05,
+      "loss": 0.6291,
+      "step": 2305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986636557735205e-05,
+      "loss": 0.5554,
+      "step": 2306
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986624940595408e-05,
+      "loss": 0.6752,
+      "step": 2307
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986613318442281e-05,
+      "loss": 0.5876,
+      "step": 2308
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9866016912758824e-05,
+      "loss": 0.605,
+      "step": 2309
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986590059096272e-05,
+      "loss": 0.613,
+      "step": 2310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9865784219035085e-05,
+      "loss": 0.6435,
+      "step": 2311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986566779697651e-05,
+      "loss": 0.6513,
+      "step": 2312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9865551324787585e-05,
+      "loss": 0.5997,
+      "step": 2313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9865434802468912e-05,
+      "loss": 0.5882,
+      "step": 2314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986531823002107e-05,
+      "loss": 0.673,
+      "step": 2315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9865201607444654e-05,
+      "loss": 0.6632,
+      "step": 2316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9865084934740265e-05,
+      "loss": 0.548,
+      "step": 2317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986496821190849e-05,
+      "loss": 0.5594,
+      "step": 2318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986485143894992e-05,
+      "loss": 0.6283,
+      "step": 2319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9864734615865153e-05,
+      "loss": 0.6084,
+      "step": 2320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986461774265478e-05,
+      "loss": 0.5668,
+      "step": 2321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9864500819319394e-05,
+      "loss": 0.6616,
+      "step": 2322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986438384585959e-05,
+      "loss": 0.5904,
+      "step": 2323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9864266822275966e-05,
+      "loss": 0.6454,
+      "step": 2324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986414974856911e-05,
+      "loss": 0.6859,
+      "step": 2325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9864032624739622e-05,
+      "loss": 0.5821,
+      "step": 2326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863915450788098e-05,
+      "loss": 0.617,
+      "step": 2327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863798226715123e-05,
+      "loss": 0.6043,
+      "step": 2328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863680952521305e-05,
+      "loss": 0.6448,
+      "step": 2329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863563628207236e-05,
+      "loss": 0.6562,
+      "step": 2330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863446253773513e-05,
+      "loss": 0.5879,
+      "step": 2331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986332882922073e-05,
+      "loss": 0.5708,
+      "step": 2332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9863211354549482e-05,
+      "loss": 0.6407,
+      "step": 2333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986309382976037e-05,
+      "loss": 0.5879,
+      "step": 2334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986297625485399e-05,
+      "loss": 0.6383,
+      "step": 2335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986285862983094e-05,
+      "loss": 0.5994,
+      "step": 2336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862740954691816e-05,
+      "loss": 0.5772,
+      "step": 2337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862623229437215e-05,
+      "loss": 0.6863,
+      "step": 2338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862505454067738e-05,
+      "loss": 0.5724,
+      "step": 2339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862387628583983e-05,
+      "loss": 0.5923,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862269752986547e-05,
+      "loss": 0.5672,
+      "step": 2341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862151827276033e-05,
+      "loss": 0.5899,
+      "step": 2342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9862033851453034e-05,
+      "loss": 0.6252,
+      "step": 2343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861915825518156e-05,
+      "loss": 0.6162,
+      "step": 2344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861797749471993e-05,
+      "loss": 0.5924,
+      "step": 2345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986167962331515e-05,
+      "loss": 0.6449,
+      "step": 2346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861561447048224e-05,
+      "loss": 0.6082,
+      "step": 2347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861443220671815e-05,
+      "loss": 0.7305,
+      "step": 2348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861324944186523e-05,
+      "loss": 0.5742,
+      "step": 2349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986120661759295e-05,
+      "loss": 0.5653,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9861088240891704e-05,
+      "loss": 0.5713,
+      "step": 2351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9860969814083376e-05,
+      "loss": 0.6115,
+      "step": 2352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9860851337168577e-05,
+      "loss": 0.6433,
+      "step": 2353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.98607328101479e-05,
+      "loss": 0.6243,
+      "step": 2354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9860614233021955e-05,
+      "loss": 0.6387,
+      "step": 2355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986049560579134e-05,
+      "loss": 0.5666,
+      "step": 2356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986037692845666e-05,
+      "loss": 0.6352,
+      "step": 2357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9860258201018515e-05,
+      "loss": 0.6133,
+      "step": 2358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.986013942347751e-05,
+      "loss": 0.5832,
+      "step": 2359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9860020595834252e-05,
+      "loss": 0.6215,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985990171808934e-05,
+      "loss": 0.6444,
+      "step": 2361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985978279024338e-05,
+      "loss": 0.6004,
+      "step": 2362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859663812296974e-05,
+      "loss": 0.6511,
+      "step": 2363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859544784250732e-05,
+      "loss": 0.6446,
+      "step": 2364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859425706105252e-05,
+      "loss": 0.5515,
+      "step": 2365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859306577861145e-05,
+      "loss": 0.6278,
+      "step": 2366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859187399519013e-05,
+      "loss": 0.614,
+      "step": 2367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9859068171079463e-05,
+      "loss": 0.6013,
+      "step": 2368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.98589488925431e-05,
+      "loss": 0.6433,
+      "step": 2369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858829563910528e-05,
+      "loss": 0.7136,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985871018518236e-05,
+      "loss": 0.5936,
+      "step": 2371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858590756359194e-05,
+      "loss": 0.5857,
+      "step": 2372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858471277441644e-05,
+      "loss": 0.5534,
+      "step": 2373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858351748430315e-05,
+      "loss": 0.6584,
+      "step": 2374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858232169325813e-05,
+      "loss": 0.5479,
+      "step": 2375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9858112540128745e-05,
+      "loss": 0.6233,
+      "step": 2376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985799286083972e-05,
+      "loss": 0.64,
+      "step": 2377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985787313145935e-05,
+      "loss": 0.565,
+      "step": 2378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9857753351988234e-05,
+      "loss": 0.6522,
+      "step": 2379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985763352242699e-05,
+      "loss": 0.5586,
+      "step": 2380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985751364277622e-05,
+      "loss": 0.6449,
+      "step": 2381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9857393713036538e-05,
+      "loss": 0.665,
+      "step": 2382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9857273733208555e-05,
+      "loss": 0.6536,
+      "step": 2383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9857153703292874e-05,
+      "loss": 0.6629,
+      "step": 2384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9857033623290106e-05,
+      "loss": 0.5569,
+      "step": 2385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856913493200863e-05,
+      "loss": 0.6307,
+      "step": 2386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856793313025762e-05,
+      "loss": 0.608,
+      "step": 2387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856673082765404e-05,
+      "loss": 0.6182,
+      "step": 2388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856552802420404e-05,
+      "loss": 0.6611,
+      "step": 2389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985643247199137e-05,
+      "loss": 0.6253,
+      "step": 2390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985631209147892e-05,
+      "loss": 0.5951,
+      "step": 2391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856191660883657e-05,
+      "loss": 0.621,
+      "step": 2392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9856071180206204e-05,
+      "loss": 0.6269,
+      "step": 2393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985595064944716e-05,
+      "loss": 0.5683,
+      "step": 2394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985583006860715e-05,
+      "loss": 0.6197,
+      "step": 2395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9855709437686777e-05,
+      "loss": 0.6362,
+      "step": 2396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985558875668666e-05,
+      "loss": 0.5923,
+      "step": 2397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985546802560741e-05,
+      "loss": 0.6657,
+      "step": 2398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985534724444964e-05,
+      "loss": 0.5799,
+      "step": 2399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9855226413213965e-05,
+      "loss": 0.6113,
+      "step": 2400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9855105531900996e-05,
+      "loss": 0.6446,
+      "step": 2401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854984600511352e-05,
+      "loss": 0.6752,
+      "step": 2402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854863619045645e-05,
+      "loss": 0.6032,
+      "step": 2403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985474258750449e-05,
+      "loss": 0.6286,
+      "step": 2404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854621505888502e-05,
+      "loss": 0.5965,
+      "step": 2405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854500374198296e-05,
+      "loss": 0.5858,
+      "step": 2406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854379192434488e-05,
+      "loss": 0.6192,
+      "step": 2407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985425796059769e-05,
+      "loss": 0.6042,
+      "step": 2408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9854136678688526e-05,
+      "loss": 0.6554,
+      "step": 2409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.98540153467076e-05,
+      "loss": 0.6277,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9853893964655545e-05,
+      "loss": 0.5966,
+      "step": 2411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9853772532532964e-05,
+      "loss": 0.611,
+      "step": 2412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985365105034048e-05,
+      "loss": 0.6619,
+      "step": 2413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985352951807871e-05,
+      "loss": 0.6653,
+      "step": 2414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9853407935748272e-05,
+      "loss": 0.6514,
+      "step": 2415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985328630334978e-05,
+      "loss": 0.6666,
+      "step": 2416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9853164620883854e-05,
+      "loss": 0.6687,
+      "step": 2417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985304288835112e-05,
+      "loss": 0.5884,
+      "step": 2418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.985292110575218e-05,
+      "loss": 0.5265,
+      "step": 2419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9852799273087665e-05,
+      "loss": 0.653,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9852677390358194e-05,
+      "loss": 0.5973,
+      "step": 2421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9852555457564383e-05,
+      "loss": 0.5729,
+      "step": 2422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9852433474706852e-05,
+      "loss": 0.5957,
+      "step": 2423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9852311441786222e-05,
+      "loss": 0.6843,
+      "step": 2424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985218935880311e-05,
+      "loss": 0.5996,
+      "step": 2425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985206722575814e-05,
+      "loss": 0.6533,
+      "step": 2426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985194504265193e-05,
+      "loss": 0.5606,
+      "step": 2427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98518228094851e-05,
+      "loss": 0.6514,
+      "step": 2428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985170052625828e-05,
+      "loss": 0.6673,
+      "step": 2429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9851578192972077e-05,
+      "loss": 0.5961,
+      "step": 2430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9851455809627123e-05,
+      "loss": 0.6459,
+      "step": 2431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9851333376224037e-05,
+      "loss": 0.642,
+      "step": 2432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985121089276344e-05,
+      "loss": 0.637,
+      "step": 2433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985108835924596e-05,
+      "loss": 0.5443,
+      "step": 2434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9850965775672208e-05,
+      "loss": 0.5817,
+      "step": 2435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9850843142042818e-05,
+      "loss": 0.623,
+      "step": 2436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985072045835841e-05,
+      "loss": 0.5836,
+      "step": 2437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98505977246196e-05,
+      "loss": 0.6226,
+      "step": 2438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9850474940827024e-05,
+      "loss": 0.6566,
+      "step": 2439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9850352106981302e-05,
+      "loss": 0.6543,
+      "step": 2440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.985022922308305e-05,
+      "loss": 0.6471,
+      "step": 2441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9850106289132902e-05,
+      "loss": 0.6396,
+      "step": 2442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9849983305131478e-05,
+      "loss": 0.6588,
+      "step": 2443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9849860271079405e-05,
+      "loss": 0.6331,
+      "step": 2444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984973718697731e-05,
+      "loss": 0.6216,
+      "step": 2445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984961405282581e-05,
+      "loss": 0.6684,
+      "step": 2446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984949086862554e-05,
+      "loss": 0.6724,
+      "step": 2447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9849367634377123e-05,
+      "loss": 0.6261,
+      "step": 2448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984924435008118e-05,
+      "loss": 0.6993,
+      "step": 2449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9849121015738347e-05,
+      "loss": 0.611,
+      "step": 2450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848997631349242e-05,
+      "loss": 0.6119,
+      "step": 2451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848874196914497e-05,
+      "loss": 0.6151,
+      "step": 2452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848750712434742e-05,
+      "loss": 0.6313,
+      "step": 2453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848627177910592e-05,
+      "loss": 0.5708,
+      "step": 2454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984850359334269e-05,
+      "loss": 0.6768,
+      "step": 2455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848379958731653e-05,
+      "loss": 0.6019,
+      "step": 2456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848256274078114e-05,
+      "loss": 0.5752,
+      "step": 2457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98481325393827e-05,
+      "loss": 0.6056,
+      "step": 2458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9848008754646043e-05,
+      "loss": 0.5419,
+      "step": 2459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9847884919868768e-05,
+      "loss": 0.6166,
+      "step": 2460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9847761035051507e-05,
+      "loss": 0.6038,
+      "step": 2461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9847637100194888e-05,
+      "loss": 0.5681,
+      "step": 2462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984751311529954e-05,
+      "loss": 0.6301,
+      "step": 2463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9847389080366092e-05,
+      "loss": 0.6582,
+      "step": 2464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9847264995395176e-05,
+      "loss": 0.6199,
+      "step": 2465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984714086038743e-05,
+      "loss": 0.644,
+      "step": 2466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984701667534347e-05,
+      "loss": 0.6104,
+      "step": 2467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846892440263934e-05,
+      "loss": 0.6671,
+      "step": 2468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984676815514946e-05,
+      "loss": 0.6324,
+      "step": 2469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846643820000665e-05,
+      "loss": 0.5899,
+      "step": 2470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846519434818193e-05,
+      "loss": 0.6143,
+      "step": 2471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846394999602674e-05,
+      "loss": 0.5904,
+      "step": 2472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846270514354734e-05,
+      "loss": 0.5732,
+      "step": 2473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846145979075015e-05,
+      "loss": 0.6384,
+      "step": 2474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9846021393764143e-05,
+      "loss": 0.6094,
+      "step": 2475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845896758422752e-05,
+      "loss": 0.5825,
+      "step": 2476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845772073051478e-05,
+      "loss": 0.6746,
+      "step": 2477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984564733765095e-05,
+      "loss": 0.6234,
+      "step": 2478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845522552221804e-05,
+      "loss": 0.6504,
+      "step": 2479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845397716764677e-05,
+      "loss": 0.5522,
+      "step": 2480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98452728312802e-05,
+      "loss": 0.5767,
+      "step": 2481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845147895769007e-05,
+      "loss": 0.6533,
+      "step": 2482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9845022910231733e-05,
+      "loss": 0.6774,
+      "step": 2483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984489787466902e-05,
+      "loss": 0.5738,
+      "step": 2484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984477278908149e-05,
+      "loss": 0.6104,
+      "step": 2485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9844647653469793e-05,
+      "loss": 0.656,
+      "step": 2486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9844522467834555e-05,
+      "loss": 0.5905,
+      "step": 2487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9844397232176414e-05,
+      "loss": 0.5752,
+      "step": 2488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984427194649601e-05,
+      "loss": 0.6908,
+      "step": 2489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984414661079398e-05,
+      "loss": 0.5795,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9844021225070955e-05,
+      "loss": 0.5951,
+      "step": 2491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843895789327572e-05,
+      "loss": 0.6306,
+      "step": 2492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843770303564475e-05,
+      "loss": 0.5987,
+      "step": 2493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843644767782296e-05,
+      "loss": 0.6208,
+      "step": 2494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843519181981677e-05,
+      "loss": 0.645,
+      "step": 2495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843393546163258e-05,
+      "loss": 0.6082,
+      "step": 2496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843267860327668e-05,
+      "loss": 0.5678,
+      "step": 2497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9843142124475552e-05,
+      "loss": 0.6277,
+      "step": 2498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984301633860755e-05,
+      "loss": 0.6795,
+      "step": 2499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98428905027243e-05,
+      "loss": 0.6217,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984276461682644e-05,
+      "loss": 0.652,
+      "step": 2501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9842638680914607e-05,
+      "loss": 0.6701,
+      "step": 2502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9842512694989447e-05,
+      "loss": 0.7348,
+      "step": 2503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.98423866590516e-05,
+      "loss": 0.6156,
+      "step": 2504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9842260573101698e-05,
+      "loss": 0.6664,
+      "step": 2505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9842134437140393e-05,
+      "loss": 0.6513,
+      "step": 2506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9842008251168318e-05,
+      "loss": 0.5905,
+      "step": 2507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984188201518612e-05,
+      "loss": 0.6379,
+      "step": 2508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9841755729194432e-05,
+      "loss": 0.6281,
+      "step": 2509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9841629393193903e-05,
+      "loss": 0.6475,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984150300718517e-05,
+      "loss": 0.6368,
+      "step": 2511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9841376571168882e-05,
+      "loss": 0.6744,
+      "step": 2512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9841250085145675e-05,
+      "loss": 0.6017,
+      "step": 2513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9841123549116198e-05,
+      "loss": 0.5833,
+      "step": 2514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840996963081087e-05,
+      "loss": 0.5984,
+      "step": 2515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984087032704099e-05,
+      "loss": 0.6626,
+      "step": 2516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840743640996546e-05,
+      "loss": 0.641,
+      "step": 2517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840616904948403e-05,
+      "loss": 0.6352,
+      "step": 2518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840490118897205e-05,
+      "loss": 0.6264,
+      "step": 2519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840363282843596e-05,
+      "loss": 0.6148,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.984023639678822e-05,
+      "loss": 0.688,
+      "step": 2521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9840109460731715e-05,
+      "loss": 0.6631,
+      "step": 2522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9839982474674733e-05,
+      "loss": 0.6153,
+      "step": 2523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9839855438617922e-05,
+      "loss": 0.615,
+      "step": 2524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983972835256192e-05,
+      "loss": 0.64,
+      "step": 2525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983960121650738e-05,
+      "loss": 0.6635,
+      "step": 2526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9839474030454945e-05,
+      "loss": 0.6202,
+      "step": 2527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983934679440526e-05,
+      "loss": 0.6071,
+      "step": 2528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983921950835897e-05,
+      "loss": 0.6657,
+      "step": 2529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9839092172316726e-05,
+      "loss": 0.6294,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9838964786279173e-05,
+      "loss": 0.6111,
+      "step": 2531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9838837350246957e-05,
+      "loss": 0.6394,
+      "step": 2532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9838709864220728e-05,
+      "loss": 0.6406,
+      "step": 2533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983858232820113e-05,
+      "loss": 0.715,
+      "step": 2534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983845474218882e-05,
+      "loss": 0.602,
+      "step": 2535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9838327106184436e-05,
+      "loss": 0.5863,
+      "step": 2536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983819942018863e-05,
+      "loss": 0.5853,
+      "step": 2537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9838071684202057e-05,
+      "loss": 0.6399,
+      "step": 2538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9837943898225355e-05,
+      "loss": 0.586,
+      "step": 2539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983781606225918e-05,
+      "loss": 0.6294,
+      "step": 2540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983768817630418e-05,
+      "loss": 0.574,
+      "step": 2541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9837560240361007e-05,
+      "loss": 0.61,
+      "step": 2542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983743225443031e-05,
+      "loss": 0.6561,
+      "step": 2543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9837304218512737e-05,
+      "loss": 0.6158,
+      "step": 2544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983717613260894e-05,
+      "loss": 0.6236,
+      "step": 2545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983704799671957e-05,
+      "loss": 0.6162,
+      "step": 2546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983691981084528e-05,
+      "loss": 0.6377,
+      "step": 2547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983679157498672e-05,
+      "loss": 0.6065,
+      "step": 2548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9836663289144537e-05,
+      "loss": 0.5942,
+      "step": 2549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9836534953319392e-05,
+      "loss": 0.6509,
+      "step": 2550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9836406567511927e-05,
+      "loss": 0.6061,
+      "step": 2551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9836278131722803e-05,
+      "loss": 0.6526,
+      "step": 2552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983614964595267e-05,
+      "loss": 0.6529,
+      "step": 2553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9836021110202177e-05,
+      "loss": 0.6324,
+      "step": 2554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983589252447198e-05,
+      "loss": 0.5695,
+      "step": 2555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835763888762736e-05,
+      "loss": 0.6534,
+      "step": 2556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835635203075095e-05,
+      "loss": 0.6282,
+      "step": 2557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835506467409707e-05,
+      "loss": 0.6236,
+      "step": 2558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835377681767235e-05,
+      "loss": 0.6365,
+      "step": 2559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835248846148324e-05,
+      "loss": 0.6226,
+      "step": 2560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9835119960553636e-05,
+      "loss": 0.6732,
+      "step": 2561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983499102498382e-05,
+      "loss": 0.6152,
+      "step": 2562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983486203943954e-05,
+      "loss": 0.5894,
+      "step": 2563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983473300392144e-05,
+      "loss": 0.6245,
+      "step": 2564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9834603918430184e-05,
+      "loss": 0.6768,
+      "step": 2565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9834474782966425e-05,
+      "loss": 0.6179,
+      "step": 2566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983434559753082e-05,
+      "loss": 0.5645,
+      "step": 2567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9834216362124025e-05,
+      "loss": 0.6375,
+      "step": 2568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9834087076746696e-05,
+      "loss": 0.6233,
+      "step": 2569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.983395774139949e-05,
+      "loss": 0.5764,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9833828356083066e-05,
+      "loss": 0.6166,
+      "step": 2571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9833698920798076e-05,
+      "loss": 0.6298,
+      "step": 2572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9833569435545186e-05,
+      "loss": 0.6476,
+      "step": 2573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983343990032505e-05,
+      "loss": 0.652,
+      "step": 2574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9833310315138326e-05,
+      "loss": 0.6237,
+      "step": 2575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983318067998567e-05,
+      "loss": 0.6043,
+      "step": 2576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9833050994867747e-05,
+      "loss": 0.6457,
+      "step": 2577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9832921259785207e-05,
+      "loss": 0.6407,
+      "step": 2578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983279147473872e-05,
+      "loss": 0.5689,
+      "step": 2579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9832661639728936e-05,
+      "loss": 0.6405,
+      "step": 2580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983253175475652e-05,
+      "loss": 0.6037,
+      "step": 2581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983240181982213e-05,
+      "loss": 0.6355,
+      "step": 2582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9832271834926428e-05,
+      "loss": 0.6169,
+      "step": 2583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983214180007007e-05,
+      "loss": 0.6593,
+      "step": 2584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9832011715253722e-05,
+      "loss": 0.6571,
+      "step": 2585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9831881580478045e-05,
+      "loss": 0.6388,
+      "step": 2586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9831751395743698e-05,
+      "loss": 0.5796,
+      "step": 2587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983162116105134e-05,
+      "loss": 0.6256,
+      "step": 2588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9831490876401637e-05,
+      "loss": 0.5771,
+      "step": 2589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983136054179525e-05,
+      "loss": 0.6052,
+      "step": 2590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983123015723284e-05,
+      "loss": 0.6061,
+      "step": 2591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983109972271507e-05,
+      "loss": 0.6376,
+      "step": 2592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830969238242605e-05,
+      "loss": 0.5932,
+      "step": 2593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830838703816105e-05,
+      "loss": 0.6302,
+      "step": 2594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983070811943623e-05,
+      "loss": 0.6118,
+      "step": 2595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830577485103656e-05,
+      "loss": 0.6288,
+      "step": 2596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830446800819035e-05,
+      "loss": 0.6195,
+      "step": 2597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830316066583037e-05,
+      "loss": 0.591,
+      "step": 2598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.983018528239632e-05,
+      "loss": 0.636,
+      "step": 2599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9830054448259555e-05,
+      "loss": 0.5689,
+      "step": 2600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829923564173407e-05,
+      "loss": 0.651,
+      "step": 2601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829792630138532e-05,
+      "loss": 0.5587,
+      "step": 2602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982966164615561e-05,
+      "loss": 0.6096,
+      "step": 2603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829530612225293e-05,
+      "loss": 0.6693,
+      "step": 2604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829399528348254e-05,
+      "loss": 0.6012,
+      "step": 2605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829268394525157e-05,
+      "loss": 0.6427,
+      "step": 2606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982913721075667e-05,
+      "loss": 0.602,
+      "step": 2607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9829005977043455e-05,
+      "loss": 0.6039,
+      "step": 2608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828874693386183e-05,
+      "loss": 0.656,
+      "step": 2609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982874335978552e-05,
+      "loss": 0.6065,
+      "step": 2610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828611976242133e-05,
+      "loss": 0.6342,
+      "step": 2611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828480542756693e-05,
+      "loss": 0.6253,
+      "step": 2612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828349059329862e-05,
+      "loss": 0.6226,
+      "step": 2613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828217525962314e-05,
+      "loss": 0.5712,
+      "step": 2614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9828085942654713e-05,
+      "loss": 0.6394,
+      "step": 2615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982795430940773e-05,
+      "loss": 0.622,
+      "step": 2616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982782262622203e-05,
+      "loss": 0.605,
+      "step": 2617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9827690893098288e-05,
+      "loss": 0.6671,
+      "step": 2618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982755911003717e-05,
+      "loss": 0.6433,
+      "step": 2619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9827427277039347e-05,
+      "loss": 0.6951,
+      "step": 2620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9827295394105483e-05,
+      "loss": 0.6643,
+      "step": 2621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9827163461236258e-05,
+      "loss": 0.6169,
+      "step": 2622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9827031478432333e-05,
+      "loss": 0.6881,
+      "step": 2623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982689944569439e-05,
+      "loss": 0.5976,
+      "step": 2624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9826767363023086e-05,
+      "loss": 0.6848,
+      "step": 2625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9826635230419104e-05,
+      "loss": 0.5885,
+      "step": 2626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9826503047883107e-05,
+      "loss": 0.6433,
+      "step": 2627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982637081541577e-05,
+      "loss": 0.6145,
+      "step": 2628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9826238533017766e-05,
+      "loss": 0.6401,
+      "step": 2629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982610620068977e-05,
+      "loss": 0.681,
+      "step": 2630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982597381843245e-05,
+      "loss": 0.613,
+      "step": 2631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9825841386246476e-05,
+      "loss": 0.6491,
+      "step": 2632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9825708904132523e-05,
+      "loss": 0.6032,
+      "step": 2633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982557637209127e-05,
+      "loss": 0.5857,
+      "step": 2634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982544379012338e-05,
+      "loss": 0.6022,
+      "step": 2635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9825311158229537e-05,
+      "loss": 0.6378,
+      "step": 2636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9825178476410413e-05,
+      "loss": 0.6079,
+      "step": 2637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9825045744666675e-05,
+      "loss": 0.5925,
+      "step": 2638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824912962999005e-05,
+      "loss": 0.6366,
+      "step": 2639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824780131408074e-05,
+      "loss": 0.6213,
+      "step": 2640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824647249894555e-05,
+      "loss": 0.6794,
+      "step": 2641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982451431845913e-05,
+      "loss": 0.5978,
+      "step": 2642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824381337102468e-05,
+      "loss": 0.556,
+      "step": 2643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824248305825247e-05,
+      "loss": 0.7332,
+      "step": 2644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9824115224628147e-05,
+      "loss": 0.6028,
+      "step": 2645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823982093511837e-05,
+      "loss": 0.5787,
+      "step": 2646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823848912477e-05,
+      "loss": 0.6317,
+      "step": 2647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823715681524304e-05,
+      "loss": 0.6212,
+      "step": 2648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823582400654436e-05,
+      "loss": 0.6123,
+      "step": 2649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982344906986807e-05,
+      "loss": 0.5867,
+      "step": 2650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982331568916588e-05,
+      "loss": 0.6115,
+      "step": 2651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823182258548544e-05,
+      "loss": 0.6133,
+      "step": 2652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9823048778016744e-05,
+      "loss": 0.618,
+      "step": 2653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822915247571156e-05,
+      "loss": 0.6049,
+      "step": 2654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982278166721246e-05,
+      "loss": 0.6023,
+      "step": 2655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822648036941333e-05,
+      "loss": 0.6477,
+      "step": 2656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822514356758457e-05,
+      "loss": 0.6475,
+      "step": 2657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822380626664506e-05,
+      "loss": 0.6295,
+      "step": 2658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822246846660162e-05,
+      "loss": 0.5832,
+      "step": 2659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9822113016746106e-05,
+      "loss": 0.5977,
+      "step": 2660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982197913692302e-05,
+      "loss": 0.6901,
+      "step": 2661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821845207191578e-05,
+      "loss": 0.6584,
+      "step": 2662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821711227552464e-05,
+      "loss": 0.649,
+      "step": 2663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982157719800636e-05,
+      "loss": 0.6035,
+      "step": 2664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821443118553947e-05,
+      "loss": 0.7077,
+      "step": 2665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821308989195902e-05,
+      "loss": 0.54,
+      "step": 2666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821174809932912e-05,
+      "loss": 0.5986,
+      "step": 2667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9821040580765655e-05,
+      "loss": 0.6242,
+      "step": 2668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9820906301694818e-05,
+      "loss": 0.5917,
+      "step": 2669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982077197272108e-05,
+      "loss": 0.6382,
+      "step": 2670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982063759384512e-05,
+      "loss": 0.6558,
+      "step": 2671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9820503165067625e-05,
+      "loss": 0.6308,
+      "step": 2672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9820368686389276e-05,
+      "loss": 0.5793,
+      "step": 2673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.982023415781076e-05,
+      "loss": 0.603,
+      "step": 2674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9820099579332756e-05,
+      "loss": 0.5915,
+      "step": 2675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819964950955953e-05,
+      "loss": 0.6299,
+      "step": 2676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819830272681032e-05,
+      "loss": 0.5968,
+      "step": 2677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819695544508678e-05,
+      "loss": 0.5643,
+      "step": 2678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981956076643957e-05,
+      "loss": 0.7023,
+      "step": 2679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819425938474403e-05,
+      "loss": 0.6096,
+      "step": 2680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819291060613858e-05,
+      "loss": 0.5831,
+      "step": 2681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819156132858613e-05,
+      "loss": 0.6451,
+      "step": 2682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9819021155209367e-05,
+      "loss": 0.6205,
+      "step": 2683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9818886127666794e-05,
+      "loss": 0.598,
+      "step": 2684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9818751050231588e-05,
+      "loss": 0.6708,
+      "step": 2685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981861592290443e-05,
+      "loss": 0.6374,
+      "step": 2686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981848074568601e-05,
+      "loss": 0.6508,
+      "step": 2687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9818345518577016e-05,
+      "loss": 0.602,
+      "step": 2688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981821024157813e-05,
+      "loss": 0.6219,
+      "step": 2689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9818074914690044e-05,
+      "loss": 0.6179,
+      "step": 2690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9817939537913446e-05,
+      "loss": 0.6399,
+      "step": 2691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981780411124902e-05,
+      "loss": 0.5881,
+      "step": 2692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981766863469745e-05,
+      "loss": 0.5974,
+      "step": 2693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981753310825944e-05,
+      "loss": 0.642,
+      "step": 2694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9817397531935664e-05,
+      "loss": 0.5977,
+      "step": 2695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9817261905726815e-05,
+      "loss": 0.6125,
+      "step": 2696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9817126229633585e-05,
+      "loss": 0.6398,
+      "step": 2697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9816990503656662e-05,
+      "loss": 0.563,
+      "step": 2698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9816854727796733e-05,
+      "loss": 0.6803,
+      "step": 2699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9816718902054495e-05,
+      "loss": 0.6208,
+      "step": 2700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981658302643063e-05,
+      "loss": 0.655,
+      "step": 2701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981644710092583e-05,
+      "loss": 0.6253,
+      "step": 2702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981631112554079e-05,
+      "loss": 0.5904,
+      "step": 2703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.98161751002762e-05,
+      "loss": 0.6158,
+      "step": 2704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9816039025132746e-05,
+      "loss": 0.6379,
+      "step": 2705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815902900111125e-05,
+      "loss": 0.6204,
+      "step": 2706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815766725212023e-05,
+      "loss": 0.6079,
+      "step": 2707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815630500436138e-05,
+      "loss": 0.6311,
+      "step": 2708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981549422578416e-05,
+      "loss": 0.6346,
+      "step": 2709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815357901256784e-05,
+      "loss": 0.6186,
+      "step": 2710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815221526854697e-05,
+      "loss": 0.5957,
+      "step": 2711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9815085102578595e-05,
+      "loss": 0.5796,
+      "step": 2712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.981494862842917e-05,
+      "loss": 0.658,
+      "step": 2713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9814812104407116e-05,
+      "loss": 0.633,
+      "step": 2714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9814675530513134e-05,
+      "loss": 0.6318,
+      "step": 2715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.98145389067479e-05,
+      "loss": 0.6277,
+      "step": 2716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9814402233112127e-05,
+      "loss": 0.6108,
+      "step": 2717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.9814265509606506e-05,
+      "loss": 0.6755,
+      "step": 2718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9814128736231718e-05,
+      "loss": 0.6041,
+      "step": 2719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9813991912988475e-05,
+      "loss": 0.6171,
+      "step": 2720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9813855039877465e-05,
+      "loss": 0.6173,
+      "step": 2721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981371811689938e-05,
+      "loss": 0.621,
+      "step": 2722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9813581144054918e-05,
+      "loss": 0.6406,
+      "step": 2723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9813444121344777e-05,
+      "loss": 0.6369,
+      "step": 2724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981330704876965e-05,
+      "loss": 0.5955,
+      "step": 2725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9813169926330244e-05,
+      "loss": 0.6883,
+      "step": 2726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981303275402724e-05,
+      "loss": 0.6021,
+      "step": 2727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9812895531861347e-05,
+      "loss": 0.6864,
+      "step": 2728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9812758259833258e-05,
+      "loss": 0.6021,
+      "step": 2729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9812620937943667e-05,
+      "loss": 0.5712,
+      "step": 2730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981248356619328e-05,
+      "loss": 0.596,
+      "step": 2731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9812346144582787e-05,
+      "loss": 0.5485,
+      "step": 2732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981220867311289e-05,
+      "loss": 0.5949,
+      "step": 2733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9812071151784286e-05,
+      "loss": 0.6053,
+      "step": 2734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811933580597676e-05,
+      "loss": 0.6384,
+      "step": 2735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811795959553758e-05,
+      "loss": 0.5811,
+      "step": 2736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811658288653233e-05,
+      "loss": 0.5789,
+      "step": 2737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811520567896797e-05,
+      "loss": 0.6639,
+      "step": 2738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811382797285153e-05,
+      "loss": 0.6532,
+      "step": 2739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9811244976818996e-05,
+      "loss": 0.5919,
+      "step": 2740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981110710649903e-05,
+      "loss": 0.5954,
+      "step": 2741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981096918632596e-05,
+      "loss": 0.621,
+      "step": 2742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981083121630048e-05,
+      "loss": 0.5728,
+      "step": 2743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9810693196423296e-05,
+      "loss": 0.6272,
+      "step": 2744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9810555126695102e-05,
+      "loss": 0.6031,
+      "step": 2745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9810417007116607e-05,
+      "loss": 0.617,
+      "step": 2746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981027883768851e-05,
+      "loss": 0.6072,
+      "step": 2747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9810140618411512e-05,
+      "loss": 0.6317,
+      "step": 2748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.981000234928632e-05,
+      "loss": 0.5938,
+      "step": 2749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980986403031363e-05,
+      "loss": 0.6027,
+      "step": 2750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9809725661494147e-05,
+      "loss": 0.6215,
+      "step": 2751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9809587242828573e-05,
+      "loss": 0.611,
+      "step": 2752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9809448774317617e-05,
+      "loss": 0.6289,
+      "step": 2753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9809310255961977e-05,
+      "loss": 0.6226,
+      "step": 2754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980917168776236e-05,
+      "loss": 0.6596,
+      "step": 2755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980903306971947e-05,
+      "loss": 0.6341,
+      "step": 2756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980889440183401e-05,
+      "loss": 0.6694,
+      "step": 2757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980875568410668e-05,
+      "loss": 0.6063,
+      "step": 2758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9808616916538197e-05,
+      "loss": 0.6857,
+      "step": 2759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9808478099129257e-05,
+      "loss": 0.6455,
+      "step": 2760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9808339231880563e-05,
+      "loss": 0.6078,
+      "step": 2761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980820031479283e-05,
+      "loss": 0.6569,
+      "step": 2762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9808061347866756e-05,
+      "loss": 0.6075,
+      "step": 2763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9807922331103052e-05,
+      "loss": 0.5815,
+      "step": 2764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980778326450242e-05,
+      "loss": 0.6357,
+      "step": 2765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9807644148065567e-05,
+      "loss": 0.6007,
+      "step": 2766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9807504981793206e-05,
+      "loss": 0.6627,
+      "step": 2767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9807365765686038e-05,
+      "loss": 0.6127,
+      "step": 2768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980722649974477e-05,
+      "loss": 0.5799,
+      "step": 2769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9807087183970118e-05,
+      "loss": 0.6147,
+      "step": 2770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980694781836278e-05,
+      "loss": 0.6681,
+      "step": 2771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9806808402923466e-05,
+      "loss": 0.653,
+      "step": 2772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980666893765289e-05,
+      "loss": 0.6198,
+      "step": 2773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9806529422551757e-05,
+      "loss": 0.6207,
+      "step": 2774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9806389857620775e-05,
+      "loss": 0.6572,
+      "step": 2775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9806250242860653e-05,
+      "loss": 0.614,
+      "step": 2776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9806110578272104e-05,
+      "loss": 0.591,
+      "step": 2777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805970863855836e-05,
+      "loss": 0.6374,
+      "step": 2778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805831099612556e-05,
+      "loss": 0.545,
+      "step": 2779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805691285542976e-05,
+      "loss": 0.6916,
+      "step": 2780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980555142164781e-05,
+      "loss": 0.61,
+      "step": 2781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805411507927757e-05,
+      "loss": 0.673,
+      "step": 2782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805271544383548e-05,
+      "loss": 0.6253,
+      "step": 2783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9805131531015874e-05,
+      "loss": 0.6284,
+      "step": 2784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9804991467825458e-05,
+      "loss": 0.5755,
+      "step": 2785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980485135481301e-05,
+      "loss": 0.6096,
+      "step": 2786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9804711191979242e-05,
+      "loss": 0.6652,
+      "step": 2787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980457097932486e-05,
+      "loss": 0.6213,
+      "step": 2788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9804430716850585e-05,
+      "loss": 0.6663,
+      "step": 2789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9804290404557127e-05,
+      "loss": 0.5875,
+      "step": 2790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.98041500424452e-05,
+      "loss": 0.6062,
+      "step": 2791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980400963051551e-05,
+      "loss": 0.6381,
+      "step": 2792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803869168768777e-05,
+      "loss": 0.5826,
+      "step": 2793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803728657205713e-05,
+      "loss": 0.63,
+      "step": 2794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803588095827035e-05,
+      "loss": 0.6121,
+      "step": 2795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803447484633454e-05,
+      "loss": 0.6328,
+      "step": 2796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803306823625687e-05,
+      "loss": 0.5823,
+      "step": 2797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803166112804444e-05,
+      "loss": 0.6319,
+      "step": 2798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9803025352170443e-05,
+      "loss": 0.6262,
+      "step": 2799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.98028845417244e-05,
+      "loss": 0.6282,
+      "step": 2800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9802743681467032e-05,
+      "loss": 0.631,
+      "step": 2801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980260277139905e-05,
+      "loss": 0.6149,
+      "step": 2802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980246181152117e-05,
+      "loss": 0.6167,
+      "step": 2803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9802320801834117e-05,
+      "loss": 0.6481,
+      "step": 2804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9802179742338595e-05,
+      "loss": 0.6402,
+      "step": 2805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9802038633035333e-05,
+      "loss": 0.6171,
+      "step": 2806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801897473925037e-05,
+      "loss": 0.6369,
+      "step": 2807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801756265008434e-05,
+      "loss": 0.6309,
+      "step": 2808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801615006286236e-05,
+      "loss": 0.6312,
+      "step": 2809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801473697759158e-05,
+      "loss": 0.6536,
+      "step": 2810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801332339427922e-05,
+      "loss": 0.5889,
+      "step": 2811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980119093129325e-05,
+      "loss": 0.5906,
+      "step": 2812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9801049473355855e-05,
+      "loss": 0.6445,
+      "step": 2813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800907965616453e-05,
+      "loss": 0.563,
+      "step": 2814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800766408075773e-05,
+      "loss": 0.6597,
+      "step": 2815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800624800734523e-05,
+      "loss": 0.6653,
+      "step": 2816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.980048314359343e-05,
+      "loss": 0.6078,
+      "step": 2817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800341436653216e-05,
+      "loss": 0.6171,
+      "step": 2818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800199679914595e-05,
+      "loss": 0.6123,
+      "step": 2819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9800057873378287e-05,
+      "loss": 0.6862,
+      "step": 2820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9799916017045017e-05,
+      "loss": 0.6211,
+      "step": 2821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.97997741109155e-05,
+      "loss": 0.6311,
+      "step": 2822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979963215499047e-05,
+      "loss": 0.6508,
+      "step": 2823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979949014927063e-05,
+      "loss": 0.7075,
+      "step": 2824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9799348093756715e-05,
+      "loss": 0.6574,
+      "step": 2825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979920598844944e-05,
+      "loss": 0.536,
+      "step": 2826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979906383334953e-05,
+      "loss": 0.5642,
+      "step": 2827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798921628457707e-05,
+      "loss": 0.6477,
+      "step": 2828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979877937377469e-05,
+      "loss": 0.6163,
+      "step": 2829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798637069301212e-05,
+      "loss": 0.6683,
+      "step": 2830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798494715037988e-05,
+      "loss": 0.5934,
+      "step": 2831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798352310985743e-05,
+      "loss": 0.5955,
+      "step": 2832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798209857145197e-05,
+      "loss": 0.621,
+      "step": 2833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9798067353517077e-05,
+      "loss": 0.6129,
+      "step": 2834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797924800102112e-05,
+      "loss": 0.6306,
+      "step": 2835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797782196901018e-05,
+      "loss": 0.6117,
+      "step": 2836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797639543914524e-05,
+      "loss": 0.5934,
+      "step": 2837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797496841143355e-05,
+      "loss": 0.6094,
+      "step": 2838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979735408858823e-05,
+      "loss": 0.5828,
+      "step": 2839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797211286249886e-05,
+      "loss": 0.6197,
+      "step": 2840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9797068434129037e-05,
+      "loss": 0.6114,
+      "step": 2841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9796925532226422e-05,
+      "loss": 0.61,
+      "step": 2842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9796782580542752e-05,
+      "loss": 0.648,
+      "step": 2843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9796639579078763e-05,
+      "loss": 0.5949,
+      "step": 2844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979649652783518e-05,
+      "loss": 0.6219,
+      "step": 2845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979635342681273e-05,
+      "loss": 0.6794,
+      "step": 2846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9796210276012136e-05,
+      "loss": 0.6171,
+      "step": 2847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979606707543413e-05,
+      "loss": 0.6351,
+      "step": 2848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9795923825079436e-05,
+      "loss": 0.6534,
+      "step": 2849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9795780524948787e-05,
+      "loss": 0.6232,
+      "step": 2850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979563717504291e-05,
+      "loss": 0.6578,
+      "step": 2851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979549377536253e-05,
+      "loss": 0.593,
+      "step": 2852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9795350325908375e-05,
+      "loss": 0.546,
+      "step": 2853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9795206826681178e-05,
+      "loss": 0.6236,
+      "step": 2854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9795063277681668e-05,
+      "loss": 0.5989,
+      "step": 2855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9794919678910573e-05,
+      "loss": 0.6188,
+      "step": 2856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979477603036862e-05,
+      "loss": 0.5797,
+      "step": 2857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9794632332056542e-05,
+      "loss": 0.5923,
+      "step": 2858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979448858397507e-05,
+      "loss": 0.6153,
+      "step": 2859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9794344786124934e-05,
+      "loss": 0.6496,
+      "step": 2860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979420093850686e-05,
+      "loss": 0.6336,
+      "step": 2861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9794057041121583e-05,
+      "loss": 0.6709,
+      "step": 2862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9793913093969838e-05,
+      "loss": 0.6262,
+      "step": 2863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.979376909705235e-05,
+      "loss": 0.6629,
+      "step": 2864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.9793625050369852e-05,
+      "loss": 0.7121,
+      "step": 2865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.979348095392308e-05,
+      "loss": 0.5759,
+      "step": 2866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.979333680771276e-05,
+      "loss": 0.5917,
+      "step": 2867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9793192611739626e-05,
+      "loss": 0.5977,
+      "step": 2868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9793048366004413e-05,
+      "loss": 0.5741,
+      "step": 2869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792904070507855e-05,
+      "loss": 0.7083,
+      "step": 2870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792759725250686e-05,
+      "loss": 0.6584,
+      "step": 2871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792615330233635e-05,
+      "loss": 0.6565,
+      "step": 2872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792470885457437e-05,
+      "loss": 0.5439,
+      "step": 2873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792326390922823e-05,
+      "loss": 0.592,
+      "step": 2874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792181846630537e-05,
+      "loss": 0.5946,
+      "step": 2875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9792037252581306e-05,
+      "loss": 0.5297,
+      "step": 2876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.979189260877586e-05,
+      "loss": 0.6388,
+      "step": 2877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791747915214946e-05,
+      "loss": 0.6658,
+      "step": 2878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791603171899292e-05,
+      "loss": 0.6511,
+      "step": 2879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791458378829635e-05,
+      "loss": 0.6361,
+      "step": 2880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791313536006708e-05,
+      "loss": 0.6276,
+      "step": 2881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791168643431252e-05,
+      "loss": 0.6491,
+      "step": 2882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9791023701104e-05,
+      "loss": 0.6256,
+      "step": 2883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.979087870902569e-05,
+      "loss": 0.6332,
+      "step": 2884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9790733667197055e-05,
+      "loss": 0.6659,
+      "step": 2885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9790588575618838e-05,
+      "loss": 0.6473,
+      "step": 2886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9790443434291773e-05,
+      "loss": 0.6197,
+      "step": 2887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9790298243216594e-05,
+      "loss": 0.6611,
+      "step": 2888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9790153002394044e-05,
+      "loss": 0.6491,
+      "step": 2889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.979000771182486e-05,
+      "loss": 0.6186,
+      "step": 2890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9789862371509783e-05,
+      "loss": 0.5877,
+      "step": 2891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9789716981449546e-05,
+      "loss": 0.6243,
+      "step": 2892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9789571541644886e-05,
+      "loss": 0.5887,
+      "step": 2893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978942605209655e-05,
+      "loss": 0.6283,
+      "step": 2894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9789280512805272e-05,
+      "loss": 0.6535,
+      "step": 2895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9789134923771794e-05,
+      "loss": 0.5649,
+      "step": 2896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9788989284996852e-05,
+      "loss": 0.6549,
+      "step": 2897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9788843596481195e-05,
+      "loss": 0.6169,
+      "step": 2898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978869785822555e-05,
+      "loss": 0.67,
+      "step": 2899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9788552070230667e-05,
+      "loss": 0.6484,
+      "step": 2900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9788406232497284e-05,
+      "loss": 0.6649,
+      "step": 2901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978826034502614e-05,
+      "loss": 0.6438,
+      "step": 2902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978811440781798e-05,
+      "loss": 0.6184,
+      "step": 2903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9787968420873548e-05,
+      "loss": 0.5658,
+      "step": 2904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978782238419358e-05,
+      "loss": 0.6169,
+      "step": 2905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9787676297778815e-05,
+      "loss": 0.5833,
+      "step": 2906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9787530161630006e-05,
+      "loss": 0.5887,
+      "step": 2907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9787383975747887e-05,
+      "loss": 0.6518,
+      "step": 2908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9787237740133205e-05,
+      "loss": 0.5719,
+      "step": 2909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.97870914547867e-05,
+      "loss": 0.5675,
+      "step": 2910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9786945119709122e-05,
+      "loss": 0.6351,
+      "step": 2911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9786798734901204e-05,
+      "loss": 0.6713,
+      "step": 2912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9786652300363697e-05,
+      "loss": 0.6137,
+      "step": 2913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9786505816097344e-05,
+      "loss": 0.6031,
+      "step": 2914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978635928210289e-05,
+      "loss": 0.6572,
+      "step": 2915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9786212698381074e-05,
+      "loss": 0.6148,
+      "step": 2916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978606606493265e-05,
+      "loss": 0.6623,
+      "step": 2917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9785919381758358e-05,
+      "loss": 0.6454,
+      "step": 2918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978577264885894e-05,
+      "loss": 0.5697,
+      "step": 2919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9785625866235148e-05,
+      "loss": 0.5786,
+      "step": 2920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9785479033887723e-05,
+      "loss": 0.571,
+      "step": 2921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9785332151817414e-05,
+      "loss": 0.6006,
+      "step": 2922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978518522002497e-05,
+      "loss": 0.6255,
+      "step": 2923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978503823851113e-05,
+      "loss": 0.601,
+      "step": 2924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978489120727665e-05,
+      "loss": 0.5966,
+      "step": 2925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9784744126322266e-05,
+      "loss": 0.6771,
+      "step": 2926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9784596995648736e-05,
+      "loss": 0.6221,
+      "step": 2927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.97844498152568e-05,
+      "loss": 0.6733,
+      "step": 2928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978430258514721e-05,
+      "loss": 0.6845,
+      "step": 2929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9784155305320713e-05,
+      "loss": 0.6028,
+      "step": 2930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978400797577806e-05,
+      "loss": 0.6234,
+      "step": 2931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9783860596519995e-05,
+      "loss": 0.5887,
+      "step": 2932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9783713167547268e-05,
+      "loss": 0.6369,
+      "step": 2933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978356568886063e-05,
+      "loss": 0.6599,
+      "step": 2934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978341816046083e-05,
+      "loss": 0.6059,
+      "step": 2935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9783270582348618e-05,
+      "loss": 0.6404,
+      "step": 2936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978312295452474e-05,
+      "loss": 0.6012,
+      "step": 2937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782975276989953e-05,
+      "loss": 0.5847,
+      "step": 2938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782827549745e-05,
+      "loss": 0.6107,
+      "step": 2939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782679772790637e-05,
+      "loss": 0.5758,
+      "step": 2940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782531946127614e-05,
+      "loss": 0.6387,
+      "step": 2941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978238406975668e-05,
+      "loss": 0.6212,
+      "step": 2942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782236143678587e-05,
+      "loss": 0.6408,
+      "step": 2943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9782088167894085e-05,
+      "loss": 0.6151,
+      "step": 2944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9781940142403932e-05,
+      "loss": 0.6455,
+      "step": 2945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9781792067208875e-05,
+      "loss": 0.7037,
+      "step": 2946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978164394230967e-05,
+      "loss": 0.6438,
+      "step": 2947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9781495767707065e-05,
+      "loss": 0.5309,
+      "step": 2948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9781347543401816e-05,
+      "loss": 0.6219,
+      "step": 2949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9781199269394675e-05,
+      "loss": 0.6027,
+      "step": 2950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.97810509456864e-05,
+      "loss": 0.5827,
+      "step": 2951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9780902572277734e-05,
+      "loss": 0.6499,
+      "step": 2952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978075414916944e-05,
+      "loss": 0.6004,
+      "step": 2953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978060567636227e-05,
+      "loss": 0.5832,
+      "step": 2954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9780457153856983e-05,
+      "loss": 0.5674,
+      "step": 2955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9780308581654323e-05,
+      "loss": 0.6359,
+      "step": 2956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.978015995975505e-05,
+      "loss": 0.6759,
+      "step": 2957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9780011288159925e-05,
+      "loss": 0.5997,
+      "step": 2958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9779862566869698e-05,
+      "loss": 0.6436,
+      "step": 2959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977971379588512e-05,
+      "loss": 0.5765,
+      "step": 2960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9779564975206956e-05,
+      "loss": 0.5581,
+      "step": 2961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9779416104835956e-05,
+      "loss": 0.6063,
+      "step": 2962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977926718477288e-05,
+      "loss": 0.6483,
+      "step": 2963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977911821501848e-05,
+      "loss": 0.5858,
+      "step": 2964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778969195573524e-05,
+      "loss": 0.6337,
+      "step": 2965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778820126438755e-05,
+      "loss": 0.5605,
+      "step": 2966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778671007614943e-05,
+      "loss": 0.6515,
+      "step": 2967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778521839102836e-05,
+      "loss": 0.5733,
+      "step": 2968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778372620903194e-05,
+      "loss": 0.6428,
+      "step": 2969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977822335301678e-05,
+      "loss": 0.5931,
+      "step": 2970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9778074035444346e-05,
+      "loss": 0.6211,
+      "step": 2971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9777924668186658e-05,
+      "loss": 0.584,
+      "step": 2972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977777525124447e-05,
+      "loss": 0.5973,
+      "step": 2973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9777625784618542e-05,
+      "loss": 0.6078,
+      "step": 2974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9777476268309635e-05,
+      "loss": 0.6183,
+      "step": 2975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9777326702318506e-05,
+      "loss": 0.6537,
+      "step": 2976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977717708664592e-05,
+      "loss": 0.6241,
+      "step": 2977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977702742129263e-05,
+      "loss": 0.6609,
+      "step": 2978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.97768777062594e-05,
+      "loss": 0.6317,
+      "step": 2979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9776727941546995e-05,
+      "loss": 0.5982,
+      "step": 2980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977657812715617e-05,
+      "loss": 0.5849,
+      "step": 2981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9776428263087687e-05,
+      "loss": 0.5964,
+      "step": 2982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977627834934231e-05,
+      "loss": 0.6106,
+      "step": 2983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9776128385920796e-05,
+      "loss": 0.5945,
+      "step": 2984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9775978372823915e-05,
+      "loss": 0.5545,
+      "step": 2985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9775828310052424e-05,
+      "loss": 0.6,
+      "step": 2986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977567819760709e-05,
+      "loss": 0.6105,
+      "step": 2987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9775528035488665e-05,
+      "loss": 0.6503,
+      "step": 2988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9775377823697922e-05,
+      "loss": 0.6207,
+      "step": 2989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9775227562235622e-05,
+      "loss": 0.6285,
+      "step": 2990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977507725110253e-05,
+      "loss": 0.635,
+      "step": 2991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9774926890299408e-05,
+      "loss": 0.6494,
+      "step": 2992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9774776479827014e-05,
+      "loss": 0.6002,
+      "step": 2993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977462601968612e-05,
+      "loss": 0.6185,
+      "step": 2994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977447550987749e-05,
+      "loss": 0.6267,
+      "step": 2995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9774324950401887e-05,
+      "loss": 0.6409,
+      "step": 2996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9774174341260078e-05,
+      "loss": 0.6097,
+      "step": 2997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9774023682452826e-05,
+      "loss": 0.6343,
+      "step": 2998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9773872973980896e-05,
+      "loss": 0.6585,
+      "step": 2999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9773722215845056e-05,
+      "loss": 0.6855,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977357140804607e-05,
+      "loss": 0.6323,
+      "step": 3001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9773420550584705e-05,
+      "loss": 0.5432,
+      "step": 3002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977326964346173e-05,
+      "loss": 0.6018,
+      "step": 3003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977311868667791e-05,
+      "loss": 0.6233,
+      "step": 3004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.977296768023401e-05,
+      "loss": 0.6572,
+      "step": 3005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772816624130798e-05,
+      "loss": 0.5956,
+      "step": 3006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772665518369045e-05,
+      "loss": 0.5996,
+      "step": 3007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772514362949517e-05,
+      "loss": 0.602,
+      "step": 3008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772363157872978e-05,
+      "loss": 0.639,
+      "step": 3009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772211903140203e-05,
+      "loss": 0.6156,
+      "step": 3010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9772060598751958e-05,
+      "loss": 0.665,
+      "step": 3011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.9771909244709005e-05,
+      "loss": 0.62,
+      "step": 3012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9771757841012127e-05,
+      "loss": 0.632,
+      "step": 3013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9771606387662083e-05,
+      "loss": 0.6144,
+      "step": 3014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9771454884659644e-05,
+      "loss": 0.5822,
+      "step": 3015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.977130333200558e-05,
+      "loss": 0.6261,
+      "step": 3016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9771151729700663e-05,
+      "loss": 0.5847,
+      "step": 3017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9771000077745665e-05,
+      "loss": 0.6339,
+      "step": 3018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.977084837614135e-05,
+      "loss": 0.6477,
+      "step": 3019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.97706966248885e-05,
+      "loss": 0.6244,
+      "step": 3020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.977054482398787e-05,
+      "loss": 0.5793,
+      "step": 3021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9770392973440246e-05,
+      "loss": 0.7121,
+      "step": 3022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9770241073246393e-05,
+      "loss": 0.6308,
+      "step": 3023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9770089123407082e-05,
+      "loss": 0.5913,
+      "step": 3024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9769937123923085e-05,
+      "loss": 0.5786,
+      "step": 3025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976978507479518e-05,
+      "loss": 0.6226,
+      "step": 3026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9769632976024134e-05,
+      "loss": 0.6378,
+      "step": 3027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9769480827610725e-05,
+      "loss": 0.6006,
+      "step": 3028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976932862955572e-05,
+      "loss": 0.6903,
+      "step": 3029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.97691763818599e-05,
+      "loss": 0.6575,
+      "step": 3030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9769024084524028e-05,
+      "loss": 0.5837,
+      "step": 3031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9768871737548884e-05,
+      "loss": 0.5606,
+      "step": 3032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976871934093524e-05,
+      "loss": 0.6456,
+      "step": 3033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9768566894683876e-05,
+      "loss": 0.6511,
+      "step": 3034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9768414398795563e-05,
+      "loss": 0.5886,
+      "step": 3035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9768261853271072e-05,
+      "loss": 0.6235,
+      "step": 3036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9768109258111186e-05,
+      "loss": 0.6319,
+      "step": 3037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767956613316673e-05,
+      "loss": 0.6148,
+      "step": 3038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767803918888313e-05,
+      "loss": 0.5665,
+      "step": 3039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767651174826882e-05,
+      "loss": 0.6442,
+      "step": 3040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767498381133152e-05,
+      "loss": 0.5918,
+      "step": 3041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767345537807903e-05,
+      "loss": 0.5981,
+      "step": 3042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767192644851913e-05,
+      "loss": 0.6491,
+      "step": 3043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9767039702265952e-05,
+      "loss": 0.6604,
+      "step": 3044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9766886710050807e-05,
+      "loss": 0.6487,
+      "step": 3045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976673366820725e-05,
+      "loss": 0.6227,
+      "step": 3046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9766580576736057e-05,
+      "loss": 0.6252,
+      "step": 3047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9766427435638005e-05,
+      "loss": 0.6343,
+      "step": 3048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9766274244913878e-05,
+      "loss": 0.639,
+      "step": 3049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9766121004564455e-05,
+      "loss": 0.5894,
+      "step": 3050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765967714590503e-05,
+      "loss": 0.6384,
+      "step": 3051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765814374992814e-05,
+      "loss": 0.6067,
+      "step": 3052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976566098577216e-05,
+      "loss": 0.6364,
+      "step": 3053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765507546929325e-05,
+      "loss": 0.5842,
+      "step": 3054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765354058465087e-05,
+      "loss": 0.6663,
+      "step": 3055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765200520380222e-05,
+      "loss": 0.6191,
+      "step": 3056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9765046932675514e-05,
+      "loss": 0.6024,
+      "step": 3057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9764893295351744e-05,
+      "loss": 0.6091,
+      "step": 3058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9764739608409692e-05,
+      "loss": 0.614,
+      "step": 3059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9764585871850135e-05,
+      "loss": 0.6509,
+      "step": 3060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976443208567386e-05,
+      "loss": 0.617,
+      "step": 3061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9764278249881644e-05,
+      "loss": 0.6345,
+      "step": 3062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976412436447427e-05,
+      "loss": 0.5838,
+      "step": 3063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9763970429452525e-05,
+      "loss": 0.6049,
+      "step": 3064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9763816444817183e-05,
+      "loss": 0.6154,
+      "step": 3065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976366241056903e-05,
+      "loss": 0.6649,
+      "step": 3066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9763508326708854e-05,
+      "loss": 0.613,
+      "step": 3067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9763354193237426e-05,
+      "loss": 0.6114,
+      "step": 3068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976320001015554e-05,
+      "loss": 0.6431,
+      "step": 3069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9763045777463973e-05,
+      "loss": 0.6042,
+      "step": 3070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9762891495163513e-05,
+      "loss": 0.6244,
+      "step": 3071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9762737163254942e-05,
+      "loss": 0.6261,
+      "step": 3072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9762582781739047e-05,
+      "loss": 0.6391,
+      "step": 3073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9762428350616606e-05,
+      "loss": 0.6609,
+      "step": 3074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976227386988841e-05,
+      "loss": 0.6199,
+      "step": 3075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9762119339555238e-05,
+      "loss": 0.643,
+      "step": 3076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976196475961788e-05,
+      "loss": 0.603,
+      "step": 3077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976181013007712e-05,
+      "loss": 0.5832,
+      "step": 3078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9761655450933745e-05,
+      "loss": 0.6325,
+      "step": 3079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976150072218854e-05,
+      "loss": 0.6414,
+      "step": 3080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976134594384229e-05,
+      "loss": 0.6504,
+      "step": 3081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9761191115895783e-05,
+      "loss": 0.6628,
+      "step": 3082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9761036238349804e-05,
+      "loss": 0.5317,
+      "step": 3083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9760881311205143e-05,
+      "loss": 0.6688,
+      "step": 3084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9760726334462587e-05,
+      "loss": 0.6334,
+      "step": 3085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9760571308122918e-05,
+      "loss": 0.5592,
+      "step": 3086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976041623218693e-05,
+      "loss": 0.5919,
+      "step": 3087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976026110665541e-05,
+      "loss": 0.5806,
+      "step": 3088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.976010593152914e-05,
+      "loss": 0.6736,
+      "step": 3089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759950706808917e-05,
+      "loss": 0.6191,
+      "step": 3090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759795432495524e-05,
+      "loss": 0.6386,
+      "step": 3091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759640108589755e-05,
+      "loss": 0.524,
+      "step": 3092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759484735092392e-05,
+      "loss": 0.6261,
+      "step": 3093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759329312004235e-05,
+      "loss": 0.566,
+      "step": 3094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759173839326064e-05,
+      "loss": 0.6437,
+      "step": 3095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9759018317058673e-05,
+      "loss": 0.6108,
+      "step": 3096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975886274520285e-05,
+      "loss": 0.6212,
+      "step": 3097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975870712375939e-05,
+      "loss": 0.6841,
+      "step": 3098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975855145272908e-05,
+      "loss": 0.5863,
+      "step": 3099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9758395732112714e-05,
+      "loss": 0.5951,
+      "step": 3100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9758239961911083e-05,
+      "loss": 0.6095,
+      "step": 3101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9758084142124972e-05,
+      "loss": 0.6159,
+      "step": 3102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975792827275518e-05,
+      "loss": 0.623,
+      "step": 3103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9757772353802497e-05,
+      "loss": 0.5692,
+      "step": 3104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975761638526771e-05,
+      "loss": 0.6402,
+      "step": 3105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975746036715162e-05,
+      "loss": 0.6028,
+      "step": 3106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975730429945502e-05,
+      "loss": 0.6127,
+      "step": 3107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9757148182178697e-05,
+      "loss": 0.6472,
+      "step": 3108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9756992015323445e-05,
+      "loss": 0.6363,
+      "step": 3109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975683579889006e-05,
+      "loss": 0.6532,
+      "step": 3110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9756679532879333e-05,
+      "loss": 0.6318,
+      "step": 3111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975652321729206e-05,
+      "loss": 0.6652,
+      "step": 3112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975636685212904e-05,
+      "loss": 0.5863,
+      "step": 3113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975621043739106e-05,
+      "loss": 0.6692,
+      "step": 3114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9756053973078917e-05,
+      "loss": 0.5697,
+      "step": 3115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9755897459193407e-05,
+      "loss": 0.6016,
+      "step": 3116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9755740895735323e-05,
+      "loss": 0.6066,
+      "step": 3117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9755584282705466e-05,
+      "loss": 0.6354,
+      "step": 3118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9755427620104626e-05,
+      "loss": 0.602,
+      "step": 3119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.97552709079336e-05,
+      "loss": 0.6266,
+      "step": 3120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975511414619319e-05,
+      "loss": 0.6748,
+      "step": 3121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975495733488418e-05,
+      "loss": 0.6222,
+      "step": 3122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9754800474007382e-05,
+      "loss": 0.596,
+      "step": 3123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9754643563563586e-05,
+      "loss": 0.6276,
+      "step": 3124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9754486603553588e-05,
+      "loss": 0.6276,
+      "step": 3125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9754329593978182e-05,
+      "loss": 0.6224,
+      "step": 3126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975417253483818e-05,
+      "loss": 0.6224,
+      "step": 3127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9754015426134363e-05,
+      "loss": 0.64,
+      "step": 3128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975385826786754e-05,
+      "loss": 0.5937,
+      "step": 3129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9753701060038505e-05,
+      "loss": 0.6346,
+      "step": 3130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9753543802648062e-05,
+      "loss": 0.6373,
+      "step": 3131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9753386495697e-05,
+      "loss": 0.6682,
+      "step": 3132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9753229139186133e-05,
+      "loss": 0.6353,
+      "step": 3133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9753071733116248e-05,
+      "loss": 0.5219,
+      "step": 3134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9752914277488146e-05,
+      "loss": 0.6082,
+      "step": 3135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9752756772302633e-05,
+      "loss": 0.6543,
+      "step": 3136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975259921756051e-05,
+      "loss": 0.5442,
+      "step": 3137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9752441613262573e-05,
+      "loss": 0.6127,
+      "step": 3138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975228395940962e-05,
+      "loss": 0.5633,
+      "step": 3139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975212625600246e-05,
+      "loss": 0.5823,
+      "step": 3140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975196850304189e-05,
+      "loss": 0.5856,
+      "step": 3141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975181070052871e-05,
+      "loss": 0.6148,
+      "step": 3142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9751652848463726e-05,
+      "loss": 0.6046,
+      "step": 3143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9751494946847736e-05,
+      "loss": 0.6279,
+      "step": 3144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9751336995681544e-05,
+      "loss": 0.5852,
+      "step": 3145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9751178994965955e-05,
+      "loss": 0.6401,
+      "step": 3146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9751020944701768e-05,
+      "loss": 0.5952,
+      "step": 3147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9750862844889788e-05,
+      "loss": 0.6955,
+      "step": 3148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.975070469553082e-05,
+      "loss": 0.635,
+      "step": 3149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9750546496625665e-05,
+      "loss": 0.5767,
+      "step": 3150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9750388248175128e-05,
+      "loss": 0.6295,
+      "step": 3151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9750229950180014e-05,
+      "loss": 0.6899,
+      "step": 3152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9750071602641123e-05,
+      "loss": 0.6435,
+      "step": 3153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9749913205559264e-05,
+      "loss": 0.6013,
+      "step": 3154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.974975475893524e-05,
+      "loss": 0.6644,
+      "step": 3155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9749596262769858e-05,
+      "loss": 0.6044,
+      "step": 3156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9749437717063923e-05,
+      "loss": 0.5925,
+      "step": 3157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9749279121818235e-05,
+      "loss": 0.5764,
+      "step": 3158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.9749120477033613e-05,
+      "loss": 0.6699,
+      "step": 3159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9748961782710848e-05,
+      "loss": 0.5866,
+      "step": 3160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9748803038850753e-05,
+      "loss": 0.574,
+      "step": 3161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974864424545414e-05,
+      "loss": 0.5931,
+      "step": 3162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9748485402521805e-05,
+      "loss": 0.6161,
+      "step": 3163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974832651005456e-05,
+      "loss": 0.6352,
+      "step": 3164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974816756805322e-05,
+      "loss": 0.6826,
+      "step": 3165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974800857651858e-05,
+      "loss": 0.6123,
+      "step": 3166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9747849535451452e-05,
+      "loss": 0.6631,
+      "step": 3167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974769044485265e-05,
+      "loss": 0.6466,
+      "step": 3168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9747531304722978e-05,
+      "loss": 0.6537,
+      "step": 3169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9747372115063243e-05,
+      "loss": 0.6231,
+      "step": 3170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9747212875874256e-05,
+      "loss": 0.5777,
+      "step": 3171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9747053587156826e-05,
+      "loss": 0.6494,
+      "step": 3172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9746894248911764e-05,
+      "loss": 0.6346,
+      "step": 3173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9746734861139873e-05,
+      "loss": 0.6235,
+      "step": 3174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974657542384197e-05,
+      "loss": 0.6453,
+      "step": 3175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9746415937018867e-05,
+      "loss": 0.5782,
+      "step": 3176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9746256400671365e-05,
+      "loss": 0.5825,
+      "step": 3177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974609681480028e-05,
+      "loss": 0.5801,
+      "step": 3178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9745937179406428e-05,
+      "loss": 0.6003,
+      "step": 3179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9745777494490608e-05,
+      "loss": 0.6187,
+      "step": 3180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974561776005364e-05,
+      "loss": 0.6189,
+      "step": 3181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9745457976096338e-05,
+      "loss": 0.6083,
+      "step": 3182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9745298142619508e-05,
+      "loss": 0.5593,
+      "step": 3183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9745138259623963e-05,
+      "loss": 0.6051,
+      "step": 3184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974497832711052e-05,
+      "loss": 0.6681,
+      "step": 3185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9744818345079983e-05,
+      "loss": 0.643,
+      "step": 3186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9744658313533175e-05,
+      "loss": 0.5691,
+      "step": 3187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.97444982324709e-05,
+      "loss": 0.5848,
+      "step": 3188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9744338101893978e-05,
+      "loss": 0.617,
+      "step": 3189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974417792180322e-05,
+      "loss": 0.6443,
+      "step": 3190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974401769219944e-05,
+      "loss": 0.6806,
+      "step": 3191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9743857413083455e-05,
+      "loss": 0.6217,
+      "step": 3192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974369708445607e-05,
+      "loss": 0.5839,
+      "step": 3193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9743536706318115e-05,
+      "loss": 0.5989,
+      "step": 3194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9743376278670393e-05,
+      "loss": 0.6449,
+      "step": 3195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974321580151372e-05,
+      "loss": 0.7199,
+      "step": 3196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9743055274848918e-05,
+      "loss": 0.6432,
+      "step": 3197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9742894698676794e-05,
+      "loss": 0.6047,
+      "step": 3198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9742734072998173e-05,
+      "loss": 0.5987,
+      "step": 3199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9742573397813864e-05,
+      "loss": 0.5808,
+      "step": 3200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9742412673124688e-05,
+      "loss": 0.6304,
+      "step": 3201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974225189893146e-05,
+      "loss": 0.6339,
+      "step": 3202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9742091075235e-05,
+      "loss": 0.6602,
+      "step": 3203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9741930202036115e-05,
+      "loss": 0.6033,
+      "step": 3204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9741769279335633e-05,
+      "loss": 0.6324,
+      "step": 3205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974160830713437e-05,
+      "loss": 0.6213,
+      "step": 3206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974144728543314e-05,
+      "loss": 0.6484,
+      "step": 3207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9741286214232764e-05,
+      "loss": 0.6281,
+      "step": 3208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9741125093534062e-05,
+      "loss": 0.63,
+      "step": 3209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974096392333785e-05,
+      "loss": 0.6537,
+      "step": 3210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9740802703644945e-05,
+      "loss": 0.6152,
+      "step": 3211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9740641434456173e-05,
+      "loss": 0.5929,
+      "step": 3212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.974048011577234e-05,
+      "loss": 0.5895,
+      "step": 3213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9740318747594285e-05,
+      "loss": 0.5708,
+      "step": 3214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9740157329922813e-05,
+      "loss": 0.5956,
+      "step": 3215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973999586275875e-05,
+      "loss": 0.6772,
+      "step": 3216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9739834346102917e-05,
+      "loss": 0.6394,
+      "step": 3217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973967277995613e-05,
+      "loss": 0.6471,
+      "step": 3218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9739511164319214e-05,
+      "loss": 0.6184,
+      "step": 3219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9739349499192992e-05,
+      "loss": 0.597,
+      "step": 3220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9739187784578284e-05,
+      "loss": 0.6167,
+      "step": 3221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973902602047591e-05,
+      "loss": 0.6434,
+      "step": 3222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738864206886686e-05,
+      "loss": 0.6147,
+      "step": 3223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738702343811446e-05,
+      "loss": 0.6609,
+      "step": 3224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738540431251006e-05,
+      "loss": 0.5926,
+      "step": 3225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738378469206194e-05,
+      "loss": 0.6183,
+      "step": 3226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738216457677825e-05,
+      "loss": 0.5701,
+      "step": 3227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9738054396666726e-05,
+      "loss": 0.6197,
+      "step": 3228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9737892286173724e-05,
+      "loss": 0.6284,
+      "step": 3229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9737730126199633e-05,
+      "loss": 0.6501,
+      "step": 3230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973756791674529e-05,
+      "loss": 0.6034,
+      "step": 3231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973740565781151e-05,
+      "loss": 0.6443,
+      "step": 3232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973724334939912e-05,
+      "loss": 0.5818,
+      "step": 3233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9737080991508944e-05,
+      "loss": 0.6105,
+      "step": 3234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973691858414181e-05,
+      "loss": 0.5832,
+      "step": 3235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9736756127298542e-05,
+      "loss": 0.615,
+      "step": 3236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9736593620979963e-05,
+      "loss": 0.577,
+      "step": 3237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9736431065186898e-05,
+      "loss": 0.6286,
+      "step": 3238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973626845992018e-05,
+      "loss": 0.6398,
+      "step": 3239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9736105805180625e-05,
+      "loss": 0.5731,
+      "step": 3240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735943100969068e-05,
+      "loss": 0.5896,
+      "step": 3241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735780347286334e-05,
+      "loss": 0.5899,
+      "step": 3242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735617544133245e-05,
+      "loss": 0.6215,
+      "step": 3243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735454691510636e-05,
+      "loss": 0.5923,
+      "step": 3244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735291789419327e-05,
+      "loss": 0.6408,
+      "step": 3245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9735128837860153e-05,
+      "loss": 0.6392,
+      "step": 3246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9734965836833936e-05,
+      "loss": 0.6476,
+      "step": 3247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973480278634151e-05,
+      "loss": 0.5847,
+      "step": 3248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9734639686383695e-05,
+      "loss": 0.6675,
+      "step": 3249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9734476536961327e-05,
+      "loss": 0.6985,
+      "step": 3250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9734313338075234e-05,
+      "loss": 0.6187,
+      "step": 3251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9734150089726242e-05,
+      "loss": 0.6512,
+      "step": 3252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973398679191518e-05,
+      "loss": 0.6291,
+      "step": 3253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973382344464289e-05,
+      "loss": 0.6308,
+      "step": 3254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9733660047910184e-05,
+      "loss": 0.619,
+      "step": 3255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9733496601717907e-05,
+      "loss": 0.6615,
+      "step": 3256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9733333106066875e-05,
+      "loss": 0.5996,
+      "step": 3257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973316956095793e-05,
+      "loss": 0.6341,
+      "step": 3258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9733005966391903e-05,
+      "loss": 0.5908,
+      "step": 3259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9732842322369623e-05,
+      "loss": 0.5956,
+      "step": 3260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9732678628891915e-05,
+      "loss": 0.5954,
+      "step": 3261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9732514885959623e-05,
+      "loss": 0.5969,
+      "step": 3262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9732351093573566e-05,
+      "loss": 0.5787,
+      "step": 3263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973218725173459e-05,
+      "loss": 0.6048,
+      "step": 3264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9732023360443513e-05,
+      "loss": 0.5748,
+      "step": 3265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973185941970118e-05,
+      "loss": 0.6114,
+      "step": 3266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9731695429508418e-05,
+      "loss": 0.6128,
+      "step": 3267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.973153138986606e-05,
+      "loss": 0.5811,
+      "step": 3268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9731367300774944e-05,
+      "loss": 0.6365,
+      "step": 3269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.97312031622359e-05,
+      "loss": 0.5967,
+      "step": 3270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9731038974249763e-05,
+      "loss": 0.6382,
+      "step": 3271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730874736817362e-05,
+      "loss": 0.603,
+      "step": 3272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730710449939542e-05,
+      "loss": 0.6777,
+      "step": 3273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730546113617133e-05,
+      "loss": 0.6414,
+      "step": 3274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730381727850968e-05,
+      "loss": 0.6142,
+      "step": 3275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730217292641883e-05,
+      "loss": 0.5976,
+      "step": 3276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9730052807990713e-05,
+      "loss": 0.6013,
+      "step": 3277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729888273898298e-05,
+      "loss": 0.6383,
+      "step": 3278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729723690365467e-05,
+      "loss": 0.541,
+      "step": 3279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729559057393065e-05,
+      "loss": 0.5835,
+      "step": 3280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729394374981924e-05,
+      "loss": 0.6113,
+      "step": 3281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729229643132876e-05,
+      "loss": 0.6342,
+      "step": 3282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9729064861846766e-05,
+      "loss": 0.6338,
+      "step": 3283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972890003112443e-05,
+      "loss": 0.6495,
+      "step": 3284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.97287351509667e-05,
+      "loss": 0.636,
+      "step": 3285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972857022137442e-05,
+      "loss": 0.5771,
+      "step": 3286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9728405242348424e-05,
+      "loss": 0.5861,
+      "step": 3287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9728240213889556e-05,
+      "loss": 0.6845,
+      "step": 3288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9728075135998644e-05,
+      "loss": 0.5644,
+      "step": 3289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972791000867654e-05,
+      "loss": 0.5893,
+      "step": 3290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9727744831924073e-05,
+      "loss": 0.642,
+      "step": 3291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9727579605742083e-05,
+      "loss": 0.5614,
+      "step": 3292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9727414330131415e-05,
+      "loss": 0.5973,
+      "step": 3293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9727249005092908e-05,
+      "loss": 0.631,
+      "step": 3294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9727083630627396e-05,
+      "loss": 0.5283,
+      "step": 3295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726918206735726e-05,
+      "loss": 0.6439,
+      "step": 3296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726752733418735e-05,
+      "loss": 0.585,
+      "step": 3297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726587210677263e-05,
+      "loss": 0.5909,
+      "step": 3298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726421638512154e-05,
+      "loss": 0.6349,
+      "step": 3299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726256016924248e-05,
+      "loss": 0.5555,
+      "step": 3300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9726090345914388e-05,
+      "loss": 0.6409,
+      "step": 3301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9725924625483415e-05,
+      "loss": 0.6098,
+      "step": 3302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972575885563217e-05,
+      "loss": 0.6074,
+      "step": 3303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.9725593036361495e-05,
+      "loss": 0.7213,
+      "step": 3304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972542716767223e-05,
+      "loss": 0.6606,
+      "step": 3305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.972526124956523e-05,
+      "loss": 0.6073,
+      "step": 3306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9725095282041322e-05,
+      "loss": 0.6605,
+      "step": 3307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972492926510136e-05,
+      "loss": 0.5959,
+      "step": 3308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9724763198746184e-05,
+      "loss": 0.5837,
+      "step": 3309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9724597082976638e-05,
+      "loss": 0.6599,
+      "step": 3310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9724430917793564e-05,
+      "loss": 0.6451,
+      "step": 3311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9724264703197813e-05,
+      "loss": 0.6564,
+      "step": 3312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9724098439190223e-05,
+      "loss": 0.6246,
+      "step": 3313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972393212577164e-05,
+      "loss": 0.6763,
+      "step": 3314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972376576294291e-05,
+      "loss": 0.5994,
+      "step": 3315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972359935070488e-05,
+      "loss": 0.5939,
+      "step": 3316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9723432889058395e-05,
+      "loss": 0.6294,
+      "step": 3317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9723266378004297e-05,
+      "loss": 0.6209,
+      "step": 3318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9723099817543436e-05,
+      "loss": 0.5929,
+      "step": 3319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9722933207676657e-05,
+      "loss": 0.6042,
+      "step": 3320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9722766548404808e-05,
+      "loss": 0.601,
+      "step": 3321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9722599839728732e-05,
+      "loss": 0.6095,
+      "step": 3322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972243308164928e-05,
+      "loss": 0.5681,
+      "step": 3323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9722266274167302e-05,
+      "loss": 0.6043,
+      "step": 3324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972209941728364e-05,
+      "loss": 0.5927,
+      "step": 3325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721932510999144e-05,
+      "loss": 0.6195,
+      "step": 3326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721765555314657e-05,
+      "loss": 0.6343,
+      "step": 3327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721598550231036e-05,
+      "loss": 0.618,
+      "step": 3328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721431495749125e-05,
+      "loss": 0.6128,
+      "step": 3329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721264391869773e-05,
+      "loss": 0.5905,
+      "step": 3330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9721097238593828e-05,
+      "loss": 0.5807,
+      "step": 3331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9720930035922144e-05,
+      "loss": 0.5964,
+      "step": 3332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9720762783855567e-05,
+      "loss": 0.6968,
+      "step": 3333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9720595482394945e-05,
+      "loss": 0.6108,
+      "step": 3334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9720428131541135e-05,
+      "loss": 0.6214,
+      "step": 3335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.972026073129498e-05,
+      "loss": 0.589,
+      "step": 3336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9720093281657332e-05,
+      "loss": 0.6452,
+      "step": 3337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719925782629047e-05,
+      "loss": 0.5868,
+      "step": 3338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719758234210972e-05,
+      "loss": 0.6279,
+      "step": 3339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719590636403957e-05,
+      "loss": 0.5558,
+      "step": 3340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719422989208854e-05,
+      "loss": 0.6596,
+      "step": 3341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719255292626522e-05,
+      "loss": 0.6061,
+      "step": 3342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9719087546657802e-05,
+      "loss": 0.5915,
+      "step": 3343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9718919751303554e-05,
+      "loss": 0.5698,
+      "step": 3344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9718751906564626e-05,
+      "loss": 0.6416,
+      "step": 3345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9718584012441878e-05,
+      "loss": 0.608,
+      "step": 3346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9718416068936154e-05,
+      "loss": 0.6444,
+      "step": 3347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9718248076048314e-05,
+      "loss": 0.5681,
+      "step": 3348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971808003377921e-05,
+      "loss": 0.658,
+      "step": 3349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9717911942129695e-05,
+      "loss": 0.5665,
+      "step": 3350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971774380110062e-05,
+      "loss": 0.6267,
+      "step": 3351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9717575610692846e-05,
+      "loss": 0.6548,
+      "step": 3352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9717407370907225e-05,
+      "loss": 0.649,
+      "step": 3353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971723908174461e-05,
+      "loss": 0.571,
+      "step": 3354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9717070743205857e-05,
+      "loss": 0.5878,
+      "step": 3355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9716902355291822e-05,
+      "loss": 0.5818,
+      "step": 3356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9716733918003362e-05,
+      "loss": 0.616,
+      "step": 3357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9716565431341328e-05,
+      "loss": 0.7247,
+      "step": 3358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9716396895306584e-05,
+      "loss": 0.642,
+      "step": 3359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971622830989998e-05,
+      "loss": 0.619,
+      "step": 3360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971605967512237e-05,
+      "loss": 0.6144,
+      "step": 3361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971589099097462e-05,
+      "loss": 0.6275,
+      "step": 3362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971572225745758e-05,
+      "loss": 0.6156,
+      "step": 3363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9715553474572112e-05,
+      "loss": 0.6125,
+      "step": 3364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971538464231907e-05,
+      "loss": 0.6259,
+      "step": 3365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9715215760699316e-05,
+      "loss": 0.5887,
+      "step": 3366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97150468297137e-05,
+      "loss": 0.6534,
+      "step": 3367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971487784936309e-05,
+      "loss": 0.5946,
+      "step": 3368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9714708819648343e-05,
+      "loss": 0.6058,
+      "step": 3369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971453974057031e-05,
+      "loss": 0.592,
+      "step": 3370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9714370612129857e-05,
+      "loss": 0.5684,
+      "step": 3371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9714201434327847e-05,
+      "loss": 0.6279,
+      "step": 3372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971403220716513e-05,
+      "loss": 0.597,
+      "step": 3373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971386293064257e-05,
+      "loss": 0.5951,
+      "step": 3374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9713693604761028e-05,
+      "loss": 0.6234,
+      "step": 3375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9713524229521364e-05,
+      "loss": 0.6272,
+      "step": 3376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971335480492444e-05,
+      "loss": 0.6338,
+      "step": 3377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9713185330971113e-05,
+      "loss": 0.6294,
+      "step": 3378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971301580766225e-05,
+      "loss": 0.5981,
+      "step": 3379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9712846234998705e-05,
+      "loss": 0.5863,
+      "step": 3380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9712676612981348e-05,
+      "loss": 0.5822,
+      "step": 3381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9712506941611035e-05,
+      "loss": 0.5796,
+      "step": 3382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9712337220888627e-05,
+      "loss": 0.6381,
+      "step": 3383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9712167450814994e-05,
+      "loss": 0.6444,
+      "step": 3384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9711997631390992e-05,
+      "loss": 0.6344,
+      "step": 3385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9711827762617485e-05,
+      "loss": 0.5802,
+      "step": 3386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9711657844495337e-05,
+      "loss": 0.6181,
+      "step": 3387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9711487877025412e-05,
+      "loss": 0.5654,
+      "step": 3388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971131786020857e-05,
+      "loss": 0.6555,
+      "step": 3389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9711147794045683e-05,
+      "loss": 0.6427,
+      "step": 3390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9710977678537605e-05,
+      "loss": 0.6373,
+      "step": 3391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971080751368521e-05,
+      "loss": 0.5733,
+      "step": 3392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9710637299489358e-05,
+      "loss": 0.6316,
+      "step": 3393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.971046703595091e-05,
+      "loss": 0.6385,
+      "step": 3394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9710296723070735e-05,
+      "loss": 0.5698,
+      "step": 3395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97101263608497e-05,
+      "loss": 0.633,
+      "step": 3396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9709955949288672e-05,
+      "loss": 0.6386,
+      "step": 3397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9709785488388512e-05,
+      "loss": 0.6445,
+      "step": 3398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970961497815009e-05,
+      "loss": 0.6696,
+      "step": 3399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9709444418574265e-05,
+      "loss": 0.644,
+      "step": 3400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9709273809661915e-05,
+      "loss": 0.6307,
+      "step": 3401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97091031514139e-05,
+      "loss": 0.5585,
+      "step": 3402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970893244383109e-05,
+      "loss": 0.5898,
+      "step": 3403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970876168691435e-05,
+      "loss": 0.6131,
+      "step": 3404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9708590880664545e-05,
+      "loss": 0.6127,
+      "step": 3405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9708420025082547e-05,
+      "loss": 0.6093,
+      "step": 3406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9708249120169228e-05,
+      "loss": 0.6915,
+      "step": 3407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970807816592545e-05,
+      "loss": 0.6532,
+      "step": 3408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9707907162352083e-05,
+      "loss": 0.6373,
+      "step": 3409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9707736109449995e-05,
+      "loss": 0.6191,
+      "step": 3410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970756500722006e-05,
+      "loss": 0.6271,
+      "step": 3411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970739385566314e-05,
+      "loss": 0.6277,
+      "step": 3412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9707222654780114e-05,
+      "loss": 0.5631,
+      "step": 3413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9707051404571846e-05,
+      "loss": 0.6264,
+      "step": 3414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9706880105039202e-05,
+      "loss": 0.6865,
+      "step": 3415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970670875618306e-05,
+      "loss": 0.6666,
+      "step": 3416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9706537358004292e-05,
+      "loss": 0.5973,
+      "step": 3417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970636591050376e-05,
+      "loss": 0.6673,
+      "step": 3418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970619441368234e-05,
+      "loss": 0.62,
+      "step": 3419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9706022867540907e-05,
+      "loss": 0.6111,
+      "step": 3420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9705851272080328e-05,
+      "loss": 0.6387,
+      "step": 3421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9705679627301477e-05,
+      "loss": 0.6263,
+      "step": 3422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9705507933205226e-05,
+      "loss": 0.569,
+      "step": 3423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9705336189792444e-05,
+      "loss": 0.6451,
+      "step": 3424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9705164397064008e-05,
+      "loss": 0.6281,
+      "step": 3425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970499255502079e-05,
+      "loss": 0.5925,
+      "step": 3426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9704820663663663e-05,
+      "loss": 0.5678,
+      "step": 3427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97046487229935e-05,
+      "loss": 0.6606,
+      "step": 3428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970447673301117e-05,
+      "loss": 0.6386,
+      "step": 3429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9704304693717556e-05,
+      "loss": 0.6108,
+      "step": 3430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9704132605113528e-05,
+      "loss": 0.6835,
+      "step": 3431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970396046719996e-05,
+      "loss": 0.5824,
+      "step": 3432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9703788279977725e-05,
+      "loss": 0.592,
+      "step": 3433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97036160434477e-05,
+      "loss": 0.6202,
+      "step": 3434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970344375761076e-05,
+      "loss": 0.6023,
+      "step": 3435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970327142246778e-05,
+      "loss": 0.5506,
+      "step": 3436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9703099038019638e-05,
+      "loss": 0.6339,
+      "step": 3437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970292660426721e-05,
+      "loss": 0.6433,
+      "step": 3438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9702754121211368e-05,
+      "loss": 0.6351,
+      "step": 3439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970258158885299e-05,
+      "loss": 0.6439,
+      "step": 3440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9702409007192954e-05,
+      "loss": 0.6249,
+      "step": 3441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9702236376232138e-05,
+      "loss": 0.6755,
+      "step": 3442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9702063695971415e-05,
+      "loss": 0.58,
+      "step": 3443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9701890966411668e-05,
+      "loss": 0.6952,
+      "step": 3444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970171818755377e-05,
+      "loss": 0.6427,
+      "step": 3445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.97015453593986e-05,
+      "loss": 0.5976,
+      "step": 3446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9701372481947037e-05,
+      "loss": 0.596,
+      "step": 3447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970119955519996e-05,
+      "loss": 0.6367,
+      "step": 3448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9701026579158243e-05,
+      "loss": 0.5924,
+      "step": 3449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9700853553822777e-05,
+      "loss": 0.5977,
+      "step": 3450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9700680479194427e-05,
+      "loss": 0.5996,
+      "step": 3451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970050735527408e-05,
+      "loss": 0.5823,
+      "step": 3452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.970033418206261e-05,
+      "loss": 0.6646,
+      "step": 3453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9700160959560907e-05,
+      "loss": 0.6536,
+      "step": 3454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9699987687769844e-05,
+      "loss": 0.5862,
+      "step": 3455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9699814366690303e-05,
+      "loss": 0.5767,
+      "step": 3456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9699640996323165e-05,
+      "loss": 0.609,
+      "step": 3457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969946757666931e-05,
+      "loss": 0.5705,
+      "step": 3458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9699294107729618e-05,
+      "loss": 0.5874,
+      "step": 3459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969912058950497e-05,
+      "loss": 0.6026,
+      "step": 3460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9698947021996254e-05,
+      "loss": 0.5651,
+      "step": 3461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9698773405204344e-05,
+      "loss": 0.6354,
+      "step": 3462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969859973913013e-05,
+      "loss": 0.5651,
+      "step": 3463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969842602377449e-05,
+      "loss": 0.6296,
+      "step": 3464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9698252259138305e-05,
+      "loss": 0.6514,
+      "step": 3465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969807844522246e-05,
+      "loss": 0.6649,
+      "step": 3466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9697904582027842e-05,
+      "loss": 0.6519,
+      "step": 3467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9697730669555325e-05,
+      "loss": 0.5667,
+      "step": 3468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9697556707805802e-05,
+      "loss": 0.6068,
+      "step": 3469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969738269678015e-05,
+      "loss": 0.6116,
+      "step": 3470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969720863647926e-05,
+      "loss": 0.6786,
+      "step": 3471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9697034526904012e-05,
+      "loss": 0.6212,
+      "step": 3472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9696860368055292e-05,
+      "loss": 0.5836,
+      "step": 3473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9696686159933982e-05,
+      "loss": 0.6598,
+      "step": 3474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9696511902540975e-05,
+      "loss": 0.6016,
+      "step": 3475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9696337595877146e-05,
+      "loss": 0.6732,
+      "step": 3476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969616323994339e-05,
+      "loss": 0.6247,
+      "step": 3477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9695988834740585e-05,
+      "loss": 0.5956,
+      "step": 3478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969581438026962e-05,
+      "loss": 0.7042,
+      "step": 3479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969563987653139e-05,
+      "loss": 0.6375,
+      "step": 3480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9695465323526767e-05,
+      "loss": 0.6095,
+      "step": 3481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969529072125665e-05,
+      "loss": 0.6196,
+      "step": 3482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969511606972192e-05,
+      "loss": 0.5875,
+      "step": 3483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694941368923464e-05,
+      "loss": 0.657,
+      "step": 3484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694766618862173e-05,
+      "loss": 0.6136,
+      "step": 3485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694591819538934e-05,
+      "loss": 0.6867,
+      "step": 3486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694416970954636e-05,
+      "loss": 0.6153,
+      "step": 3487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694242073110165e-05,
+      "loss": 0.6175,
+      "step": 3488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9694067126006408e-05,
+      "loss": 0.5683,
+      "step": 3489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969389212964426e-05,
+      "loss": 0.6145,
+      "step": 3490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9693717084024605e-05,
+      "loss": 0.6182,
+      "step": 3491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9693541989148338e-05,
+      "loss": 0.5776,
+      "step": 3492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9693366845016345e-05,
+      "loss": 0.6272,
+      "step": 3493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9693191651629512e-05,
+      "loss": 0.6594,
+      "step": 3494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969301640898874e-05,
+      "loss": 0.618,
+      "step": 3495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9692841117094906e-05,
+      "loss": 0.6047,
+      "step": 3496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9692665775948912e-05,
+      "loss": 0.6427,
+      "step": 3497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969249038555164e-05,
+      "loss": 0.6505,
+      "step": 3498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9692314945903992e-05,
+      "loss": 0.6404,
+      "step": 3499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969213945700685e-05,
+      "loss": 0.5811,
+      "step": 3500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969196391886111e-05,
+      "loss": 0.5838,
+      "step": 3501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969178833146766e-05,
+      "loss": 0.6135,
+      "step": 3502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9691612694827397e-05,
+      "loss": 0.6118,
+      "step": 3503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9691437008941212e-05,
+      "loss": 0.6063,
+      "step": 3504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969126127381e-05,
+      "loss": 0.6006,
+      "step": 3505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9691085489434645e-05,
+      "loss": 0.6201,
+      "step": 3506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969090965581605e-05,
+      "loss": 0.5974,
+      "step": 3507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9690733772955103e-05,
+      "loss": 0.6251,
+      "step": 3508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9690557840852703e-05,
+      "loss": 0.5697,
+      "step": 3509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.969038185950974e-05,
+      "loss": 0.5972,
+      "step": 3510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9690205828927107e-05,
+      "loss": 0.6227,
+      "step": 3511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9690029749105702e-05,
+      "loss": 0.6423,
+      "step": 3512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9689853620046414e-05,
+      "loss": 0.6505,
+      "step": 3513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9689677441750148e-05,
+      "loss": 0.6108,
+      "step": 3514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968950121421779e-05,
+      "loss": 0.5955,
+      "step": 3515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968932493745024e-05,
+      "loss": 0.6245,
+      "step": 3516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968914861144839e-05,
+      "loss": 0.6514,
+      "step": 3517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9688972236213136e-05,
+      "loss": 0.604,
+      "step": 3518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9688795811745382e-05,
+      "loss": 0.642,
+      "step": 3519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9688619338046016e-05,
+      "loss": 0.6164,
+      "step": 3520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968844281511594e-05,
+      "loss": 0.6346,
+      "step": 3521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9688266242956045e-05,
+      "loss": 0.6195,
+      "step": 3522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9688089621567236e-05,
+      "loss": 0.629,
+      "step": 3523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9687912950950406e-05,
+      "loss": 0.5972,
+      "step": 3524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9687736231106448e-05,
+      "loss": 0.6822,
+      "step": 3525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9687559462036268e-05,
+      "loss": 0.5952,
+      "step": 3526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9687382643740762e-05,
+      "loss": 0.5849,
+      "step": 3527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968720577622083e-05,
+      "loss": 0.6432,
+      "step": 3528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9687028859477365e-05,
+      "loss": 0.6068,
+      "step": 3529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968685189351127e-05,
+      "loss": 0.5917,
+      "step": 3530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9686674878323443e-05,
+      "loss": 0.6078,
+      "step": 3531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9686497813914786e-05,
+      "loss": 0.511,
+      "step": 3532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9686320700286193e-05,
+      "loss": 0.6118,
+      "step": 3533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9686143537438572e-05,
+      "loss": 0.6156,
+      "step": 3534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9685966325372817e-05,
+      "loss": 0.5651,
+      "step": 3535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9685789064089834e-05,
+      "loss": 0.5847,
+      "step": 3536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9685611753590516e-05,
+      "loss": 0.6585,
+      "step": 3537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9685434393875768e-05,
+      "loss": 0.6178,
+      "step": 3538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9685256984946495e-05,
+      "loss": 0.6348,
+      "step": 3539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968507952680359e-05,
+      "loss": 0.5893,
+      "step": 3540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9684902019447962e-05,
+      "loss": 0.6313,
+      "step": 3541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968472446288051e-05,
+      "loss": 0.636,
+      "step": 3542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968454685710214e-05,
+      "loss": 0.5805,
+      "step": 3543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9684369202113743e-05,
+      "loss": 0.6542,
+      "step": 3544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968419149791624e-05,
+      "loss": 0.6251,
+      "step": 3545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9684013744510518e-05,
+      "loss": 0.6348,
+      "step": 3546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9683835941897485e-05,
+      "loss": 0.6214,
+      "step": 3547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968365809007805e-05,
+      "loss": 0.6479,
+      "step": 3548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968348018905311e-05,
+      "loss": 0.6409,
+      "step": 3549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968330223882357e-05,
+      "loss": 0.629,
+      "step": 3550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9683124239390336e-05,
+      "loss": 0.606,
+      "step": 3551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9682946190754313e-05,
+      "loss": 0.6361,
+      "step": 3552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9682768092916404e-05,
+      "loss": 0.5584,
+      "step": 3553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9682589945877515e-05,
+      "loss": 0.6042,
+      "step": 3554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968241174963855e-05,
+      "loss": 0.6296,
+      "step": 3555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9682233504200414e-05,
+      "loss": 0.6088,
+      "step": 3556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9682055209564014e-05,
+      "loss": 0.6201,
+      "step": 3557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968187686573026e-05,
+      "loss": 0.611,
+      "step": 3558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968169847270005e-05,
+      "loss": 0.6146,
+      "step": 3559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.96815200304743e-05,
+      "loss": 0.6499,
+      "step": 3560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9681341539053905e-05,
+      "loss": 0.5563,
+      "step": 3561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968116299843978e-05,
+      "loss": 0.6239,
+      "step": 3562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9680984408632834e-05,
+      "loss": 0.6577,
+      "step": 3563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9680805769633967e-05,
+      "loss": 0.5611,
+      "step": 3564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968062708144409e-05,
+      "loss": 0.6625,
+      "step": 3565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9680448344064116e-05,
+      "loss": 0.6404,
+      "step": 3566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.968026955749495e-05,
+      "loss": 0.5891,
+      "step": 3567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9680090721737493e-05,
+      "loss": 0.5923,
+      "step": 3568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9679911836792666e-05,
+      "loss": 0.5956,
+      "step": 3569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9679732902661367e-05,
+      "loss": 0.6073,
+      "step": 3570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9679553919344514e-05,
+      "loss": 0.6296,
+      "step": 3571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9679374886843014e-05,
+      "loss": 0.6013,
+      "step": 3572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9679195805157772e-05,
+      "loss": 0.556,
+      "step": 3573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.96790166742897e-05,
+      "loss": 0.6167,
+      "step": 3574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9678837494239713e-05,
+      "loss": 0.6667,
+      "step": 3575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9678658265008715e-05,
+      "loss": 0.595,
+      "step": 3576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9678478986597625e-05,
+      "loss": 0.5998,
+      "step": 3577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9678299659007345e-05,
+      "loss": 0.6307,
+      "step": 3578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967812028223879e-05,
+      "loss": 0.6385,
+      "step": 3579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9677940856292874e-05,
+      "loss": 0.5805,
+      "step": 3580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9677761381170505e-05,
+      "loss": 0.6294,
+      "step": 3581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9677581856872594e-05,
+      "loss": 0.6094,
+      "step": 3582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9677402283400058e-05,
+      "loss": 0.6421,
+      "step": 3583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9677222660753806e-05,
+      "loss": 0.7149,
+      "step": 3584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967704298893475e-05,
+      "loss": 0.6271,
+      "step": 3585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9676863267943806e-05,
+      "loss": 0.5737,
+      "step": 3586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9676683497781884e-05,
+      "loss": 0.5982,
+      "step": 3587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9676503678449903e-05,
+      "loss": 0.6226,
+      "step": 3588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967632380994877e-05,
+      "loss": 0.6336,
+      "step": 3589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9676143892279405e-05,
+      "loss": 0.5943,
+      "step": 3590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9675963925442718e-05,
+      "loss": 0.676,
+      "step": 3591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967578390943962e-05,
+      "loss": 0.6651,
+      "step": 3592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9675603844271034e-05,
+      "loss": 0.6014,
+      "step": 3593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967542372993787e-05,
+      "loss": 0.5815,
+      "step": 3594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9675243566441047e-05,
+      "loss": 0.6475,
+      "step": 3595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9675063353781477e-05,
+      "loss": 0.6329,
+      "step": 3596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9674883091960074e-05,
+      "loss": 0.5799,
+      "step": 3597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967470278097776e-05,
+      "loss": 0.6039,
+      "step": 3598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9674522420835445e-05,
+      "loss": 0.6614,
+      "step": 3599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.967434201153405e-05,
+      "loss": 0.6183,
+      "step": 3600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9674161553074488e-05,
+      "loss": 0.6665,
+      "step": 3601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967398104545768e-05,
+      "loss": 0.583,
+      "step": 3602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9673800488684544e-05,
+      "loss": 0.6531,
+      "step": 3603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967361988275599e-05,
+      "loss": 0.6131,
+      "step": 3604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967343922767294e-05,
+      "loss": 0.6112,
+      "step": 3605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9673258523436314e-05,
+      "loss": 0.5417,
+      "step": 3606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9673077770047027e-05,
+      "loss": 0.6367,
+      "step": 3607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9672896967506e-05,
+      "loss": 0.6016,
+      "step": 3608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9672716115814153e-05,
+      "loss": 0.5997,
+      "step": 3609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.96725352149724e-05,
+      "loss": 0.642,
+      "step": 3610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9672354264981663e-05,
+      "loss": 0.6196,
+      "step": 3611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9672173265842863e-05,
+      "loss": 0.5738,
+      "step": 3612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9671992217556915e-05,
+      "loss": 0.5827,
+      "step": 3613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9671811120124742e-05,
+      "loss": 0.5335,
+      "step": 3614
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9671629973547265e-05,
+      "loss": 0.6224,
+      "step": 3615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967144877782541e-05,
+      "loss": 0.5986,
+      "step": 3616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9671267532960082e-05,
+      "loss": 0.5352,
+      "step": 3617
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9671086238952217e-05,
+      "loss": 0.6438,
+      "step": 3618
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967090489580273e-05,
+      "loss": 0.6451,
+      "step": 3619
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967072350351254e-05,
+      "loss": 0.6402,
+      "step": 3620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967054206208257e-05,
+      "loss": 0.6262,
+      "step": 3621
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9670360571513747e-05,
+      "loss": 0.6861,
+      "step": 3622
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.967017903180699e-05,
+      "loss": 0.684,
+      "step": 3623
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9669997442963223e-05,
+      "loss": 0.5963,
+      "step": 3624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9669815804983362e-05,
+      "loss": 0.6302,
+      "step": 3625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966963411786834e-05,
+      "loss": 0.6351,
+      "step": 3626
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9669452381619074e-05,
+      "loss": 0.6442,
+      "step": 3627
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9669270596236486e-05,
+      "loss": 0.5599,
+      "step": 3628
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9669088761721505e-05,
+      "loss": 0.6377,
+      "step": 3629
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9668906878075052e-05,
+      "loss": 0.6135,
+      "step": 3630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9668724945298052e-05,
+      "loss": 0.6042,
+      "step": 3631
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9668542963391427e-05,
+      "loss": 0.7084,
+      "step": 3632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9668360932356104e-05,
+      "loss": 0.6699,
+      "step": 3633
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9668178852193007e-05,
+      "loss": 0.665,
+      "step": 3634
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9667996722903062e-05,
+      "loss": 0.6505,
+      "step": 3635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.96678145444872e-05,
+      "loss": 0.5763,
+      "step": 3636
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966763231694633e-05,
+      "loss": 0.5788,
+      "step": 3637
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9667450040281398e-05,
+      "loss": 0.6329,
+      "step": 3638
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9667267714493318e-05,
+      "loss": 0.5676,
+      "step": 3639
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966708533958302e-05,
+      "loss": 0.559,
+      "step": 3640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966690291555143e-05,
+      "loss": 0.6339,
+      "step": 3641
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9666720442399475e-05,
+      "loss": 0.6782,
+      "step": 3642
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9666537920128082e-05,
+      "loss": 0.5843,
+      "step": 3643
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966635534873818e-05,
+      "loss": 0.5507,
+      "step": 3644
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9666172728230696e-05,
+      "loss": 0.6205,
+      "step": 3645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9665990058606552e-05,
+      "loss": 0.6364,
+      "step": 3646
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9665807339866688e-05,
+      "loss": 0.6421,
+      "step": 3647
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9665624572012024e-05,
+      "loss": 0.6624,
+      "step": 3648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966544175504349e-05,
+      "loss": 0.6409,
+      "step": 3649
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9665258888962018e-05,
+      "loss": 0.6216,
+      "step": 3650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9665075973768535e-05,
+      "loss": 0.6206,
+      "step": 3651
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9664893009463966e-05,
+      "loss": 0.6078,
+      "step": 3652
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966470999604925e-05,
+      "loss": 0.616,
+      "step": 3653
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966452693352531e-05,
+      "loss": 0.5545,
+      "step": 3654
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966434382189308e-05,
+      "loss": 0.676,
+      "step": 3655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9664160661153486e-05,
+      "loss": 0.6003,
+      "step": 3656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9663977451307463e-05,
+      "loss": 0.6347,
+      "step": 3657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9663794192355937e-05,
+      "loss": 0.5997,
+      "step": 3658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966361088429985e-05,
+      "loss": 0.6504,
+      "step": 3659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966342752714012e-05,
+      "loss": 0.6658,
+      "step": 3660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9663244120877685e-05,
+      "loss": 0.5897,
+      "step": 3661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966306066551348e-05,
+      "loss": 0.6473,
+      "step": 3662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966287716104843e-05,
+      "loss": 0.6156,
+      "step": 3663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9662693607483473e-05,
+      "loss": 0.625,
+      "step": 3664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966251000481954e-05,
+      "loss": 0.6321,
+      "step": 3665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9662326353057563e-05,
+      "loss": 0.6457,
+      "step": 3666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966214265219848e-05,
+      "loss": 0.6678,
+      "step": 3667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9661958902243213e-05,
+      "loss": 0.6526,
+      "step": 3668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966177510319271e-05,
+      "loss": 0.6387,
+      "step": 3669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9661591255047896e-05,
+      "loss": 0.671,
+      "step": 3670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966140735780971e-05,
+      "loss": 0.6032,
+      "step": 3671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966122341147908e-05,
+      "loss": 0.5982,
+      "step": 3672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9661039416056943e-05,
+      "loss": 0.6941,
+      "step": 3673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9660855371544237e-05,
+      "loss": 0.6318,
+      "step": 3674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9660671277941895e-05,
+      "loss": 0.6111,
+      "step": 3675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9660487135250854e-05,
+      "loss": 0.6356,
+      "step": 3676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.966030294347205e-05,
+      "loss": 0.5824,
+      "step": 3677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9660118702606414e-05,
+      "loss": 0.5988,
+      "step": 3678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965993441265489e-05,
+      "loss": 0.597,
+      "step": 3679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9659750073618406e-05,
+      "loss": 0.5852,
+      "step": 3680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9659565685497904e-05,
+      "loss": 0.679,
+      "step": 3681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9659381248294324e-05,
+      "loss": 0.6163,
+      "step": 3682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9659196762008595e-05,
+      "loss": 0.6184,
+      "step": 3683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965901222664166e-05,
+      "loss": 0.6261,
+      "step": 3684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9658827642194454e-05,
+      "loss": 0.5687,
+      "step": 3685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9658643008667915e-05,
+      "loss": 0.6429,
+      "step": 3686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9658458326062985e-05,
+      "loss": 0.5955,
+      "step": 3687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.96582735943806e-05,
+      "loss": 0.6249,
+      "step": 3688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9658088813621693e-05,
+      "loss": 0.6482,
+      "step": 3689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9657903983787215e-05,
+      "loss": 0.6264,
+      "step": 3690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9657719104878093e-05,
+      "loss": 0.5553,
+      "step": 3691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9657534176895275e-05,
+      "loss": 0.612,
+      "step": 3692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9657349199839693e-05,
+      "loss": 0.6619,
+      "step": 3693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.96571641737123e-05,
+      "loss": 0.5913,
+      "step": 3694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965697909851402e-05,
+      "loss": 0.5673,
+      "step": 3695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9656793974245802e-05,
+      "loss": 0.6402,
+      "step": 3696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9656608800908587e-05,
+      "loss": 0.6482,
+      "step": 3697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9656423578503316e-05,
+      "loss": 0.6341,
+      "step": 3698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965623830703093e-05,
+      "loss": 0.587,
+      "step": 3699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9656052986492367e-05,
+      "loss": 0.6656,
+      "step": 3700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9655867616888572e-05,
+      "loss": 0.5762,
+      "step": 3701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965568219822048e-05,
+      "loss": 0.6829,
+      "step": 3702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9655496730489045e-05,
+      "loss": 0.5947,
+      "step": 3703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9655311213695203e-05,
+      "loss": 0.6028,
+      "step": 3704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9655125647839895e-05,
+      "loss": 0.6485,
+      "step": 3705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9654940032924066e-05,
+      "loss": 0.6485,
+      "step": 3706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965475436894866e-05,
+      "loss": 0.5888,
+      "step": 3707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9654568655914622e-05,
+      "loss": 0.6426,
+      "step": 3708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9654382893822888e-05,
+      "loss": 0.6075,
+      "step": 3709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965419708267441e-05,
+      "loss": 0.5793,
+      "step": 3710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965401122247013e-05,
+      "loss": 0.5873,
+      "step": 3711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965382531321099e-05,
+      "loss": 0.5679,
+      "step": 3712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965363935489794e-05,
+      "loss": 0.5632,
+      "step": 3713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9653453347531918e-05,
+      "loss": 0.6375,
+      "step": 3714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965326729111387e-05,
+      "loss": 0.5899,
+      "step": 3715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9653081185644748e-05,
+      "loss": 0.6416,
+      "step": 3716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965289503112549e-05,
+      "loss": 0.6712,
+      "step": 3717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965270882755705e-05,
+      "loss": 0.6009,
+      "step": 3718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9652522574940366e-05,
+      "loss": 0.6504,
+      "step": 3719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965233627327639e-05,
+      "loss": 0.6041,
+      "step": 3720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9652149922566065e-05,
+      "loss": 0.6398,
+      "step": 3721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9651963522810337e-05,
+      "loss": 0.5735,
+      "step": 3722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965177707401016e-05,
+      "loss": 0.5416,
+      "step": 3723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9651590576166474e-05,
+      "loss": 0.6036,
+      "step": 3724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9651404029280232e-05,
+      "loss": 0.6131,
+      "step": 3725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965121743335238e-05,
+      "loss": 0.5827,
+      "step": 3726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965103078838386e-05,
+      "loss": 0.6719,
+      "step": 3727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9650844094375634e-05,
+      "loss": 0.5895,
+      "step": 3728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965065735132864e-05,
+      "loss": 0.6329,
+      "step": 3729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.965047055924383e-05,
+      "loss": 0.5721,
+      "step": 3730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9650283718122153e-05,
+      "loss": 0.6897,
+      "step": 3731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9650096827964554e-05,
+      "loss": 0.6711,
+      "step": 3732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964990988877199e-05,
+      "loss": 0.5692,
+      "step": 3733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964972290054541e-05,
+      "loss": 0.6413,
+      "step": 3734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964953586328576e-05,
+      "loss": 0.5661,
+      "step": 3735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964934877699399e-05,
+      "loss": 0.6381,
+      "step": 3736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9649161641671054e-05,
+      "loss": 0.643,
+      "step": 3737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9648974457317903e-05,
+      "loss": 0.5948,
+      "step": 3738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9648787223935486e-05,
+      "loss": 0.6393,
+      "step": 3739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964859994152476e-05,
+      "loss": 0.5739,
+      "step": 3740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9648412610086664e-05,
+      "loss": 0.5926,
+      "step": 3741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9648225229622163e-05,
+      "loss": 0.625,
+      "step": 3742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.96480378001322e-05,
+      "loss": 0.5997,
+      "step": 3743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9647850321617734e-05,
+      "loss": 0.6818,
+      "step": 3744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9647662794079715e-05,
+      "loss": 0.6584,
+      "step": 3745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9647475217519097e-05,
+      "loss": 0.575,
+      "step": 3746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.964728759193683e-05,
+      "loss": 0.6077,
+      "step": 3747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964709991733387e-05,
+      "loss": 0.6907,
+      "step": 3748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9646912193711173e-05,
+      "loss": 0.5763,
+      "step": 3749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964672442106969e-05,
+      "loss": 0.6017,
+      "step": 3750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9646536599410367e-05,
+      "loss": 0.6278,
+      "step": 3751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9646348728734174e-05,
+      "loss": 0.6121,
+      "step": 3752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9646160809042055e-05,
+      "loss": 0.676,
+      "step": 3753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9645972840334968e-05,
+      "loss": 0.626,
+      "step": 3754
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964578482261387e-05,
+      "loss": 0.6669,
+      "step": 3755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9645596755879713e-05,
+      "loss": 0.6229,
+      "step": 3756
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9645408640133456e-05,
+      "loss": 0.6683,
+      "step": 3757
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964522047537605e-05,
+      "loss": 0.668,
+      "step": 3758
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9645032261608453e-05,
+      "loss": 0.641,
+      "step": 3759
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9644843998831626e-05,
+      "loss": 0.6271,
+      "step": 3760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964465568704652e-05,
+      "loss": 0.6038,
+      "step": 3761
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9644467326254094e-05,
+      "loss": 0.6458,
+      "step": 3762
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9644278916455308e-05,
+      "loss": 0.6125,
+      "step": 3763
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9644090457651112e-05,
+      "loss": 0.656,
+      "step": 3764
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964390194984247e-05,
+      "loss": 0.5819,
+      "step": 3765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964371339303034e-05,
+      "loss": 0.6096,
+      "step": 3766
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9643524787215672e-05,
+      "loss": 0.7037,
+      "step": 3767
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9643336132399436e-05,
+      "loss": 0.5715,
+      "step": 3768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9643147428582585e-05,
+      "loss": 0.6189,
+      "step": 3769
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9642958675766073e-05,
+      "loss": 0.6335,
+      "step": 3770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9642769873950868e-05,
+      "loss": 0.609,
+      "step": 3771
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9642581023137923e-05,
+      "loss": 0.6238,
+      "step": 3772
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.96423921233282e-05,
+      "loss": 0.6356,
+      "step": 3773
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964220317452266e-05,
+      "loss": 0.5787,
+      "step": 3774
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9642014176722265e-05,
+      "loss": 0.6257,
+      "step": 3775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9641825129927967e-05,
+      "loss": 0.6461,
+      "step": 3776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964163603414073e-05,
+      "loss": 0.5969,
+      "step": 3777
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9641446889361522e-05,
+      "loss": 0.6092,
+      "step": 3778
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9641257695591297e-05,
+      "loss": 0.5419,
+      "step": 3779
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9641068452831017e-05,
+      "loss": 0.6253,
+      "step": 3780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9640879161081644e-05,
+      "loss": 0.6225,
+      "step": 3781
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.964068982034414e-05,
+      "loss": 0.5542,
+      "step": 3782
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9640500430619474e-05,
+      "loss": 0.5902,
+      "step": 3783
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9640310991908595e-05,
+      "loss": 0.6116,
+      "step": 3784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9640121504212475e-05,
+      "loss": 0.6447,
+      "step": 3785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9639931967532074e-05,
+      "loss": 0.5997,
+      "step": 3786
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9639742381868358e-05,
+      "loss": 0.6145,
+      "step": 3787
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9639552747222284e-05,
+      "loss": 0.6654,
+      "step": 3788
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9639363063594823e-05,
+      "loss": 0.689,
+      "step": 3789
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9639173330986933e-05,
+      "loss": 0.7336,
+      "step": 3790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963898354939958e-05,
+      "loss": 0.6389,
+      "step": 3791
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963879371883373e-05,
+      "loss": 0.6009,
+      "step": 3792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9638603839290345e-05,
+      "loss": 0.6458,
+      "step": 3793
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963841391077039e-05,
+      "loss": 0.5627,
+      "step": 3794
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9638223933274837e-05,
+      "loss": 0.5955,
+      "step": 3795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9638033906804638e-05,
+      "loss": 0.642,
+      "step": 3796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963784383136077e-05,
+      "loss": 0.6315,
+      "step": 3797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9637653706944196e-05,
+      "loss": 0.6391,
+      "step": 3798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9637463533555877e-05,
+      "loss": 0.6422,
+      "step": 3799
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9637273311196786e-05,
+      "loss": 0.63,
+      "step": 3800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9637083039867888e-05,
+      "loss": 0.6086,
+      "step": 3801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963689271957015e-05,
+      "loss": 0.6009,
+      "step": 3802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963670235030453e-05,
+      "loss": 0.6137,
+      "step": 3803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963651193207201e-05,
+      "loss": 0.5899,
+      "step": 3804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963632146487355e-05,
+      "loss": 0.6793,
+      "step": 3805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9636130948710117e-05,
+      "loss": 0.5977,
+      "step": 3806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963594038358268e-05,
+      "loss": 0.5808,
+      "step": 3807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963574976949221e-05,
+      "loss": 0.5753,
+      "step": 3808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963555910643967e-05,
+      "loss": 0.6067,
+      "step": 3809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9635368394426036e-05,
+      "loss": 0.6022,
+      "step": 3810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9635177633452273e-05,
+      "loss": 0.6224,
+      "step": 3811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963498682351935e-05,
+      "loss": 0.5993,
+      "step": 3812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9634795964628236e-05,
+      "loss": 0.5505,
+      "step": 3813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9634605056779902e-05,
+      "loss": 0.6442,
+      "step": 3814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963441409997532e-05,
+      "loss": 0.5799,
+      "step": 3815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963422309421546e-05,
+      "loss": 0.6359,
+      "step": 3816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9634032039501286e-05,
+      "loss": 0.6002,
+      "step": 3817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9633840935833775e-05,
+      "loss": 0.6125,
+      "step": 3818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.96336497832139e-05,
+      "loss": 0.6337,
+      "step": 3819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9633458581642627e-05,
+      "loss": 0.6715,
+      "step": 3820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9633267331120927e-05,
+      "loss": 0.6967,
+      "step": 3821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963307603164978e-05,
+      "loss": 0.6048,
+      "step": 3822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9632884683230147e-05,
+      "loss": 0.6345,
+      "step": 3823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9632693285863008e-05,
+      "loss": 0.6106,
+      "step": 3824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9632501839549335e-05,
+      "loss": 0.618,
+      "step": 3825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.96323103442901e-05,
+      "loss": 0.5971,
+      "step": 3826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9632118800086272e-05,
+      "loss": 0.6311,
+      "step": 3827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9631927206938827e-05,
+      "loss": 0.572,
+      "step": 3828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963173556484874e-05,
+      "loss": 0.6864,
+      "step": 3829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9631543873816987e-05,
+      "loss": 0.5911,
+      "step": 3830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9631352133844536e-05,
+      "loss": 0.6376,
+      "step": 3831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.963116034493236e-05,
+      "loss": 0.6498,
+      "step": 3832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630968507081446e-05,
+      "loss": 0.6574,
+      "step": 3833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630776620292757e-05,
+      "loss": 0.6814,
+      "step": 3834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630584684567272e-05,
+      "loss": 0.5979,
+      "step": 3835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630392699905966e-05,
+      "loss": 0.6109,
+      "step": 3836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630200666309812e-05,
+      "loss": 0.5946,
+      "step": 3837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9630008583779788e-05,
+      "loss": 0.6137,
+      "step": 3838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962981645231687e-05,
+      "loss": 0.5811,
+      "step": 3839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9629624271922037e-05,
+      "loss": 0.6648,
+      "step": 3840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962943204259626e-05,
+      "loss": 0.7128,
+      "step": 3841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962923976434052e-05,
+      "loss": 0.6448,
+      "step": 3842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9629047437155794e-05,
+      "loss": 0.586,
+      "step": 3843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9628855061043057e-05,
+      "loss": 0.6185,
+      "step": 3844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9628662636003288e-05,
+      "loss": 0.6059,
+      "step": 3845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9628470162037465e-05,
+      "loss": 0.6432,
+      "step": 3846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9628277639146563e-05,
+      "loss": 0.5353,
+      "step": 3847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9628085067331562e-05,
+      "loss": 0.6026,
+      "step": 3848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9627892446593443e-05,
+      "loss": 0.5768,
+      "step": 3849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962769977693318e-05,
+      "loss": 0.6444,
+      "step": 3850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962750705835176e-05,
+      "loss": 0.5508,
+      "step": 3851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9627314290850154e-05,
+      "loss": 0.6984,
+      "step": 3852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962712147442935e-05,
+      "loss": 0.5511,
+      "step": 3853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9626928609090313e-05,
+      "loss": 0.5829,
+      "step": 3854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9626735694834037e-05,
+      "loss": 0.5653,
+      "step": 3855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9626542731661494e-05,
+      "loss": 0.6264,
+      "step": 3856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9626349719573675e-05,
+      "loss": 0.6299,
+      "step": 3857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962615665857155e-05,
+      "loss": 0.5721,
+      "step": 3858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9625963548656105e-05,
+      "loss": 0.5996,
+      "step": 3859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962577038982832e-05,
+      "loss": 0.6001,
+      "step": 3860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9625577182089174e-05,
+      "loss": 0.6294,
+      "step": 3861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9625383925439654e-05,
+      "loss": 0.6275,
+      "step": 3862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962519061988074e-05,
+      "loss": 0.6276,
+      "step": 3863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9624997265413408e-05,
+      "loss": 0.631,
+      "step": 3864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962480386203865e-05,
+      "loss": 0.6528,
+      "step": 3865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9624610409757446e-05,
+      "loss": 0.607,
+      "step": 3866
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9624416908570777e-05,
+      "loss": 0.5926,
+      "step": 3867
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9624223358479628e-05,
+      "loss": 0.6594,
+      "step": 3868
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962402975948498e-05,
+      "loss": 0.6653,
+      "step": 3869
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962383611158782e-05,
+      "loss": 0.683,
+      "step": 3870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9623642414789125e-05,
+      "loss": 0.6473,
+      "step": 3871
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962344866908989e-05,
+      "loss": 0.6165,
+      "step": 3872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9623254874491092e-05,
+      "loss": 0.59,
+      "step": 3873
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9623061030993717e-05,
+      "loss": 0.6362,
+      "step": 3874
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9622867138598752e-05,
+      "loss": 0.6797,
+      "step": 3875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962267319730718e-05,
+      "loss": 0.6491,
+      "step": 3876
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962247920711999e-05,
+      "loss": 0.6032,
+      "step": 3877
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9622285168038163e-05,
+      "loss": 0.6646,
+      "step": 3878
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9622091080062688e-05,
+      "loss": 0.6043,
+      "step": 3879
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962189694319455e-05,
+      "loss": 0.6126,
+      "step": 3880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962170275743473e-05,
+      "loss": 0.6213,
+      "step": 3881
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962150852278423e-05,
+      "loss": 0.6667,
+      "step": 3882
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9621314239244022e-05,
+      "loss": 0.6089,
+      "step": 3883
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.96211199068151e-05,
+      "loss": 0.6587,
+      "step": 3884
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.962092552549845e-05,
+      "loss": 0.6804,
+      "step": 3885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9620731095295057e-05,
+      "loss": 0.5991,
+      "step": 3886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9620536616205913e-05,
+      "loss": 0.6054,
+      "step": 3887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9620342088232008e-05,
+      "loss": 0.6152,
+      "step": 3888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9620147511374326e-05,
+      "loss": 0.6176,
+      "step": 3889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.961995288563386e-05,
+      "loss": 0.6149,
+      "step": 3890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.961975821101159e-05,
+      "loss": 0.5958,
+      "step": 3891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.961956348750851e-05,
+      "loss": 0.6147,
+      "step": 3892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9619368715125617e-05,
+      "loss": 0.5507,
+      "step": 3893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961917389386389e-05,
+      "loss": 0.6305,
+      "step": 3894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9618979023724328e-05,
+      "loss": 0.6015,
+      "step": 3895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9618784104707915e-05,
+      "loss": 0.661,
+      "step": 3896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961858913681564e-05,
+      "loss": 0.6242,
+      "step": 3897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9618394120048498e-05,
+      "loss": 0.6017,
+      "step": 3898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961819905440748e-05,
+      "loss": 0.6278,
+      "step": 3899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9618003939893575e-05,
+      "loss": 0.5833,
+      "step": 3900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9617808776507773e-05,
+      "loss": 0.6318,
+      "step": 3901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961761356425107e-05,
+      "loss": 0.6012,
+      "step": 3902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9617418303124458e-05,
+      "loss": 0.6021,
+      "step": 3903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9617222993128925e-05,
+      "loss": 0.5702,
+      "step": 3904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9617027634265463e-05,
+      "loss": 0.6787,
+      "step": 3905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961683222653507e-05,
+      "loss": 0.5607,
+      "step": 3906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961663676993874e-05,
+      "loss": 0.6847,
+      "step": 3907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9616441264477455e-05,
+      "loss": 0.5309,
+      "step": 3908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9616245710152216e-05,
+      "loss": 0.6144,
+      "step": 3909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961605010696402e-05,
+      "loss": 0.6105,
+      "step": 3910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9615854454913858e-05,
+      "loss": 0.622,
+      "step": 3911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9615658754002717e-05,
+      "loss": 0.6357,
+      "step": 3912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.96154630042316e-05,
+      "loss": 0.65,
+      "step": 3913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9615267205601505e-05,
+      "loss": 0.658,
+      "step": 3914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9615071358113416e-05,
+      "loss": 0.6385,
+      "step": 3915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961487546176834e-05,
+      "loss": 0.6112,
+      "step": 3916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9614679516567257e-05,
+      "loss": 0.6078,
+      "step": 3917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9614483522511176e-05,
+      "loss": 0.6452,
+      "step": 3918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9614287479601087e-05,
+      "loss": 0.5969,
+      "step": 3919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961409138783799e-05,
+      "loss": 0.6748,
+      "step": 3920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9613895247222878e-05,
+      "loss": 0.5625,
+      "step": 3921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961369905775675e-05,
+      "loss": 0.5857,
+      "step": 3922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9613502819440598e-05,
+      "loss": 0.6346,
+      "step": 3923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9613306532275427e-05,
+      "loss": 0.606,
+      "step": 3924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9613110196262227e-05,
+      "loss": 0.5649,
+      "step": 3925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9612913811402e-05,
+      "loss": 0.5913,
+      "step": 3926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961271737769574e-05,
+      "loss": 0.6067,
+      "step": 3927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9612520895144453e-05,
+      "loss": 0.6099,
+      "step": 3928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9612324363749132e-05,
+      "loss": 0.7008,
+      "step": 3929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961212778351077e-05,
+      "loss": 0.6649,
+      "step": 3930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.961193115443038e-05,
+      "loss": 0.625,
+      "step": 3931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9611734476508942e-05,
+      "loss": 0.5856,
+      "step": 3932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9611537749747474e-05,
+      "loss": 0.6337,
+      "step": 3933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9611340974146964e-05,
+      "loss": 0.6153,
+      "step": 3934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9611144149708418e-05,
+      "loss": 0.6689,
+      "step": 3935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9610947276432835e-05,
+      "loss": 0.6451,
+      "step": 3936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9610750354321212e-05,
+      "loss": 0.608,
+      "step": 3937
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9610553383374552e-05,
+      "loss": 0.5782,
+      "step": 3938
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9610356363593857e-05,
+      "loss": 0.6127,
+      "step": 3939
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9610159294980124e-05,
+      "loss": 0.5775,
+      "step": 3940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960996217753436e-05,
+      "loss": 0.6367,
+      "step": 3941
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9609765011257563e-05,
+      "loss": 0.6086,
+      "step": 3942
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9609567796150735e-05,
+      "loss": 0.6813,
+      "step": 3943
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9609370532214877e-05,
+      "loss": 0.5999,
+      "step": 3944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9609173219450998e-05,
+      "loss": 0.7049,
+      "step": 3945
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960897585786009e-05,
+      "loss": 0.6355,
+      "step": 3946
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9608778447443166e-05,
+      "loss": 0.6313,
+      "step": 3947
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9608580988201226e-05,
+      "loss": 0.6549,
+      "step": 3948
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9608383480135266e-05,
+      "loss": 0.5883,
+      "step": 3949
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.96081859232463e-05,
+      "loss": 0.669,
+      "step": 3950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9607988317535325e-05,
+      "loss": 0.6391,
+      "step": 3951
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960779066300335e-05,
+      "loss": 0.6059,
+      "step": 3952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9607592959651375e-05,
+      "loss": 0.5479,
+      "step": 3953
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9607395207480406e-05,
+      "loss": 0.6166,
+      "step": 3954
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960719740649145e-05,
+      "loss": 0.6388,
+      "step": 3955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9606999556685507e-05,
+      "loss": 0.5921,
+      "step": 3956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960680165806359e-05,
+      "loss": 0.6187,
+      "step": 3957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9606603710626698e-05,
+      "loss": 0.6496,
+      "step": 3958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9606405714375844e-05,
+      "loss": 0.5976,
+      "step": 3959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9606207669312023e-05,
+      "loss": 0.6733,
+      "step": 3960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960600957543625e-05,
+      "loss": 0.5994,
+      "step": 3961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9605811432749527e-05,
+      "loss": 0.6169,
+      "step": 3962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9605613241252866e-05,
+      "loss": 0.6601,
+      "step": 3963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960541500094727e-05,
+      "loss": 0.6309,
+      "step": 3964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9605216711833743e-05,
+      "loss": 0.6575,
+      "step": 3965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.96050183739133e-05,
+      "loss": 0.5998,
+      "step": 3966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9604819987186947e-05,
+      "loss": 0.5958,
+      "step": 3967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9604621551655692e-05,
+      "loss": 0.6214,
+      "step": 3968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960442306732054e-05,
+      "loss": 0.6593,
+      "step": 3969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.96042245341825e-05,
+      "loss": 0.5856,
+      "step": 3970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9604025952242587e-05,
+      "loss": 0.6122,
+      "step": 3971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.96038273215018e-05,
+      "loss": 0.5887,
+      "step": 3972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9603628641961158e-05,
+      "loss": 0.6355,
+      "step": 3973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9603429913621665e-05,
+      "loss": 0.5979,
+      "step": 3974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960323113648433e-05,
+      "loss": 0.6548,
+      "step": 3975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960303231055017e-05,
+      "loss": 0.6316,
+      "step": 3976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9602833435820185e-05,
+      "loss": 0.5888,
+      "step": 3977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9602634512295397e-05,
+      "loss": 0.6896,
+      "step": 3978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9602435539976805e-05,
+      "loss": 0.5683,
+      "step": 3979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9602236518865426e-05,
+      "loss": 0.6612,
+      "step": 3980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9602037448962276e-05,
+      "loss": 0.6617,
+      "step": 3981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9601838330268358e-05,
+      "loss": 0.6058,
+      "step": 3982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9601639162784688e-05,
+      "loss": 0.5769,
+      "step": 3983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9601439946512276e-05,
+      "loss": 0.6311,
+      "step": 3984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960124068145214e-05,
+      "loss": 0.6732,
+      "step": 3985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9601041367605282e-05,
+      "loss": 0.6184,
+      "step": 3986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9600842004972726e-05,
+      "loss": 0.6084,
+      "step": 3987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960064259355548e-05,
+      "loss": 0.6163,
+      "step": 3988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9600443133354554e-05,
+      "loss": 0.6126,
+      "step": 3989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9600243624370967e-05,
+      "loss": 0.5269,
+      "step": 3990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.960004406660573e-05,
+      "loss": 0.6507,
+      "step": 3991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959984446005986e-05,
+      "loss": 0.6084,
+      "step": 3992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9599644804734366e-05,
+      "loss": 0.6013,
+      "step": 3993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9599445100630266e-05,
+      "loss": 0.6031,
+      "step": 3994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9599245347748574e-05,
+      "loss": 0.6574,
+      "step": 3995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9599045546090306e-05,
+      "loss": 0.5846,
+      "step": 3996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9598845695656473e-05,
+      "loss": 0.6046,
+      "step": 3997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9598645796448098e-05,
+      "loss": 0.6213,
+      "step": 3998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9598445848466193e-05,
+      "loss": 0.5645,
+      "step": 3999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959824585171177e-05,
+      "loss": 0.6,
+      "step": 4000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959804580618585e-05,
+      "loss": 0.5935,
+      "step": 4001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959784571188945e-05,
+      "loss": 0.5833,
+      "step": 4002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9597645568823583e-05,
+      "loss": 0.6768,
+      "step": 4003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9597445376989272e-05,
+      "loss": 0.605,
+      "step": 4004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9597245136387526e-05,
+      "loss": 0.6642,
+      "step": 4005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959704484701937e-05,
+      "loss": 0.5331,
+      "step": 4006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9596844508885816e-05,
+      "loss": 0.6397,
+      "step": 4007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9596644121987884e-05,
+      "loss": 0.6409,
+      "step": 4008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9596443686326594e-05,
+      "loss": 0.614,
+      "step": 4009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9596243201902966e-05,
+      "loss": 0.5682,
+      "step": 4010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959604266871801e-05,
+      "loss": 0.5815,
+      "step": 4011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9595842086772753e-05,
+      "loss": 0.6212,
+      "step": 4012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9595641456068212e-05,
+      "loss": 0.6118,
+      "step": 4013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959544077660541e-05,
+      "loss": 0.6261,
+      "step": 4014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959524004838536e-05,
+      "loss": 0.5465,
+      "step": 4015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9595039271409087e-05,
+      "loss": 0.5828,
+      "step": 4016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9594838445677608e-05,
+      "loss": 0.6363,
+      "step": 4017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9594637571191947e-05,
+      "loss": 0.6171,
+      "step": 4018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959443664795312e-05,
+      "loss": 0.6193,
+      "step": 4019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959423567596215e-05,
+      "loss": 0.4922,
+      "step": 4020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959403465522006e-05,
+      "loss": 0.6559,
+      "step": 4021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959383358572787e-05,
+      "loss": 0.7047,
+      "step": 4022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9593632467486604e-05,
+      "loss": 0.6167,
+      "step": 4023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959343130049728e-05,
+      "loss": 0.6843,
+      "step": 4024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959323008476092e-05,
+      "loss": 0.5816,
+      "step": 4025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959302882027855e-05,
+      "loss": 0.6264,
+      "step": 4026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9592827507051192e-05,
+      "loss": 0.5904,
+      "step": 4027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9592626145079868e-05,
+      "loss": 0.6126,
+      "step": 4028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.95924247343656e-05,
+      "loss": 0.6271,
+      "step": 4029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9592223274909413e-05,
+      "loss": 0.5696,
+      "step": 4030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9592021766712332e-05,
+      "loss": 0.6396,
+      "step": 4031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959182020977538e-05,
+      "loss": 0.6369,
+      "step": 4032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9591618604099575e-05,
+      "loss": 0.6161,
+      "step": 4033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959141694968595e-05,
+      "loss": 0.6221,
+      "step": 4034
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959121524653553e-05,
+      "loss": 0.6059,
+      "step": 4035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959101349464933e-05,
+      "loss": 0.6407,
+      "step": 4036
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9590811694028387e-05,
+      "loss": 0.5691,
+      "step": 4037
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.959060984467372e-05,
+      "loss": 0.5822,
+      "step": 4038
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9590407946586353e-05,
+      "loss": 0.5791,
+      "step": 4039
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9590205999767317e-05,
+      "loss": 0.6095,
+      "step": 4040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9590004004217633e-05,
+      "loss": 0.6613,
+      "step": 4041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9589801959938332e-05,
+      "loss": 0.5968,
+      "step": 4042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958959986693044e-05,
+      "loss": 0.6085,
+      "step": 4043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9589397725194977e-05,
+      "loss": 0.6631,
+      "step": 4044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9589195534732982e-05,
+      "loss": 0.6906,
+      "step": 4045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9588993295545475e-05,
+      "loss": 0.6138,
+      "step": 4046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9588791007633485e-05,
+      "loss": 0.6243,
+      "step": 4047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958858867099804e-05,
+      "loss": 0.6187,
+      "step": 4048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9588386285640163e-05,
+      "loss": 0.5933,
+      "step": 4049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958818385156089e-05,
+      "loss": 0.6931,
+      "step": 4050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958798136876125e-05,
+      "loss": 0.6582,
+      "step": 4051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9587778837242265e-05,
+      "loss": 0.5656,
+      "step": 4052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9587576257004968e-05,
+      "loss": 0.6295,
+      "step": 4053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9587373628050388e-05,
+      "loss": 0.635,
+      "step": 4054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9587170950379558e-05,
+      "loss": 0.6936,
+      "step": 4055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9586968223993502e-05,
+      "loss": 0.6706,
+      "step": 4056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9586765448893253e-05,
+      "loss": 0.6253,
+      "step": 4057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958656262507984e-05,
+      "loss": 0.6645,
+      "step": 4058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9586359752554297e-05,
+      "loss": 0.6494,
+      "step": 4059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9586156831317652e-05,
+      "loss": 0.6454,
+      "step": 4060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9585953861370938e-05,
+      "loss": 0.6136,
+      "step": 4061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958575084271518e-05,
+      "loss": 0.6075,
+      "step": 4062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958554777535142e-05,
+      "loss": 0.6687,
+      "step": 4063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9585344659280682e-05,
+      "loss": 0.6016,
+      "step": 4064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9585141494504003e-05,
+      "loss": 0.5705,
+      "step": 4065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958493828102241e-05,
+      "loss": 0.6426,
+      "step": 4066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958473501883694e-05,
+      "loss": 0.6302,
+      "step": 4067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9584531707948623e-05,
+      "loss": 0.6326,
+      "step": 4068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9584328348358496e-05,
+      "loss": 0.6383,
+      "step": 4069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9584124940067587e-05,
+      "loss": 0.6087,
+      "step": 4070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9583921483076935e-05,
+      "loss": 0.6571,
+      "step": 4071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958371797738757e-05,
+      "loss": 0.6859,
+      "step": 4072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9583514423000527e-05,
+      "loss": 0.5805,
+      "step": 4073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958331081991684e-05,
+      "loss": 0.6458,
+      "step": 4074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9583107168137546e-05,
+      "loss": 0.6228,
+      "step": 4075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9582903467663675e-05,
+      "loss": 0.6613,
+      "step": 4076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958269971849627e-05,
+      "loss": 0.5973,
+      "step": 4077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9582495920636354e-05,
+      "loss": 0.5481,
+      "step": 4078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9582292074084973e-05,
+      "loss": 0.6217,
+      "step": 4079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958208817884316e-05,
+      "loss": 0.6149,
+      "step": 4080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9581884234911953e-05,
+      "loss": 0.6194,
+      "step": 4081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9581680242292383e-05,
+      "loss": 0.5839,
+      "step": 4082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9581476200985492e-05,
+      "loss": 0.6379,
+      "step": 4083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9581272110992313e-05,
+      "loss": 0.5679,
+      "step": 4084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9581067972313885e-05,
+      "loss": 0.5998,
+      "step": 4085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9580863784951246e-05,
+      "loss": 0.6647,
+      "step": 4086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958065954890543e-05,
+      "loss": 0.6714,
+      "step": 4087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9580455264177477e-05,
+      "loss": 0.6075,
+      "step": 4088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9580250930768425e-05,
+      "loss": 0.5897,
+      "step": 4089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.958004654867931e-05,
+      "loss": 0.6257,
+      "step": 4090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9579842117911177e-05,
+      "loss": 0.6259,
+      "step": 4091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9579637638465063e-05,
+      "loss": 0.6091,
+      "step": 4092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9579433110341997e-05,
+      "loss": 0.5467,
+      "step": 4093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957922853354303e-05,
+      "loss": 0.6226,
+      "step": 4094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9579023908069196e-05,
+      "loss": 0.6342,
+      "step": 4095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957881923392154e-05,
+      "loss": 0.6209,
+      "step": 4096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9578614511101094e-05,
+      "loss": 0.5767,
+      "step": 4097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9578409739608902e-05,
+      "loss": 0.6291,
+      "step": 4098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957820491944601e-05,
+      "loss": 0.5908,
+      "step": 4099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957800005061345e-05,
+      "loss": 0.6097,
+      "step": 4100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9577795133112268e-05,
+      "loss": 0.5795,
+      "step": 4101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9577590166943502e-05,
+      "loss": 0.6725,
+      "step": 4102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9577385152108197e-05,
+      "loss": 0.5967,
+      "step": 4103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9577180088607393e-05,
+      "loss": 0.6097,
+      "step": 4104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957697497644213e-05,
+      "loss": 0.6229,
+      "step": 4105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9576769815613453e-05,
+      "loss": 0.6276,
+      "step": 4106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9576564606122408e-05,
+      "loss": 0.6573,
+      "step": 4107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957635934797003e-05,
+      "loss": 0.6582,
+      "step": 4108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9576154041157366e-05,
+      "loss": 0.6004,
+      "step": 4109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957594868568546e-05,
+      "loss": 0.5707,
+      "step": 4110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9575743281555355e-05,
+      "loss": 0.5854,
+      "step": 4111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957553782876809e-05,
+      "loss": 0.5934,
+      "step": 4112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9575332327324718e-05,
+      "loss": 0.6207,
+      "step": 4113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9575126777226275e-05,
+      "loss": 0.6513,
+      "step": 4114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9574921178473808e-05,
+      "loss": 0.5969,
+      "step": 4115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9574715531068365e-05,
+      "loss": 0.6274,
+      "step": 4116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9574509835010987e-05,
+      "loss": 0.5971,
+      "step": 4117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957430409030272e-05,
+      "loss": 0.6367,
+      "step": 4118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957409829694461e-05,
+      "loss": 0.6882,
+      "step": 4119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9573892454937703e-05,
+      "loss": 0.6182,
+      "step": 4120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9573686564283047e-05,
+      "loss": 0.6547,
+      "step": 4121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957348062498168e-05,
+      "loss": 0.6238,
+      "step": 4122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9573274637034662e-05,
+      "loss": 0.578,
+      "step": 4123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9573068600443025e-05,
+      "loss": 0.6627,
+      "step": 4124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9572862515207827e-05,
+      "loss": 0.667,
+      "step": 4125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957265638133011e-05,
+      "loss": 0.6114,
+      "step": 4126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957245019881092e-05,
+      "loss": 0.6654,
+      "step": 4127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957224396765131e-05,
+      "loss": 0.7107,
+      "step": 4128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9572037687852322e-05,
+      "loss": 0.6617,
+      "step": 4129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957183135941501e-05,
+      "loss": 0.6458,
+      "step": 4130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957162498234042e-05,
+      "loss": 0.6594,
+      "step": 4131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9571418556629598e-05,
+      "loss": 0.6372,
+      "step": 4132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.95712120822836e-05,
+      "loss": 0.6007,
+      "step": 4133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9571005559303465e-05,
+      "loss": 0.6088,
+      "step": 4134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957079898769025e-05,
+      "loss": 0.6775,
+      "step": 4135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9570592367445e-05,
+      "loss": 0.6174,
+      "step": 4136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.957038569856877e-05,
+      "loss": 0.6073,
+      "step": 4137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9570178981062607e-05,
+      "loss": 0.5528,
+      "step": 4138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956997221492756e-05,
+      "loss": 0.6295,
+      "step": 4139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9569765400164684e-05,
+      "loss": 0.6166,
+      "step": 4140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9569558536775026e-05,
+      "loss": 0.6896,
+      "step": 4141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956935162475964e-05,
+      "loss": 0.6729,
+      "step": 4142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9569144664119576e-05,
+      "loss": 0.6346,
+      "step": 4143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9568937654855882e-05,
+      "loss": 0.668,
+      "step": 4144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9568730596969617e-05,
+      "loss": 0.6001,
+      "step": 4145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9568523490461825e-05,
+      "loss": 0.6008,
+      "step": 4146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956831633533357e-05,
+      "loss": 0.6061,
+      "step": 4147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9568109131585892e-05,
+      "loss": 0.6137,
+      "step": 4148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956790187921985e-05,
+      "loss": 0.6045,
+      "step": 4149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9567694578236497e-05,
+      "loss": 0.6711,
+      "step": 4150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9567487228636886e-05,
+      "loss": 0.6223,
+      "step": 4151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956727983042207e-05,
+      "loss": 0.6288,
+      "step": 4152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9567072383593105e-05,
+      "loss": 0.6171,
+      "step": 4153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956686488815104e-05,
+      "loss": 0.612,
+      "step": 4154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956665734409694e-05,
+      "loss": 0.6033,
+      "step": 4155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9566449751431845e-05,
+      "loss": 0.6151,
+      "step": 4156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9566242110156815e-05,
+      "loss": 0.6109,
+      "step": 4157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9566034420272915e-05,
+      "loss": 0.6146,
+      "step": 4158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956582668178119e-05,
+      "loss": 0.5917,
+      "step": 4159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9565618894682698e-05,
+      "loss": 0.6434,
+      "step": 4160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956541105897849e-05,
+      "loss": 0.6206,
+      "step": 4161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9565203174669634e-05,
+      "loss": 0.6608,
+      "step": 4162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9564995241757175e-05,
+      "loss": 0.6263,
+      "step": 4163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9564787260242176e-05,
+      "loss": 0.6662,
+      "step": 4164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956457923012569e-05,
+      "loss": 0.6699,
+      "step": 4165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956437115140878e-05,
+      "loss": 0.6384,
+      "step": 4166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9564163024092494e-05,
+      "loss": 0.6453,
+      "step": 4167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9563954848177897e-05,
+      "loss": 0.5768,
+      "step": 4168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9563746623666046e-05,
+      "loss": 0.6885,
+      "step": 4169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9563538350557996e-05,
+      "loss": 0.6074,
+      "step": 4170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956333002885481e-05,
+      "loss": 0.6599,
+      "step": 4171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956312165855754e-05,
+      "loss": 0.5578,
+      "step": 4172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9562913239667245e-05,
+      "loss": 0.5844,
+      "step": 4173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9562704772184992e-05,
+      "loss": 0.6715,
+      "step": 4174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956249625611183e-05,
+      "loss": 0.6041,
+      "step": 4175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956228769144883e-05,
+      "loss": 0.6321,
+      "step": 4176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9562079078197044e-05,
+      "loss": 0.6114,
+      "step": 4177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9561870416357535e-05,
+      "loss": 0.6418,
+      "step": 4178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956166170593136e-05,
+      "loss": 0.6739,
+      "step": 4179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9561452946919583e-05,
+      "loss": 0.6092,
+      "step": 4180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9561244139323264e-05,
+      "loss": 0.6055,
+      "step": 4181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956103528314346e-05,
+      "loss": 0.6061,
+      "step": 4182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9560826378381237e-05,
+      "loss": 0.6551,
+      "step": 4183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9560617425037657e-05,
+      "loss": 0.6211,
+      "step": 4184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.956040842311378e-05,
+      "loss": 0.5719,
+      "step": 4185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9560199372610668e-05,
+      "loss": 0.6602,
+      "step": 4186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.955999027352938e-05,
+      "loss": 0.6097,
+      "step": 4187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9559781125870983e-05,
+      "loss": 0.6054,
+      "step": 4188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9559571929636538e-05,
+      "loss": 0.5958,
+      "step": 4189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955936268482711e-05,
+      "loss": 0.6352,
+      "step": 4190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955915339144376e-05,
+      "loss": 0.6492,
+      "step": 4191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955894404948755e-05,
+      "loss": 0.6055,
+      "step": 4192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955873465895955e-05,
+      "loss": 0.6851,
+      "step": 4193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955852521986082e-05,
+      "loss": 0.5761,
+      "step": 4194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955831573219242e-05,
+      "loss": 0.6024,
+      "step": 4195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955810619595542e-05,
+      "loss": 0.6243,
+      "step": 4196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9557896611150884e-05,
+      "loss": 0.6865,
+      "step": 4197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9557686977779875e-05,
+      "loss": 0.5928,
+      "step": 4198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955747729584346e-05,
+      "loss": 0.613,
+      "step": 4199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9557267565342704e-05,
+      "loss": 0.5807,
+      "step": 4200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955705778627867e-05,
+      "loss": 0.6384,
+      "step": 4201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955684795865243e-05,
+      "loss": 0.6485,
+      "step": 4202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9556638082465046e-05,
+      "loss": 0.6329,
+      "step": 4203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9556428157717583e-05,
+      "loss": 0.6284,
+      "step": 4204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955621818441111e-05,
+      "loss": 0.5625,
+      "step": 4205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9556008162546696e-05,
+      "loss": 0.6093,
+      "step": 4206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9555798092125405e-05,
+      "loss": 0.6061,
+      "step": 4207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9555587973148307e-05,
+      "loss": 0.6447,
+      "step": 4208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9555377805616464e-05,
+      "loss": 0.6102,
+      "step": 4209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955516758953095e-05,
+      "loss": 0.6483,
+      "step": 4210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955495732489283e-05,
+      "loss": 0.6295,
+      "step": 4211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9554747011703176e-05,
+      "loss": 0.6029,
+      "step": 4212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9554536649963054e-05,
+      "loss": 0.601,
+      "step": 4213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955432623967353e-05,
+      "loss": 0.6305,
+      "step": 4214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955411578083568e-05,
+      "loss": 0.5827,
+      "step": 4215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955390527345057e-05,
+      "loss": 0.6698,
+      "step": 4216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9553694717519267e-05,
+      "loss": 0.6081,
+      "step": 4217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9553484113042843e-05,
+      "loss": 0.6395,
+      "step": 4218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9553273460022368e-05,
+      "loss": 0.6285,
+      "step": 4219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9553062758458916e-05,
+      "loss": 0.6783,
+      "step": 4220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955285200835355e-05,
+      "loss": 0.646,
+      "step": 4221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955264120970735e-05,
+      "loss": 0.5942,
+      "step": 4222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955243036252138e-05,
+      "loss": 0.5983,
+      "step": 4223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955221946679671e-05,
+      "loss": 0.597,
+      "step": 4224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955200852253442e-05,
+      "loss": 0.6769,
+      "step": 4225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9551797529735577e-05,
+      "loss": 0.5797,
+      "step": 4226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9551586488401253e-05,
+      "loss": 0.6338,
+      "step": 4227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955137539853252e-05,
+      "loss": 0.6275,
+      "step": 4228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955116426013045e-05,
+      "loss": 0.5941,
+      "step": 4229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955095307319612e-05,
+      "loss": 0.6106,
+      "step": 4230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.95507418377306e-05,
+      "loss": 0.6162,
+      "step": 4231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9550530553734962e-05,
+      "loss": 0.719,
+      "step": 4232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.955031922121028e-05,
+      "loss": 0.6343,
+      "step": 4233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9550107840157632e-05,
+      "loss": 0.6659,
+      "step": 4234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954989641057809e-05,
+      "loss": 0.6793,
+      "step": 4235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9549684932472727e-05,
+      "loss": 0.6162,
+      "step": 4236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9549473405842615e-05,
+      "loss": 0.6239,
+      "step": 4237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9549261830688833e-05,
+      "loss": 0.6276,
+      "step": 4238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9549050207012456e-05,
+      "loss": 0.6354,
+      "step": 4239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954883853481456e-05,
+      "loss": 0.6148,
+      "step": 4240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9548626814096216e-05,
+      "loss": 0.6562,
+      "step": 4241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9548415044858504e-05,
+      "loss": 0.7051,
+      "step": 4242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.95482032271025e-05,
+      "loss": 0.6302,
+      "step": 4243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9547991360829278e-05,
+      "loss": 0.5899,
+      "step": 4244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9547779446039915e-05,
+      "loss": 0.6056,
+      "step": 4245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9547567482735487e-05,
+      "loss": 0.5936,
+      "step": 4246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9547355470917077e-05,
+      "loss": 0.64,
+      "step": 4247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9547143410585752e-05,
+      "loss": 0.6372,
+      "step": 4248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9546931301742598e-05,
+      "loss": 0.6102,
+      "step": 4249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954671914438869e-05,
+      "loss": 0.5968,
+      "step": 4250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9546506938525106e-05,
+      "loss": 0.6451,
+      "step": 4251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9546294684152926e-05,
+      "loss": 0.6442,
+      "step": 4252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954608238127323e-05,
+      "loss": 0.6314,
+      "step": 4253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9545870029887087e-05,
+      "loss": 0.6036,
+      "step": 4254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9545657629995585e-05,
+      "loss": 0.5738,
+      "step": 4255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9545445181599802e-05,
+      "loss": 0.6148,
+      "step": 4256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9545232684700816e-05,
+      "loss": 0.6259,
+      "step": 4257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954502013929971e-05,
+      "loss": 0.6686,
+      "step": 4258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9544807545397554e-05,
+      "loss": 0.6563,
+      "step": 4259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954459490299544e-05,
+      "loss": 0.6438,
+      "step": 4260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9544382212094447e-05,
+      "loss": 0.5506,
+      "step": 4261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9544169472695647e-05,
+      "loss": 0.6109,
+      "step": 4262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954395668480013e-05,
+      "loss": 0.5361,
+      "step": 4263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9543743848408974e-05,
+      "loss": 0.6161,
+      "step": 4264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954353096352326e-05,
+      "loss": 0.5974,
+      "step": 4265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9543318030144072e-05,
+      "loss": 0.5981,
+      "step": 4266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954310504827249e-05,
+      "loss": 0.6234,
+      "step": 4267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9542892017909594e-05,
+      "loss": 0.612,
+      "step": 4268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954267893905647e-05,
+      "loss": 0.5891,
+      "step": 4269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.95424658117142e-05,
+      "loss": 0.5882,
+      "step": 4270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9542252635883866e-05,
+      "loss": 0.6133,
+      "step": 4271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954203941156655e-05,
+      "loss": 0.5979,
+      "step": 4272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9541826138763337e-05,
+      "loss": 0.6493,
+      "step": 4273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9541612817475315e-05,
+      "loss": 0.598,
+      "step": 4274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9541399447703563e-05,
+      "loss": 0.6036,
+      "step": 4275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9541186029449164e-05,
+      "loss": 0.6348,
+      "step": 4276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9540972562713204e-05,
+      "loss": 0.6327,
+      "step": 4277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.954075904749677e-05,
+      "loss": 0.6244,
+      "step": 4278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9540545483800947e-05,
+      "loss": 0.6333,
+      "step": 4279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9540331871626815e-05,
+      "loss": 0.6332,
+      "step": 4280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9540118210975465e-05,
+      "loss": 0.6163,
+      "step": 4281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9539904501847977e-05,
+      "loss": 0.5936,
+      "step": 4282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9539690744245443e-05,
+      "loss": 0.623,
+      "step": 4283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9539476938168945e-05,
+      "loss": 0.6488,
+      "step": 4284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9539263083619574e-05,
+      "loss": 0.5713,
+      "step": 4285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9539049180598413e-05,
+      "loss": 0.5589,
+      "step": 4286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9538835229106548e-05,
+      "loss": 0.6204,
+      "step": 4287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9538621229145066e-05,
+      "loss": 0.6047,
+      "step": 4288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953840718071506e-05,
+      "loss": 0.6387,
+      "step": 4289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953819308381761e-05,
+      "loss": 0.6113,
+      "step": 4290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9537978938453808e-05,
+      "loss": 0.6396,
+      "step": 4291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9537764744624742e-05,
+      "loss": 0.5872,
+      "step": 4292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9537550502331504e-05,
+      "loss": 0.5942,
+      "step": 4293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9537336211575175e-05,
+      "loss": 0.6478,
+      "step": 4294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9537121872356848e-05,
+      "loss": 0.605,
+      "step": 4295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953690748467761e-05,
+      "loss": 0.662,
+      "step": 4296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9536693048538552e-05,
+      "loss": 0.5735,
+      "step": 4297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953647856394077e-05,
+      "loss": 0.6148,
+      "step": 4298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953626403088534e-05,
+      "loss": 0.6442,
+      "step": 4299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9536049449373365e-05,
+      "loss": 0.6809,
+      "step": 4300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9535834819405924e-05,
+      "loss": 0.652,
+      "step": 4301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953562014098412e-05,
+      "loss": 0.5757,
+      "step": 4302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953540541410903e-05,
+      "loss": 0.589,
+      "step": 4303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953519063878176e-05,
+      "loss": 0.6607,
+      "step": 4304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9534975815003393e-05,
+      "loss": 0.6267,
+      "step": 4305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9534760942775018e-05,
+      "loss": 0.5788,
+      "step": 4306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953454602209773e-05,
+      "loss": 0.6318,
+      "step": 4307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9534331052972622e-05,
+      "loss": 0.6422,
+      "step": 4308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953411603540079e-05,
+      "loss": 0.5608,
+      "step": 4309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953390096938332e-05,
+      "loss": 0.6114,
+      "step": 4310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9533685854921302e-05,
+      "loss": 0.6044,
+      "step": 4311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953347069201584e-05,
+      "loss": 0.6937,
+      "step": 4312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9533255480668018e-05,
+      "loss": 0.6459,
+      "step": 4313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9533040220878935e-05,
+      "loss": 0.6148,
+      "step": 4314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9532824912649682e-05,
+      "loss": 0.645,
+      "step": 4315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9532609555981355e-05,
+      "loss": 0.5929,
+      "step": 4316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9532394150875043e-05,
+      "loss": 0.5547,
+      "step": 4317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9532178697331853e-05,
+      "loss": 0.5926,
+      "step": 4318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9531963195352863e-05,
+      "loss": 0.6146,
+      "step": 4319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953174764493918e-05,
+      "loss": 0.6486,
+      "step": 4320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9531532046091896e-05,
+      "loss": 0.6276,
+      "step": 4321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9531316398812105e-05,
+      "loss": 0.6314,
+      "step": 4322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9531100703100904e-05,
+      "loss": 0.607,
+      "step": 4323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.953088495895939e-05,
+      "loss": 0.6348,
+      "step": 4324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9530669166388656e-05,
+      "loss": 0.6055,
+      "step": 4325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9530453325389804e-05,
+      "loss": 0.5918,
+      "step": 4326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9530237435963927e-05,
+      "loss": 0.6108,
+      "step": 4327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9530021498112123e-05,
+      "loss": 0.6177,
+      "step": 4328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9529805511835485e-05,
+      "loss": 0.5521,
+      "step": 4329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9529589477135116e-05,
+      "loss": 0.5999,
+      "step": 4330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9529373394012113e-05,
+      "loss": 0.623,
+      "step": 4331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.952915726246757e-05,
+      "loss": 0.6374,
+      "step": 4332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9528941082502592e-05,
+      "loss": 0.6026,
+      "step": 4333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9528724854118273e-05,
+      "loss": 0.6325,
+      "step": 4334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952850857731571e-05,
+      "loss": 0.6274,
+      "step": 4335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9528292252096008e-05,
+      "loss": 0.6004,
+      "step": 4336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9528075878460264e-05,
+      "loss": 0.6133,
+      "step": 4337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952785945640957e-05,
+      "loss": 0.5565,
+      "step": 4338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9527642985945037e-05,
+      "loss": 0.6058,
+      "step": 4339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9527426467067757e-05,
+      "loss": 0.5522,
+      "step": 4340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952720989977883e-05,
+      "loss": 0.5682,
+      "step": 4341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9526993284079365e-05,
+      "loss": 0.5907,
+      "step": 4342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9526776619970453e-05,
+      "loss": 0.5962,
+      "step": 4343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95265599074532e-05,
+      "loss": 0.5564,
+      "step": 4344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9526343146528708e-05,
+      "loss": 0.6562,
+      "step": 4345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9526126337198074e-05,
+      "loss": 0.6056,
+      "step": 4346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9525909479462404e-05,
+      "loss": 0.5959,
+      "step": 4347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9525692573322795e-05,
+      "loss": 0.6179,
+      "step": 4348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952547561878036e-05,
+      "loss": 0.638,
+      "step": 4349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9525258615836185e-05,
+      "loss": 0.6139,
+      "step": 4350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9525041564491385e-05,
+      "loss": 0.6697,
+      "step": 4351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952482446474706e-05,
+      "loss": 0.605,
+      "step": 4352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9524607316604312e-05,
+      "loss": 0.6593,
+      "step": 4353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9524390120064245e-05,
+      "loss": 0.6204,
+      "step": 4354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952417287512796e-05,
+      "loss": 0.5628,
+      "step": 4355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9523955581796565e-05,
+      "loss": 0.5532,
+      "step": 4356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9523738240071165e-05,
+      "loss": 0.6174,
+      "step": 4357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9523520849952858e-05,
+      "loss": 0.6706,
+      "step": 4358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9523303411442756e-05,
+      "loss": 0.6163,
+      "step": 4359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9523085924541956e-05,
+      "loss": 0.5993,
+      "step": 4360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952286838925157e-05,
+      "loss": 0.6197,
+      "step": 4361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95226508055727e-05,
+      "loss": 0.6539,
+      "step": 4362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9522433173506454e-05,
+      "loss": 0.68,
+      "step": 4363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9522215493053937e-05,
+      "loss": 0.6349,
+      "step": 4364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952199776421625e-05,
+      "loss": 0.637,
+      "step": 4365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9521779986994506e-05,
+      "loss": 0.5958,
+      "step": 4366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952156216138981e-05,
+      "loss": 0.5842,
+      "step": 4367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952134428740327e-05,
+      "loss": 0.5264,
+      "step": 4368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9521126365035987e-05,
+      "loss": 0.6415,
+      "step": 4369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9520908394289077e-05,
+      "loss": 0.5655,
+      "step": 4370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952069037516364e-05,
+      "loss": 0.5802,
+      "step": 4371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.952047230766079e-05,
+      "loss": 0.6382,
+      "step": 4372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9520254191781623e-05,
+      "loss": 0.5782,
+      "step": 4373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9520036027527264e-05,
+      "loss": 0.5845,
+      "step": 4374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9519817814898813e-05,
+      "loss": 0.6321,
+      "step": 4375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951959955389738e-05,
+      "loss": 0.6547,
+      "step": 4376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9519381244524073e-05,
+      "loss": 0.5943,
+      "step": 4377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9519162886780006e-05,
+      "loss": 0.6267,
+      "step": 4378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951894448066628e-05,
+      "loss": 0.6602,
+      "step": 4379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951872602618401e-05,
+      "loss": 0.5999,
+      "step": 4380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9518507523334304e-05,
+      "loss": 0.593,
+      "step": 4381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9518288972118276e-05,
+      "loss": 0.6508,
+      "step": 4382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9518070372537033e-05,
+      "loss": 0.6121,
+      "step": 4383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9517851724591686e-05,
+      "loss": 0.6068,
+      "step": 4384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951763302828335e-05,
+      "loss": 0.6224,
+      "step": 4385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9517414283613128e-05,
+      "loss": 0.5965,
+      "step": 4386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9517195490582143e-05,
+      "loss": 0.6351,
+      "step": 4387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95169766491915e-05,
+      "loss": 0.5865,
+      "step": 4388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9516757759442305e-05,
+      "loss": 0.6152,
+      "step": 4389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951653882133568e-05,
+      "loss": 0.5787,
+      "step": 4390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951631983487274e-05,
+      "loss": 0.6353,
+      "step": 4391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9516100800054583e-05,
+      "loss": 0.6159,
+      "step": 4392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9515881716882336e-05,
+      "loss": 0.677,
+      "step": 4393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9515662585357103e-05,
+      "loss": 0.6062,
+      "step": 4394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951544340548001e-05,
+      "loss": 0.6249,
+      "step": 4395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9515224177252156e-05,
+      "loss": 0.6388,
+      "step": 4396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9515004900674663e-05,
+      "loss": 0.6185,
+      "step": 4397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951478557574864e-05,
+      "loss": 0.5538,
+      "step": 4398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9514566202475208e-05,
+      "loss": 0.6156,
+      "step": 4399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951434678085548e-05,
+      "loss": 0.597,
+      "step": 4400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9514127310890566e-05,
+      "loss": 0.6339,
+      "step": 4401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951390779258159e-05,
+      "loss": 0.5849,
+      "step": 4402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9513688225929657e-05,
+      "loss": 0.5661,
+      "step": 4403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951346861093589e-05,
+      "loss": 0.6332,
+      "step": 4404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9513248947601402e-05,
+      "loss": 0.6007,
+      "step": 4405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951302923592731e-05,
+      "loss": 0.6422,
+      "step": 4406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951280947591473e-05,
+      "loss": 0.6356,
+      "step": 4407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9512589667564782e-05,
+      "loss": 0.6564,
+      "step": 4408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951236981087858e-05,
+      "loss": 0.625,
+      "step": 4409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9512149905857237e-05,
+      "loss": 0.6162,
+      "step": 4410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951192995250188e-05,
+      "loss": 0.6044,
+      "step": 4411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951170995081362e-05,
+      "loss": 0.6038,
+      "step": 4412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951148990079357e-05,
+      "loss": 0.5706,
+      "step": 4413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951126980244286e-05,
+      "loss": 0.6036,
+      "step": 4414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95110496557626e-05,
+      "loss": 0.6328,
+      "step": 4415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.951082946075391e-05,
+      "loss": 0.6546,
+      "step": 4416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9510609217417915e-05,
+      "loss": 0.5904,
+      "step": 4417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9510388925755728e-05,
+      "loss": 0.6627,
+      "step": 4418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9510168585768468e-05,
+      "loss": 0.6432,
+      "step": 4419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9509948197457257e-05,
+      "loss": 0.6621,
+      "step": 4420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9509727760823212e-05,
+      "loss": 0.6141,
+      "step": 4421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950950727586746e-05,
+      "loss": 0.606,
+      "step": 4422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9509286742591113e-05,
+      "loss": 0.5724,
+      "step": 4423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9509066160995294e-05,
+      "loss": 0.6001,
+      "step": 4424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9508845531081125e-05,
+      "loss": 0.638,
+      "step": 4425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950862485284973e-05,
+      "loss": 0.6002,
+      "step": 4426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9508404126302227e-05,
+      "loss": 0.6099,
+      "step": 4427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9508183351439737e-05,
+      "loss": 0.6287,
+      "step": 4428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9507962528263383e-05,
+      "loss": 0.6098,
+      "step": 4429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950774165677429e-05,
+      "loss": 0.6403,
+      "step": 4430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950752073697357e-05,
+      "loss": 0.608,
+      "step": 4431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9507299768862356e-05,
+      "loss": 0.5976,
+      "step": 4432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9507078752441767e-05,
+      "loss": 0.5972,
+      "step": 4433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950685768771293e-05,
+      "loss": 0.6241,
+      "step": 4434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9506636574676963e-05,
+      "loss": 0.6001,
+      "step": 4435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9506415413334993e-05,
+      "loss": 0.6464,
+      "step": 4436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9506194203688137e-05,
+      "loss": 0.6509,
+      "step": 4437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9505972945737527e-05,
+      "loss": 0.5996,
+      "step": 4438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9505751639484286e-05,
+      "loss": 0.6079,
+      "step": 4439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9505530284929537e-05,
+      "loss": 0.6,
+      "step": 4440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9505308882074402e-05,
+      "loss": 0.5759,
+      "step": 4441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950508743092001e-05,
+      "loss": 0.6143,
+      "step": 4442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9504865931467486e-05,
+      "loss": 0.583,
+      "step": 4443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9504644383717956e-05,
+      "loss": 0.6001,
+      "step": 4444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9504422787672542e-05,
+      "loss": 0.6001,
+      "step": 4445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9504201143332372e-05,
+      "loss": 0.6398,
+      "step": 4446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9503979450698575e-05,
+      "loss": 0.5983,
+      "step": 4447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950375770977227e-05,
+      "loss": 0.6175,
+      "step": 4448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9503535920554597e-05,
+      "loss": 0.5976,
+      "step": 4449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950331408304667e-05,
+      "loss": 0.6297,
+      "step": 4450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9503092197249617e-05,
+      "loss": 0.6243,
+      "step": 4451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9502870263164575e-05,
+      "loss": 0.5794,
+      "step": 4452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9502648280792665e-05,
+      "loss": 0.569,
+      "step": 4453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9502426250135015e-05,
+      "loss": 0.6178,
+      "step": 4454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9502204171192757e-05,
+      "loss": 0.6355,
+      "step": 4455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9501982043967015e-05,
+      "loss": 0.704,
+      "step": 4456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9501759868458918e-05,
+      "loss": 0.5569,
+      "step": 4457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95015376446696e-05,
+      "loss": 0.6095,
+      "step": 4458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950131537260018e-05,
+      "loss": 0.5815,
+      "step": 4459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.95010930522518e-05,
+      "loss": 0.6753,
+      "step": 4460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.950087068362558e-05,
+      "loss": 0.591,
+      "step": 4461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9500648266722658e-05,
+      "loss": 0.6306,
+      "step": 4462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9500425801544156e-05,
+      "loss": 0.6236,
+      "step": 4463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9500203288091208e-05,
+      "loss": 0.6037,
+      "step": 4464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9499980726364947e-05,
+      "loss": 0.5638,
+      "step": 4465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9499758116366502e-05,
+      "loss": 0.5986,
+      "step": 4466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9499535458097005e-05,
+      "loss": 0.6756,
+      "step": 4467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9499312751557583e-05,
+      "loss": 0.6331,
+      "step": 4468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.949908999674937e-05,
+      "loss": 0.6372,
+      "step": 4469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.94988671936735e-05,
+      "loss": 0.6002,
+      "step": 4470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9498644342331107e-05,
+      "loss": 0.6466,
+      "step": 4471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9498421442723315e-05,
+      "loss": 0.639,
+      "step": 4472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9498198494851264e-05,
+      "loss": 0.5633,
+      "step": 4473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.949797549871609e-05,
+      "loss": 0.5614,
+      "step": 4474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9497752454318915e-05,
+      "loss": 0.6151,
+      "step": 4475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.949752936166088e-05,
+      "loss": 0.622,
+      "step": 4476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9497306220743115e-05,
+      "loss": 0.6116,
+      "step": 4477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.949708303156676e-05,
+      "loss": 0.668,
+      "step": 4478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9496859794132942e-05,
+      "loss": 0.6087,
+      "step": 4479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9496636508442796e-05,
+      "loss": 0.5959,
+      "step": 4480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.949641317449746e-05,
+      "loss": 0.621,
+      "step": 4481
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949618979229807e-05,
+      "loss": 0.5625,
+      "step": 4482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9495966361845758e-05,
+      "loss": 0.6118,
+      "step": 4483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949574288314166e-05,
+      "loss": 0.5972,
+      "step": 4484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9495519356186907e-05,
+      "loss": 0.5959,
+      "step": 4485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9495295780982645e-05,
+      "loss": 0.649,
+      "step": 4486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949507215753e-05,
+      "loss": 0.607,
+      "step": 4487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9494848485830116e-05,
+      "loss": 0.6193,
+      "step": 4488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9494624765884125e-05,
+      "loss": 0.5806,
+      "step": 4489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949440099769316e-05,
+      "loss": 0.6252,
+      "step": 4490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949417718125837e-05,
+      "loss": 0.5846,
+      "step": 4491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9493953316580883e-05,
+      "loss": 0.658,
+      "step": 4492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9493729403661837e-05,
+      "loss": 0.635,
+      "step": 4493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9493505442502374e-05,
+      "loss": 0.6297,
+      "step": 4494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949328143310363e-05,
+      "loss": 0.6085,
+      "step": 4495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949305737546674e-05,
+      "loss": 0.5836,
+      "step": 4496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9492833269592845e-05,
+      "loss": 0.5811,
+      "step": 4497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9492609115483088e-05,
+      "loss": 0.6078,
+      "step": 4498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9492384913138603e-05,
+      "loss": 0.5807,
+      "step": 4499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9492160662560525e-05,
+      "loss": 0.588,
+      "step": 4500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9491936363750006e-05,
+      "loss": 0.5888,
+      "step": 4501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9491712016708177e-05,
+      "loss": 0.6564,
+      "step": 4502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9491487621436177e-05,
+      "loss": 0.6096,
+      "step": 4503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949126317793515e-05,
+      "loss": 0.6223,
+      "step": 4504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9491038686206233e-05,
+      "loss": 0.6228,
+      "step": 4505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949081414625057e-05,
+      "loss": 0.6461,
+      "step": 4506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9490589558069303e-05,
+      "loss": 0.6023,
+      "step": 4507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9490364921663567e-05,
+      "loss": 0.6082,
+      "step": 4508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.949014023703451e-05,
+      "loss": 0.6262,
+      "step": 4509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948991550418327e-05,
+      "loss": 0.72,
+      "step": 4510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9489690723110994e-05,
+      "loss": 0.6141,
+      "step": 4511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9489465893818817e-05,
+      "loss": 0.5839,
+      "step": 4512
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9489241016307883e-05,
+      "loss": 0.5273,
+      "step": 4513
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9489016090579338e-05,
+      "loss": 0.6815,
+      "step": 4514
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9488791116634324e-05,
+      "loss": 0.6812,
+      "step": 4515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948856609447398e-05,
+      "loss": 0.6254,
+      "step": 4516
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9488341024099456e-05,
+      "loss": 0.5977,
+      "step": 4517
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9488115905511894e-05,
+      "loss": 0.6509,
+      "step": 4518
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9487890738712435e-05,
+      "loss": 0.5938,
+      "step": 4519
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9487665523702224e-05,
+      "loss": 0.6345,
+      "step": 4520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9487440260482406e-05,
+      "loss": 0.6211,
+      "step": 4521
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9487214949054126e-05,
+      "loss": 0.6416,
+      "step": 4522
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9486989589418527e-05,
+      "loss": 0.6336,
+      "step": 4523
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9486764181576755e-05,
+      "loss": 0.6184,
+      "step": 4524
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948653872552996e-05,
+      "loss": 0.6297,
+      "step": 4525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948631322127928e-05,
+      "loss": 0.59,
+      "step": 4526
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9486087668825866e-05,
+      "loss": 0.6559,
+      "step": 4527
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948586206817086e-05,
+      "loss": 0.6285,
+      "step": 4528
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948563641931541e-05,
+      "loss": 0.5869,
+      "step": 4529
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9485410722260666e-05,
+      "loss": 0.5968,
+      "step": 4530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9485184977007775e-05,
+      "loss": 0.5952,
+      "step": 4531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9484959183557876e-05,
+      "loss": 0.6044,
+      "step": 4532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9484733341912124e-05,
+      "loss": 0.5557,
+      "step": 4533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9484507452071664e-05,
+      "loss": 0.5698,
+      "step": 4534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9484281514037642e-05,
+      "loss": 0.6665,
+      "step": 4535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9484055527811212e-05,
+      "loss": 0.629,
+      "step": 4536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9483829493393514e-05,
+      "loss": 0.6306,
+      "step": 4537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9483603410785706e-05,
+      "loss": 0.6366,
+      "step": 4538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9483377279988927e-05,
+      "loss": 0.6569,
+      "step": 4539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948315110100433e-05,
+      "loss": 0.6086,
+      "step": 4540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9482924873833068e-05,
+      "loss": 0.6283,
+      "step": 4541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948269859847629e-05,
+      "loss": 0.6994,
+      "step": 4542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9482472274935136e-05,
+      "loss": 0.5938,
+      "step": 4543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948224590321077e-05,
+      "loss": 0.6772,
+      "step": 4544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948201948330433e-05,
+      "loss": 0.5963,
+      "step": 4545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9481793015216975e-05,
+      "loss": 0.7028,
+      "step": 4546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9481566498949857e-05,
+      "loss": 0.6088,
+      "step": 4547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9481339934504118e-05,
+      "loss": 0.5656,
+      "step": 4548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9481113321880913e-05,
+      "loss": 0.5767,
+      "step": 4549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.94808866610814e-05,
+      "loss": 0.6401,
+      "step": 4550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948065995210672e-05,
+      "loss": 0.66,
+      "step": 4551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.948043319495803e-05,
+      "loss": 0.5567,
+      "step": 4552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9480206389636484e-05,
+      "loss": 0.6352,
+      "step": 4553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9479979536143232e-05,
+      "loss": 0.6307,
+      "step": 4554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947975263447943e-05,
+      "loss": 0.6562,
+      "step": 4555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947952568464623e-05,
+      "loss": 0.6141,
+      "step": 4556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9479298686644778e-05,
+      "loss": 0.5826,
+      "step": 4557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9479071640476237e-05,
+      "loss": 0.6064,
+      "step": 4558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9478844546141756e-05,
+      "loss": 0.5528,
+      "step": 4559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947861740364249e-05,
+      "loss": 0.5954,
+      "step": 4560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9478390212979593e-05,
+      "loss": 0.6185,
+      "step": 4561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947816297415422e-05,
+      "loss": 0.6547,
+      "step": 4562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9477935687167524e-05,
+      "loss": 0.6582,
+      "step": 4563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9477708352020662e-05,
+      "loss": 0.6071,
+      "step": 4564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947748096871479e-05,
+      "loss": 0.5855,
+      "step": 4565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947725353725106e-05,
+      "loss": 0.5868,
+      "step": 4566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9477026057630625e-05,
+      "loss": 0.6093,
+      "step": 4567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947679852985465e-05,
+      "loss": 0.6218,
+      "step": 4568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9476570953924287e-05,
+      "loss": 0.5683,
+      "step": 4569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9476343329840687e-05,
+      "loss": 0.6596,
+      "step": 4570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9476115657605015e-05,
+      "loss": 0.5952,
+      "step": 4571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9475887937218424e-05,
+      "loss": 0.6684,
+      "step": 4572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9475660168682073e-05,
+      "loss": 0.6197,
+      "step": 4573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9475432351997115e-05,
+      "loss": 0.6517,
+      "step": 4574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9475204487164708e-05,
+      "loss": 0.5966,
+      "step": 4575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9474976574186013e-05,
+      "loss": 0.6458,
+      "step": 4576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9474748613062193e-05,
+      "loss": 0.6654,
+      "step": 4577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9474520603794395e-05,
+      "loss": 0.5593,
+      "step": 4578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9474292546383782e-05,
+      "loss": 0.6299,
+      "step": 4579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9474064440831516e-05,
+      "loss": 0.6373,
+      "step": 4580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9473836287138753e-05,
+      "loss": 0.6496,
+      "step": 4581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9473608085306655e-05,
+      "loss": 0.6312,
+      "step": 4582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947337983533638e-05,
+      "loss": 0.6497,
+      "step": 4583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9473151537229086e-05,
+      "loss": 0.5727,
+      "step": 4584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9472923190985934e-05,
+      "loss": 0.6407,
+      "step": 4585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9472694796608088e-05,
+      "loss": 0.6471,
+      "step": 4586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9472466354096703e-05,
+      "loss": 0.6492,
+      "step": 4587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9472237863452946e-05,
+      "loss": 0.6385,
+      "step": 4588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9472009324677968e-05,
+      "loss": 0.6537,
+      "step": 4589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947178073777294e-05,
+      "loss": 0.6154,
+      "step": 4590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9471552102739022e-05,
+      "loss": 0.6179,
+      "step": 4591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947132341957737e-05,
+      "loss": 0.5681,
+      "step": 4592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9471094688289155e-05,
+      "loss": 0.6346,
+      "step": 4593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947086590887553e-05,
+      "loss": 0.6694,
+      "step": 4594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9470637081337664e-05,
+      "loss": 0.6142,
+      "step": 4595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9470408205676717e-05,
+      "loss": 0.6244,
+      "step": 4596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.947017928189385e-05,
+      "loss": 0.6329,
+      "step": 4597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946995030999023e-05,
+      "loss": 0.6023,
+      "step": 4598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946972128996702e-05,
+      "loss": 0.6523,
+      "step": 4599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9469492221825385e-05,
+      "loss": 0.6775,
+      "step": 4600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946926310556648e-05,
+      "loss": 0.6574,
+      "step": 4601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946903394119148e-05,
+      "loss": 0.5733,
+      "step": 4602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9468804728701543e-05,
+      "loss": 0.6058,
+      "step": 4603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946857546809784e-05,
+      "loss": 0.6117,
+      "step": 4604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946834615938153e-05,
+      "loss": 0.6254,
+      "step": 4605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946811680255378e-05,
+      "loss": 0.6288,
+      "step": 4606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946788739761576e-05,
+      "loss": 0.6312,
+      "step": 4607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9467657944568627e-05,
+      "loss": 0.6374,
+      "step": 4608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946742844341355e-05,
+      "loss": 0.673,
+      "step": 4609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.94671988941517e-05,
+      "loss": 0.653,
+      "step": 4610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9466969296784238e-05,
+      "loss": 0.612,
+      "step": 4611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9466739651312332e-05,
+      "loss": 0.6164,
+      "step": 4612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946650995773715e-05,
+      "loss": 0.6188,
+      "step": 4613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946628021605986e-05,
+      "loss": 0.6645,
+      "step": 4614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9466050426281623e-05,
+      "loss": 0.651,
+      "step": 4615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9465820588403618e-05,
+      "loss": 0.6664,
+      "step": 4616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9465590702427003e-05,
+      "loss": 0.5768,
+      "step": 4617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9465360768352947e-05,
+      "loss": 0.6299,
+      "step": 4618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9465130786182625e-05,
+      "loss": 0.5724,
+      "step": 4619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.94649007559172e-05,
+      "loss": 0.5906,
+      "step": 4620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9464670677557844e-05,
+      "loss": 0.5534,
+      "step": 4621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9464440551105723e-05,
+      "loss": 0.5306,
+      "step": 4622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9464210376562005e-05,
+      "loss": 0.65,
+      "step": 4623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9463980153927866e-05,
+      "loss": 0.6756,
+      "step": 4624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946374988320447e-05,
+      "loss": 0.6153,
+      "step": 4625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.946351956439299e-05,
+      "loss": 0.6115,
+      "step": 4626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9463289197494595e-05,
+      "loss": 0.6287,
+      "step": 4627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9463058782510458e-05,
+      "loss": 0.7113,
+      "step": 4628
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9462828319441748e-05,
+      "loss": 0.5799,
+      "step": 4629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9462597808289637e-05,
+      "loss": 0.578,
+      "step": 4630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9462367249055295e-05,
+      "loss": 0.55,
+      "step": 4631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.946213664173989e-05,
+      "loss": 0.6186,
+      "step": 4632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94619059863446e-05,
+      "loss": 0.6281,
+      "step": 4633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9461675282870595e-05,
+      "loss": 0.5897,
+      "step": 4634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9461444531319045e-05,
+      "loss": 0.582,
+      "step": 4635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9461213731691127e-05,
+      "loss": 0.622,
+      "step": 4636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9460982883988012e-05,
+      "loss": 0.5991,
+      "step": 4637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9460751988210867e-05,
+      "loss": 0.6301,
+      "step": 4638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9460521044360872e-05,
+      "loss": 0.674,
+      "step": 4639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9460290052439203e-05,
+      "loss": 0.5731,
+      "step": 4640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9460059012447025e-05,
+      "loss": 0.6233,
+      "step": 4641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945982792438552e-05,
+      "loss": 0.6899,
+      "step": 4642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945959678825586e-05,
+      "loss": 0.6303,
+      "step": 4643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9459365604059215e-05,
+      "loss": 0.5424,
+      "step": 4644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9459134371796765e-05,
+      "loss": 0.6716,
+      "step": 4645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945890309146968e-05,
+      "loss": 0.6129,
+      "step": 4646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945867176307914e-05,
+      "loss": 0.5783,
+      "step": 4647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945844038662632e-05,
+      "loss": 0.5716,
+      "step": 4648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9458208962112393e-05,
+      "loss": 0.633,
+      "step": 4649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9457977489538537e-05,
+      "loss": 0.5862,
+      "step": 4650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9457745968905926e-05,
+      "loss": 0.6285,
+      "step": 4651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945751440021574e-05,
+      "loss": 0.6246,
+      "step": 4652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9457282783469154e-05,
+      "loss": 0.5859,
+      "step": 4653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9457051118667344e-05,
+      "loss": 0.5913,
+      "step": 4654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9456819405811488e-05,
+      "loss": 0.6523,
+      "step": 4655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945658764490276e-05,
+      "loss": 0.5723,
+      "step": 4656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945635583594234e-05,
+      "loss": 0.5138,
+      "step": 4657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9456123978931408e-05,
+      "loss": 0.6359,
+      "step": 4658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9455892073871142e-05,
+      "loss": 0.676,
+      "step": 4659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9455660120762722e-05,
+      "loss": 0.6149,
+      "step": 4660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945542811960732e-05,
+      "loss": 0.6128,
+      "step": 4661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945519607040612e-05,
+      "loss": 0.5547,
+      "step": 4662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9454963973160298e-05,
+      "loss": 0.6321,
+      "step": 4663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9454731827871036e-05,
+      "loss": 0.5995,
+      "step": 4664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9454499634539515e-05,
+      "loss": 0.5851,
+      "step": 4665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9454267393166908e-05,
+      "loss": 0.6078,
+      "step": 4666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94540351037544e-05,
+      "loss": 0.6149,
+      "step": 4667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9453802766303174e-05,
+      "loss": 0.6741,
+      "step": 4668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9453570380814406e-05,
+      "loss": 0.5903,
+      "step": 4669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9453337947289278e-05,
+      "loss": 0.5985,
+      "step": 4670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945310546572897e-05,
+      "loss": 0.6222,
+      "step": 4671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945287293613467e-05,
+      "loss": 0.6261,
+      "step": 4672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945264035850755e-05,
+      "loss": 0.593,
+      "step": 4673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9452407732848796e-05,
+      "loss": 0.6121,
+      "step": 4674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945217505915959e-05,
+      "loss": 0.6174,
+      "step": 4675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9451942337441115e-05,
+      "loss": 0.6269,
+      "step": 4676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945170956769455e-05,
+      "loss": 0.5556,
+      "step": 4677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9451476749921087e-05,
+      "loss": 0.5766,
+      "step": 4678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9451243884121896e-05,
+      "loss": 0.6546,
+      "step": 4679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945101097029817e-05,
+      "loss": 0.6279,
+      "step": 4680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945077800845109e-05,
+      "loss": 0.627,
+      "step": 4681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9450544998581836e-05,
+      "loss": 0.5838,
+      "step": 4682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94503119406916e-05,
+      "loss": 0.613,
+      "step": 4683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.945007883478156e-05,
+      "loss": 0.639,
+      "step": 4684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94498456808529e-05,
+      "loss": 0.6275,
+      "step": 4685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9449612478906808e-05,
+      "loss": 0.6226,
+      "step": 4686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9449379228944464e-05,
+      "loss": 0.6761,
+      "step": 4687
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944914593096706e-05,
+      "loss": 0.6354,
+      "step": 4688
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944891258497578e-05,
+      "loss": 0.5645,
+      "step": 4689
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9448679190971806e-05,
+      "loss": 0.5968,
+      "step": 4690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9448445748956325e-05,
+      "loss": 0.6069,
+      "step": 4691
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9448212258930526e-05,
+      "loss": 0.6343,
+      "step": 4692
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944797872089559e-05,
+      "loss": 0.5467,
+      "step": 4693
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944774513485271e-05,
+      "loss": 0.5853,
+      "step": 4694
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944751150080307e-05,
+      "loss": 0.6333,
+      "step": 4695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9447277818747854e-05,
+      "loss": 0.6054,
+      "step": 4696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944704408868826e-05,
+      "loss": 0.6354,
+      "step": 4697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9446810310625463e-05,
+      "loss": 0.635,
+      "step": 4698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9446576484560653e-05,
+      "loss": 0.6178,
+      "step": 4699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9446342610495025e-05,
+      "loss": 0.5158,
+      "step": 4700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9446108688429764e-05,
+      "loss": 0.6371,
+      "step": 4701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944587471836606e-05,
+      "loss": 0.6768,
+      "step": 4702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9445640700305098e-05,
+      "loss": 0.5992,
+      "step": 4703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9445406634248067e-05,
+      "loss": 0.6581,
+      "step": 4704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9445172520196162e-05,
+      "loss": 0.6154,
+      "step": 4705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944493835815057e-05,
+      "loss": 0.592,
+      "step": 4706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9444704148112477e-05,
+      "loss": 0.5837,
+      "step": 4707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944446989008308e-05,
+      "loss": 0.6129,
+      "step": 4708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9444235584063564e-05,
+      "loss": 0.671,
+      "step": 4709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944400123005512e-05,
+      "loss": 0.6347,
+      "step": 4710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9443766828058937e-05,
+      "loss": 0.6188,
+      "step": 4711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9443532378076214e-05,
+      "loss": 0.6392,
+      "step": 4712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944329788010814e-05,
+      "loss": 0.6175,
+      "step": 4713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9443063334155896e-05,
+      "loss": 0.5667,
+      "step": 4714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9442828740220686e-05,
+      "loss": 0.6412,
+      "step": 4715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94425940983037e-05,
+      "loss": 0.6239,
+      "step": 4716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944235940840612e-05,
+      "loss": 0.5811,
+      "step": 4717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9442124670529154e-05,
+      "loss": 0.6472,
+      "step": 4718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9441889884673984e-05,
+      "loss": 0.5938,
+      "step": 4719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944165505084181e-05,
+      "loss": 0.6083,
+      "step": 4720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944142016903382e-05,
+      "loss": 0.6303,
+      "step": 4721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9441185239251204e-05,
+      "loss": 0.6139,
+      "step": 4722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9440950261495163e-05,
+      "loss": 0.5907,
+      "step": 4723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9440715235766894e-05,
+      "loss": 0.598,
+      "step": 4724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944048016206758e-05,
+      "loss": 0.6118,
+      "step": 4725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9440245040398427e-05,
+      "loss": 0.5938,
+      "step": 4726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.944000987076062e-05,
+      "loss": 0.5862,
+      "step": 4727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9439774653155358e-05,
+      "loss": 0.6141,
+      "step": 4728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9439539387583838e-05,
+      "loss": 0.6815,
+      "step": 4729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9439304074047257e-05,
+      "loss": 0.5943,
+      "step": 4730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9439068712546803e-05,
+      "loss": 0.5705,
+      "step": 4731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9438833303083677e-05,
+      "loss": 0.6035,
+      "step": 4732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9438597845659076e-05,
+      "loss": 0.6741,
+      "step": 4733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9438362340274195e-05,
+      "loss": 0.6229,
+      "step": 4734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943812678693023e-05,
+      "loss": 0.6353,
+      "step": 4735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943789118562838e-05,
+      "loss": 0.5983,
+      "step": 4736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943765553636984e-05,
+      "loss": 0.6158,
+      "step": 4737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943741983915581e-05,
+      "loss": 0.6544,
+      "step": 4738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943718409398748e-05,
+      "loss": 0.6591,
+      "step": 4739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943694830086606e-05,
+      "loss": 0.6557,
+      "step": 4740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9436712459792743e-05,
+      "loss": 0.6292,
+      "step": 4741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943647657076872e-05,
+      "loss": 0.6734,
+      "step": 4742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94362406337952e-05,
+      "loss": 0.5891,
+      "step": 4743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9436004648873376e-05,
+      "loss": 0.6079,
+      "step": 4744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943576861600445e-05,
+      "loss": 0.643,
+      "step": 4745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943553253518962e-05,
+      "loss": 0.6034,
+      "step": 4746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943529640643009e-05,
+      "loss": 0.6118,
+      "step": 4747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9435060229727048e-05,
+      "loss": 0.6146,
+      "step": 4748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9434824005081708e-05,
+      "loss": 0.6522,
+      "step": 4749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943458773249526e-05,
+      "loss": 0.6003,
+      "step": 4750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943435141196891e-05,
+      "loss": 0.5854,
+      "step": 4751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9434115043503858e-05,
+      "loss": 0.6531,
+      "step": 4752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9433878627101307e-05,
+      "loss": 0.6278,
+      "step": 4753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943364216276245e-05,
+      "loss": 0.6039,
+      "step": 4754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94334056504885e-05,
+      "loss": 0.6378,
+      "step": 4755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943316909028065e-05,
+      "loss": 0.6014,
+      "step": 4756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9432932482140104e-05,
+      "loss": 0.6693,
+      "step": 4757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943269582606807e-05,
+      "loss": 0.6959,
+      "step": 4758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9432459122065747e-05,
+      "loss": 0.5885,
+      "step": 4759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943222237013433e-05,
+      "loss": 0.6144,
+      "step": 4760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943198557027503e-05,
+      "loss": 0.6377,
+      "step": 4761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9431748722489055e-05,
+      "loss": 0.6343,
+      "step": 4762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94315118267776e-05,
+      "loss": 0.5746,
+      "step": 4763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9431274883141867e-05,
+      "loss": 0.569,
+      "step": 4764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9431037891583067e-05,
+      "loss": 0.6653,
+      "step": 4765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9430800852102403e-05,
+      "loss": 0.5807,
+      "step": 4766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9430563764701076e-05,
+      "loss": 0.5718,
+      "step": 4767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9430326629380293e-05,
+      "loss": 0.5913,
+      "step": 4768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.943008944614126e-05,
+      "loss": 0.618,
+      "step": 4769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9429852214985182e-05,
+      "loss": 0.6375,
+      "step": 4770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.942961493591326e-05,
+      "loss": 0.6036,
+      "step": 4771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.94293776089267e-05,
+      "loss": 0.5886,
+      "step": 4772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.942914023402672e-05,
+      "loss": 0.5835,
+      "step": 4773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.942890281121451e-05,
+      "loss": 0.6165,
+      "step": 4774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9428665340491287e-05,
+      "loss": 0.5859,
+      "step": 4775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942842782185825e-05,
+      "loss": 0.6438,
+      "step": 4776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9428190255316612e-05,
+      "loss": 0.581,
+      "step": 4777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9427952640867577e-05,
+      "loss": 0.6503,
+      "step": 4778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9427714978512354e-05,
+      "loss": 0.6023,
+      "step": 4779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9427477268252157e-05,
+      "loss": 0.6401,
+      "step": 4780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9427239510088177e-05,
+      "loss": 0.6173,
+      "step": 4781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9427001704021637e-05,
+      "loss": 0.638,
+      "step": 4782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9426763850053742e-05,
+      "loss": 0.6458,
+      "step": 4783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9426525948185696e-05,
+      "loss": 0.6468,
+      "step": 4784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9426287998418713e-05,
+      "loss": 0.5919,
+      "step": 4785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9426050000754e-05,
+      "loss": 0.5582,
+      "step": 4786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9425811955192763e-05,
+      "loss": 0.5944,
+      "step": 4787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9425573861736215e-05,
+      "loss": 0.6094,
+      "step": 4788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942533572038557e-05,
+      "loss": 0.6493,
+      "step": 4789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942509753114203e-05,
+      "loss": 0.6597,
+      "step": 4790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9424859294006812e-05,
+      "loss": 0.5781,
+      "step": 4791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9424621008981125e-05,
+      "loss": 0.6024,
+      "step": 4792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9424382676066175e-05,
+      "loss": 0.6151,
+      "step": 4793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942414429526318e-05,
+      "loss": 0.6627,
+      "step": 4794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9423905866573346e-05,
+      "loss": 0.6527,
+      "step": 4795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9423667389997884e-05,
+      "loss": 0.6616,
+      "step": 4796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9423428865538013e-05,
+      "loss": 0.6193,
+      "step": 4797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9423190293194935e-05,
+      "loss": 0.5355,
+      "step": 4798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9422951672969868e-05,
+      "loss": 0.6521,
+      "step": 4799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9422713004864028e-05,
+      "loss": 0.5675,
+      "step": 4800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942247428887862e-05,
+      "loss": 0.672,
+      "step": 4801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9422235525014864e-05,
+      "loss": 0.5484,
+      "step": 4802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9421996713273967e-05,
+      "loss": 0.6327,
+      "step": 4803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9421757853657145e-05,
+      "loss": 0.592,
+      "step": 4804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9421518946165614e-05,
+      "loss": 0.5928,
+      "step": 4805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9421279990800587e-05,
+      "loss": 0.6552,
+      "step": 4806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9421040987563274e-05,
+      "loss": 0.6252,
+      "step": 4807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9420801936454895e-05,
+      "loss": 0.6133,
+      "step": 4808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942056283747666e-05,
+      "loss": 0.6223,
+      "step": 4809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.942032369062979e-05,
+      "loss": 0.6003,
+      "step": 4810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9420084495915494e-05,
+      "loss": 0.5992,
+      "step": 4811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941984525333499e-05,
+      "loss": 0.5779,
+      "step": 4812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9419605962889494e-05,
+      "loss": 0.6151,
+      "step": 4813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941936662458022e-05,
+      "loss": 0.6609,
+      "step": 4814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941912723840839e-05,
+      "loss": 0.6391,
+      "step": 4815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9418887804375212e-05,
+      "loss": 0.6137,
+      "step": 4816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9418648322481908e-05,
+      "loss": 0.5364,
+      "step": 4817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9418408792729694e-05,
+      "loss": 0.6927,
+      "step": 4818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9418169215119784e-05,
+      "loss": 0.6384,
+      "step": 4819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.94179295896534e-05,
+      "loss": 0.6151,
+      "step": 4820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9417689916331757e-05,
+      "loss": 0.5914,
+      "step": 4821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9417450195156074e-05,
+      "loss": 0.6457,
+      "step": 4822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941721042612757e-05,
+      "loss": 0.5531,
+      "step": 4823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9416970609247457e-05,
+      "loss": 0.5892,
+      "step": 4824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941673074451696e-05,
+      "loss": 0.5857,
+      "step": 4825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9416490831937298e-05,
+      "loss": 0.6178,
+      "step": 4826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941625087150969e-05,
+      "loss": 0.576,
+      "step": 4827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9416010863235352e-05,
+      "loss": 0.5443,
+      "step": 4828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9415770807115502e-05,
+      "loss": 0.5839,
+      "step": 4829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941553070315137e-05,
+      "loss": 0.6121,
+      "step": 4830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9415290551344163e-05,
+      "loss": 0.6182,
+      "step": 4831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941505035169511e-05,
+      "loss": 0.6441,
+      "step": 4832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9414810104205425e-05,
+      "loss": 0.6079,
+      "step": 4833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9414569808876336e-05,
+      "loss": 0.5832,
+      "step": 4834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9414329465709063e-05,
+      "loss": 0.5381,
+      "step": 4835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941408907470482e-05,
+      "loss": 0.6027,
+      "step": 4836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9413848635864837e-05,
+      "loss": 0.6309,
+      "step": 4837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9413608149190327e-05,
+      "loss": 0.5962,
+      "step": 4838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9413367614682522e-05,
+      "loss": 0.6225,
+      "step": 4839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9413127032342638e-05,
+      "loss": 0.5827,
+      "step": 4840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.94128864021719e-05,
+      "loss": 0.7138,
+      "step": 4841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9412645724171527e-05,
+      "loss": 0.6241,
+      "step": 4842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9412404998342747e-05,
+      "loss": 0.627,
+      "step": 4843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9412164224686777e-05,
+      "loss": 0.6518,
+      "step": 4844
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941192340320485e-05,
+      "loss": 0.6026,
+      "step": 4845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.941168253389818e-05,
+      "loss": 0.6013,
+      "step": 4846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9411441616767993e-05,
+      "loss": 0.6123,
+      "step": 4847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9411200651815517e-05,
+      "loss": 0.5896,
+      "step": 4848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9410959639041976e-05,
+      "loss": 0.6523,
+      "step": 4849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9410718578448588e-05,
+      "loss": 0.5817,
+      "step": 4850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9410477470036587e-05,
+      "loss": 0.6263,
+      "step": 4851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9410236313807195e-05,
+      "loss": 0.6014,
+      "step": 4852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940999510976163e-05,
+      "loss": 0.556,
+      "step": 4853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9409753857901132e-05,
+      "loss": 0.6349,
+      "step": 4854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9409512558226914e-05,
+      "loss": 0.6086,
+      "step": 4855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940927121074021e-05,
+      "loss": 0.6642,
+      "step": 4856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940902981544224e-05,
+      "loss": 0.6258,
+      "step": 4857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9408788372334237e-05,
+      "loss": 0.6212,
+      "step": 4858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940854688141742e-05,
+      "loss": 0.5512,
+      "step": 4859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9408305342693026e-05,
+      "loss": 0.595,
+      "step": 4860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9408063756162274e-05,
+      "loss": 0.6472,
+      "step": 4861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9407822121826395e-05,
+      "loss": 0.6115,
+      "step": 4862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940758043968662e-05,
+      "loss": 0.5854,
+      "step": 4863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940733870974417e-05,
+      "loss": 0.6221,
+      "step": 4864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940709693200028e-05,
+      "loss": 0.5993,
+      "step": 4865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9406855106456174e-05,
+      "loss": 0.6713,
+      "step": 4866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940661323311308e-05,
+      "loss": 0.6365,
+      "step": 4867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940637131197223e-05,
+      "loss": 0.6266,
+      "step": 4868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9406129343034855e-05,
+      "loss": 0.627,
+      "step": 4869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9405887326302185e-05,
+      "loss": 0.6159,
+      "step": 4870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940564526177544e-05,
+      "loss": 0.6023,
+      "step": 4871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940540314945586e-05,
+      "loss": 0.6173,
+      "step": 4872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9405160989344673e-05,
+      "loss": 0.6238,
+      "step": 4873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9404918781443104e-05,
+      "loss": 0.6414,
+      "step": 4874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9404676525752393e-05,
+      "loss": 0.5983,
+      "step": 4875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9404434222273764e-05,
+      "loss": 0.568,
+      "step": 4876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9404191871008454e-05,
+      "loss": 0.6251,
+      "step": 4877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9403949471957688e-05,
+      "loss": 0.6094,
+      "step": 4878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.94037070251227e-05,
+      "loss": 0.6362,
+      "step": 4879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9403464530504724e-05,
+      "loss": 0.6155,
+      "step": 4880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9403221988104992e-05,
+      "loss": 0.6172,
+      "step": 4881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940297939792473e-05,
+      "loss": 0.6017,
+      "step": 4882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9402736759965185e-05,
+      "loss": 0.6462,
+      "step": 4883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9402494074227574e-05,
+      "loss": 0.6197,
+      "step": 4884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9402251340713136e-05,
+      "loss": 0.6244,
+      "step": 4885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940200855942311e-05,
+      "loss": 0.6698,
+      "step": 4886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9401765730358725e-05,
+      "loss": 0.6493,
+      "step": 4887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.940152285352121e-05,
+      "loss": 0.6145,
+      "step": 4888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9401279928911807e-05,
+      "loss": 0.6113,
+      "step": 4889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9401036956531747e-05,
+      "loss": 0.5861,
+      "step": 4890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9400793936382265e-05,
+      "loss": 0.597,
+      "step": 4891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9400550868464597e-05,
+      "loss": 0.5992,
+      "step": 4892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9400307752779973e-05,
+      "loss": 0.5765,
+      "step": 4893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9400064589329637e-05,
+      "loss": 0.6289,
+      "step": 4894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9399821378114817e-05,
+      "loss": 0.5396,
+      "step": 4895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9399578119136752e-05,
+      "loss": 0.6782,
+      "step": 4896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939933481239668e-05,
+      "loss": 0.59,
+      "step": 4897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9399091457895834e-05,
+      "loss": 0.6616,
+      "step": 4898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939884805563545e-05,
+      "loss": 0.5971,
+      "step": 4899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939860460561677e-05,
+      "loss": 0.5461,
+      "step": 4900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9398361107841023e-05,
+      "loss": 0.6458,
+      "step": 4901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9398117562309452e-05,
+      "loss": 0.6249,
+      "step": 4902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.93978739690233e-05,
+      "loss": 0.5232,
+      "step": 4903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939763032798379e-05,
+      "loss": 0.5821,
+      "step": 4904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9397386639192167e-05,
+      "loss": 0.6754,
+      "step": 4905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939714290264968e-05,
+      "loss": 0.5705,
+      "step": 4906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9396899118357545e-05,
+      "loss": 0.5644,
+      "step": 4907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9396655286317025e-05,
+      "loss": 0.6427,
+      "step": 4908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939641140652934e-05,
+      "loss": 0.6344,
+      "step": 4909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939616747899574e-05,
+      "loss": 0.6641,
+      "step": 4910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939592350371746e-05,
+      "loss": 0.6366,
+      "step": 4911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9395679480695745e-05,
+      "loss": 0.6334,
+      "step": 4912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9395435409931823e-05,
+      "loss": 0.6034,
+      "step": 4913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.939519129142695e-05,
+      "loss": 0.5873,
+      "step": 4914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9394947125182354e-05,
+      "loss": 0.582,
+      "step": 4915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9394702911199282e-05,
+      "loss": 0.5919,
+      "step": 4916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9394458649478972e-05,
+      "loss": 0.6588,
+      "step": 4917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9394214340022667e-05,
+      "loss": 0.5931,
+      "step": 4918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9393969982831607e-05,
+      "loss": 0.6033,
+      "step": 4919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9393725577907038e-05,
+      "loss": 0.6292,
+      "step": 4920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9393481125250194e-05,
+      "loss": 0.6208,
+      "step": 4921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9393236624862322e-05,
+      "loss": 0.6444,
+      "step": 4922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9392992076744665e-05,
+      "loss": 0.5707,
+      "step": 4923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9392747480898463e-05,
+      "loss": 0.5689,
+      "step": 4924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9392502837324964e-05,
+      "loss": 0.6259,
+      "step": 4925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9392258146025403e-05,
+      "loss": 0.5596,
+      "step": 4926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9392013407001027e-05,
+      "loss": 0.6489,
+      "step": 4927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9391768620253085e-05,
+      "loss": 0.6401,
+      "step": 4928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9391523785782814e-05,
+      "loss": 0.6177,
+      "step": 4929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9391278903591456e-05,
+      "loss": 0.6095,
+      "step": 4930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9391033973680265e-05,
+      "loss": 0.6394,
+      "step": 4931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.939078899605048e-05,
+      "loss": 0.6394,
+      "step": 4932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9390543970703343e-05,
+      "loss": 0.6287,
+      "step": 4933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9390298897640104e-05,
+      "loss": 0.604,
+      "step": 4934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9390053776862004e-05,
+      "loss": 0.5999,
+      "step": 4935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938980860837029e-05,
+      "loss": 0.5858,
+      "step": 4936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9389563392166212e-05,
+      "loss": 0.6501,
+      "step": 4937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9389318128251014e-05,
+      "loss": 0.6382,
+      "step": 4938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938907281662594e-05,
+      "loss": 0.5996,
+      "step": 4939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938882745729223e-05,
+      "loss": 0.5864,
+      "step": 4940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9388582050251144e-05,
+      "loss": 0.5922,
+      "step": 4941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9388336595503926e-05,
+      "loss": 0.6493,
+      "step": 4942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9388091093051814e-05,
+      "loss": 0.5843,
+      "step": 4943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9387845542896065e-05,
+      "loss": 0.6076,
+      "step": 4944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9387599945037923e-05,
+      "loss": 0.6415,
+      "step": 4945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9387354299478637e-05,
+      "loss": 0.5953,
+      "step": 4946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9387108606219452e-05,
+      "loss": 0.5986,
+      "step": 4947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9386862865261623e-05,
+      "loss": 0.6178,
+      "step": 4948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938661707660639e-05,
+      "loss": 0.5456,
+      "step": 4949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9386371240255012e-05,
+      "loss": 0.5963,
+      "step": 4950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938612535620873e-05,
+      "loss": 0.5614,
+      "step": 4951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9385879424468796e-05,
+      "loss": 0.6192,
+      "step": 4952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938563344503646e-05,
+      "loss": 0.5814,
+      "step": 4953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938538741791297e-05,
+      "loss": 0.5936,
+      "step": 4954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938514134309958e-05,
+      "loss": 0.6711,
+      "step": 4955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9384895220597538e-05,
+      "loss": 0.6062,
+      "step": 4956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9384649050408094e-05,
+      "loss": 0.6533,
+      "step": 4957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.93844028325325e-05,
+      "loss": 0.6113,
+      "step": 4958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938415656697201e-05,
+      "loss": 0.5553,
+      "step": 4959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938391025372787e-05,
+      "loss": 0.662,
+      "step": 4960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9383663892801333e-05,
+      "loss": 0.6498,
+      "step": 4961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9383417484193653e-05,
+      "loss": 0.6456,
+      "step": 4962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9383171027906078e-05,
+      "loss": 0.6492,
+      "step": 4963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9382924523939866e-05,
+      "loss": 0.5768,
+      "step": 4964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9382677972296266e-05,
+      "loss": 0.5911,
+      "step": 4965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938243137297653e-05,
+      "loss": 0.5947,
+      "step": 4966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9382184725981915e-05,
+      "loss": 0.5627,
+      "step": 4967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938193803131367e-05,
+      "loss": 0.5999,
+      "step": 4968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938169128897305e-05,
+      "loss": 0.6421,
+      "step": 4969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938144449896131e-05,
+      "loss": 0.5742,
+      "step": 4970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.93811976612797e-05,
+      "loss": 0.6167,
+      "step": 4971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9380950775929477e-05,
+      "loss": 0.6196,
+      "step": 4972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9380703842911902e-05,
+      "loss": 0.584,
+      "step": 4973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.938045686222822e-05,
+      "loss": 0.57,
+      "step": 4974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9380209833879687e-05,
+      "loss": 0.6635,
+      "step": 4975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9379962757867564e-05,
+      "loss": 0.6422,
+      "step": 4976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.93797156341931e-05,
+      "loss": 0.6199,
+      "step": 4977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9379468462857558e-05,
+      "loss": 0.5762,
+      "step": 4978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9379221243862188e-05,
+      "loss": 0.5626,
+      "step": 4979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937897397720825e-05,
+      "loss": 0.6342,
+      "step": 4980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9378726662896996e-05,
+      "loss": 0.5994,
+      "step": 4981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937847930092968e-05,
+      "loss": 0.6045,
+      "step": 4982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937823189130757e-05,
+      "loss": 0.6284,
+      "step": 4983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937798443403192e-05,
+      "loss": 0.6374,
+      "step": 4984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937773692910398e-05,
+      "loss": 0.6374,
+      "step": 4985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9377489376525016e-05,
+      "loss": 0.5878,
+      "step": 4986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9377241776296277e-05,
+      "loss": 0.6346,
+      "step": 4987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9376994128419032e-05,
+      "loss": 0.6088,
+      "step": 4988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9376746432894527e-05,
+      "loss": 0.5799,
+      "step": 4989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9376498689724036e-05,
+      "loss": 0.6354,
+      "step": 4990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9376250898908802e-05,
+      "loss": 0.6493,
+      "step": 4991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9376003060450092e-05,
+      "loss": 0.6245,
+      "step": 4992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9375755174349168e-05,
+      "loss": 0.5905,
+      "step": 4993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937550724060728e-05,
+      "loss": 0.5866,
+      "step": 4994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9375259259225704e-05,
+      "loss": 0.5819,
+      "step": 4995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9375011230205683e-05,
+      "loss": 0.597,
+      "step": 4996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9374763153548484e-05,
+      "loss": 0.6001,
+      "step": 4997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937451502925537e-05,
+      "loss": 0.6081,
+      "step": 4998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9374266857327598e-05,
+      "loss": 0.6355,
+      "step": 4999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937401863776643e-05,
+      "loss": 0.6124,
+      "step": 5000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9373770370573132e-05,
+      "loss": 0.667,
+      "step": 5001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9373522055748958e-05,
+      "loss": 0.5816,
+      "step": 5002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9373273693295176e-05,
+      "loss": 0.6355,
+      "step": 5003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9373025283213043e-05,
+      "loss": 0.646,
+      "step": 5004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9372776825503826e-05,
+      "loss": 0.5641,
+      "step": 5005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9372528320168782e-05,
+      "loss": 0.6178,
+      "step": 5006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9372279767209176e-05,
+      "loss": 0.6455,
+      "step": 5007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9372031166626277e-05,
+      "loss": 0.5899,
+      "step": 5008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937178251842134e-05,
+      "loss": 0.5962,
+      "step": 5009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9371533822595626e-05,
+      "loss": 0.5861,
+      "step": 5010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9371285079150412e-05,
+      "loss": 0.6726,
+      "step": 5011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937103628808695e-05,
+      "loss": 0.6336,
+      "step": 5012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9370787449406512e-05,
+      "loss": 0.6338,
+      "step": 5013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9370538563110354e-05,
+      "loss": 0.627,
+      "step": 5014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.937028962919975e-05,
+      "loss": 0.563,
+      "step": 5015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9370040647675952e-05,
+      "loss": 0.5707,
+      "step": 5016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936979161854024e-05,
+      "loss": 0.5647,
+      "step": 5017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936954254179387e-05,
+      "loss": 0.6295,
+      "step": 5018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9369293417438113e-05,
+      "loss": 0.7314,
+      "step": 5019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9369044245474235e-05,
+      "loss": 0.6286,
+      "step": 5020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9368795025903494e-05,
+      "loss": 0.6092,
+      "step": 5021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9368545758727163e-05,
+      "loss": 0.6594,
+      "step": 5022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9368296443946506e-05,
+      "loss": 0.6139,
+      "step": 5023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936804708156279e-05,
+      "loss": 0.5952,
+      "step": 5024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9367797671577285e-05,
+      "loss": 0.5959,
+      "step": 5025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9367548213991258e-05,
+      "loss": 0.6359,
+      "step": 5026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9367298708805977e-05,
+      "loss": 0.6027,
+      "step": 5027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9367049156022703e-05,
+      "loss": 0.5844,
+      "step": 5028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9366799555642713e-05,
+      "loss": 0.633,
+      "step": 5029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9366549907667268e-05,
+      "loss": 0.6222,
+      "step": 5030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936630021209764e-05,
+      "loss": 0.6463,
+      "step": 5031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.93660504689351e-05,
+      "loss": 0.6142,
+      "step": 5032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9365800678180914e-05,
+      "loss": 0.6417,
+      "step": 5033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936555083983635e-05,
+      "loss": 0.6003,
+      "step": 5034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936530095390268e-05,
+      "loss": 0.6153,
+      "step": 5035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9365051020381176e-05,
+      "loss": 0.6312,
+      "step": 5036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9364801039273103e-05,
+      "loss": 0.6202,
+      "step": 5037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9364551010579733e-05,
+      "loss": 0.5803,
+      "step": 5038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9364300934302335e-05,
+      "loss": 0.6219,
+      "step": 5039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9364050810442184e-05,
+      "loss": 0.6352,
+      "step": 5040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9363800639000547e-05,
+      "loss": 0.6221,
+      "step": 5041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9363550419978697e-05,
+      "loss": 0.6347,
+      "step": 5042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9363300153377905e-05,
+      "loss": 0.6005,
+      "step": 5043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9363049839199446e-05,
+      "loss": 0.6253,
+      "step": 5044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9362799477444586e-05,
+      "loss": 0.6741,
+      "step": 5045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.93625490681146e-05,
+      "loss": 0.6321,
+      "step": 5046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936229861121076e-05,
+      "loss": 0.6399,
+      "step": 5047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9362048106734337e-05,
+      "loss": 0.5933,
+      "step": 5048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9361797554686607e-05,
+      "loss": 0.6159,
+      "step": 5049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9361546955068845e-05,
+      "loss": 0.6504,
+      "step": 5050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9361296307882317e-05,
+      "loss": 0.5916,
+      "step": 5051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9361045613128305e-05,
+      "loss": 0.5938,
+      "step": 5052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9360794870808074e-05,
+      "loss": 0.658,
+      "step": 5053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9360544080922904e-05,
+      "loss": 0.6278,
+      "step": 5054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.936029324347407e-05,
+      "loss": 0.6116,
+      "step": 5055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9360042358462845e-05,
+      "loss": 0.5977,
+      "step": 5056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9359791425890503e-05,
+      "loss": 0.6265,
+      "step": 5057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9359540445758318e-05,
+      "loss": 0.668,
+      "step": 5058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9359289418067568e-05,
+      "loss": 0.6159,
+      "step": 5059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9359038342819526e-05,
+      "loss": 0.627,
+      "step": 5060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9358787220015473e-05,
+      "loss": 0.5934,
+      "step": 5061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.935853604965668e-05,
+      "loss": 0.5597,
+      "step": 5062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9358284831744423e-05,
+      "loss": 0.6213,
+      "step": 5063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.935803356627998e-05,
+      "loss": 0.5777,
+      "step": 5064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.935778225326463e-05,
+      "loss": 0.5941,
+      "step": 5065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9357530892699645e-05,
+      "loss": 0.5876,
+      "step": 5066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.935727948458631e-05,
+      "loss": 0.549,
+      "step": 5067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.935702802892589e-05,
+      "loss": 0.6484,
+      "step": 5068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9356776525719677e-05,
+      "loss": 0.6068,
+      "step": 5069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9356524974968938e-05,
+      "loss": 0.615,
+      "step": 5070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9356273376674958e-05,
+      "loss": 0.6539,
+      "step": 5071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.935602173083901e-05,
+      "loss": 0.6105,
+      "step": 5072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.935577003746238e-05,
+      "loss": 0.6027,
+      "step": 5073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9355518296546337e-05,
+      "loss": 0.605,
+      "step": 5074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9355266508092172e-05,
+      "loss": 0.6172,
+      "step": 5075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9355014672101154e-05,
+      "loss": 0.6066,
+      "step": 5076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9354762788574568e-05,
+      "loss": 0.6145,
+      "step": 5077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9354510857513692e-05,
+      "loss": 0.5702,
+      "step": 5078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9354258878919805e-05,
+      "loss": 0.5785,
+      "step": 5079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9354006852794192e-05,
+      "loss": 0.5453,
+      "step": 5080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.935375477913813e-05,
+      "loss": 0.6454,
+      "step": 5081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.93535026579529e-05,
+      "loss": 0.6288,
+      "step": 5082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9353250489239782e-05,
+      "loss": 0.637,
+      "step": 5083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.935299827300006e-05,
+      "loss": 0.6241,
+      "step": 5084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9352746009235015e-05,
+      "loss": 0.5562,
+      "step": 5085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9352493697945932e-05,
+      "loss": 0.592,
+      "step": 5086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9352241339134082e-05,
+      "loss": 0.614,
+      "step": 5087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.935198893280076e-05,
+      "loss": 0.6119,
+      "step": 5088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9351736478947246e-05,
+      "loss": 0.6444,
+      "step": 5089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9351483977574815e-05,
+      "loss": 0.563,
+      "step": 5090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9351231428684757e-05,
+      "loss": 0.6341,
+      "step": 5091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9350978832278353e-05,
+      "loss": 0.652,
+      "step": 5092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9350726188356887e-05,
+      "loss": 0.5979,
+      "step": 5093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9350473496921643e-05,
+      "loss": 0.6036,
+      "step": 5094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9350220757973905e-05,
+      "loss": 0.6325,
+      "step": 5095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9349967971514958e-05,
+      "loss": 0.5855,
+      "step": 5096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9349715137546082e-05,
+      "loss": 0.6459,
+      "step": 5097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934946225606857e-05,
+      "loss": 0.6454,
+      "step": 5098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.93492093270837e-05,
+      "loss": 0.6488,
+      "step": 5099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9348956350592758e-05,
+      "loss": 0.5285,
+      "step": 5100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934870332659703e-05,
+      "loss": 0.5748,
+      "step": 5101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934845025509781e-05,
+      "loss": 0.6144,
+      "step": 5102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9348197136096366e-05,
+      "loss": 0.634,
+      "step": 5103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9347943969594e-05,
+      "loss": 0.6872,
+      "step": 5104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9347690755591994e-05,
+      "loss": 0.5941,
+      "step": 5105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934743749409163e-05,
+      "loss": 0.5918,
+      "step": 5106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.93471841850942e-05,
+      "loss": 0.6278,
+      "step": 5107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934693082860099e-05,
+      "loss": 0.5982,
+      "step": 5108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934667742461329e-05,
+      "loss": 0.6679,
+      "step": 5109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9346423973132383e-05,
+      "loss": 0.6036,
+      "step": 5110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9346170474159556e-05,
+      "loss": 0.6363,
+      "step": 5111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9345916927696103e-05,
+      "loss": 0.592,
+      "step": 5112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9345663333743306e-05,
+      "loss": 0.6013,
+      "step": 5113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9345409692302458e-05,
+      "loss": 0.633,
+      "step": 5114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934515600337485e-05,
+      "loss": 0.6341,
+      "step": 5115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934490226696176e-05,
+      "loss": 0.6146,
+      "step": 5116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934464848306449e-05,
+      "loss": 0.6053,
+      "step": 5117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934439465168432e-05,
+      "loss": 0.6168,
+      "step": 5118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934414077282255e-05,
+      "loss": 0.6069,
+      "step": 5119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9343886846480463e-05,
+      "loss": 0.6093,
+      "step": 5120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934363287265935e-05,
+      "loss": 0.5773,
+      "step": 5121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9343378851360498e-05,
+      "loss": 0.5624,
+      "step": 5122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9343124782585206e-05,
+      "loss": 0.6116,
+      "step": 5123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9342870666334757e-05,
+      "loss": 0.6591,
+      "step": 5124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934261650261045e-05,
+      "loss": 0.5755,
+      "step": 5125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934236229141357e-05,
+      "loss": 0.5883,
+      "step": 5126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.934210803274541e-05,
+      "loss": 0.6529,
+      "step": 5127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9341853726607268e-05,
+      "loss": 0.6118,
+      "step": 5128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9341599373000427e-05,
+      "loss": 0.6042,
+      "step": 5129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9341344971926184e-05,
+      "loss": 0.6108,
+      "step": 5130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9341090523385832e-05,
+      "loss": 0.646,
+      "step": 5131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9340836027380667e-05,
+      "loss": 0.59,
+      "step": 5132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9340581483911973e-05,
+      "loss": 0.6663,
+      "step": 5133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9340326892981052e-05,
+      "loss": 0.6345,
+      "step": 5134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9340072254589194e-05,
+      "loss": 0.6012,
+      "step": 5135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9339817568737695e-05,
+      "loss": 0.6544,
+      "step": 5136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9339562835427847e-05,
+      "loss": 0.6164,
+      "step": 5137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9339308054660942e-05,
+      "loss": 0.603,
+      "step": 5138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933905322643828e-05,
+      "loss": 0.638,
+      "step": 5139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933879835076115e-05,
+      "loss": 0.6345,
+      "step": 5140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9338543427630857e-05,
+      "loss": 0.6358,
+      "step": 5141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9338288457048687e-05,
+      "loss": 0.6157,
+      "step": 5142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9338033439015937e-05,
+      "loss": 0.6137,
+      "step": 5143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9337778373533906e-05,
+      "loss": 0.6215,
+      "step": 5144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9337523260603888e-05,
+      "loss": 0.6893,
+      "step": 5145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933726810022718e-05,
+      "loss": 0.5676,
+      "step": 5146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9337012892405078e-05,
+      "loss": 0.5307,
+      "step": 5147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9336757637138877e-05,
+      "loss": 0.6691,
+      "step": 5148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933650233442988e-05,
+      "loss": 0.6266,
+      "step": 5149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933624698427938e-05,
+      "loss": 0.639,
+      "step": 5150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933599158668867e-05,
+      "loss": 0.6559,
+      "step": 5151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9335736141659058e-05,
+      "loss": 0.5589,
+      "step": 5152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9335480649191835e-05,
+      "loss": 0.5995,
+      "step": 5153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.93352251092883e-05,
+      "loss": 0.5958,
+      "step": 5154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933496952194975e-05,
+      "loss": 0.6216,
+      "step": 5155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933471388717749e-05,
+      "loss": 0.6173,
+      "step": 5156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9334458204972813e-05,
+      "loss": 0.6039,
+      "step": 5157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9334202475337022e-05,
+      "loss": 0.6082,
+      "step": 5158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9333946698271414e-05,
+      "loss": 0.6114,
+      "step": 5159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9333690873777288e-05,
+      "loss": 0.6362,
+      "step": 5160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9333435001855943e-05,
+      "loss": 0.6364,
+      "step": 5161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9333179082508686e-05,
+      "loss": 0.6192,
+      "step": 5162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933292311573681e-05,
+      "loss": 0.6307,
+      "step": 5163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9332667101541617e-05,
+      "loss": 0.6207,
+      "step": 5164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933241103992441e-05,
+      "loss": 0.588,
+      "step": 5165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9332154930886494e-05,
+      "loss": 0.6061,
+      "step": 5166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933189877442916e-05,
+      "loss": 0.6081,
+      "step": 5167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933164257055372e-05,
+      "loss": 0.6361,
+      "step": 5168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9331386319261467e-05,
+      "loss": 0.582,
+      "step": 5169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933113002055371e-05,
+      "loss": 0.5955,
+      "step": 5170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9330873674431747e-05,
+      "loss": 0.6045,
+      "step": 5171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9330617280896882e-05,
+      "loss": 0.6839,
+      "step": 5172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9330360839950418e-05,
+      "loss": 0.5807,
+      "step": 5173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.933010435159366e-05,
+      "loss": 0.6626,
+      "step": 5174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932984781582791e-05,
+      "loss": 0.6314,
+      "step": 5175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932959123265447e-05,
+      "loss": 0.585,
+      "step": 5176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9329334602074642e-05,
+      "loss": 0.6288,
+      "step": 5177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9329077924089735e-05,
+      "loss": 0.6362,
+      "step": 5178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932882119870105e-05,
+      "loss": 0.6269,
+      "step": 5179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9328564425909892e-05,
+      "loss": 0.7325,
+      "step": 5180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932830760571757e-05,
+      "loss": 0.6453,
+      "step": 5181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932805073812538e-05,
+      "loss": 0.6175,
+      "step": 5182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9327793823134633e-05,
+      "loss": 0.5901,
+      "step": 5183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9327536860746637e-05,
+      "loss": 0.6379,
+      "step": 5184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9327279850962692e-05,
+      "loss": 0.583,
+      "step": 5185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9327022793784106e-05,
+      "loss": 0.6016,
+      "step": 5186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932676568921219e-05,
+      "loss": 0.5988,
+      "step": 5187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9326508537248242e-05,
+      "loss": 0.5786,
+      "step": 5188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9326251337893573e-05,
+      "loss": 0.6087,
+      "step": 5189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9325994091149488e-05,
+      "loss": 0.6517,
+      "step": 5190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.93257367970173e-05,
+      "loss": 0.6293,
+      "step": 5191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932547945549831e-05,
+      "loss": 0.569,
+      "step": 5192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9325222066593828e-05,
+      "loss": 0.6591,
+      "step": 5193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932496463030516e-05,
+      "loss": 0.6189,
+      "step": 5194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9324707146633618e-05,
+      "loss": 0.6184,
+      "step": 5195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9324449615580508e-05,
+      "loss": 0.612,
+      "step": 5196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932419203714714e-05,
+      "loss": 0.5994,
+      "step": 5197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932393441133482e-05,
+      "loss": 0.6643,
+      "step": 5198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932367673814486e-05,
+      "loss": 0.5973,
+      "step": 5199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9323419017578567e-05,
+      "loss": 0.5744,
+      "step": 5200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932316124963725e-05,
+      "loss": 0.6344,
+      "step": 5201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932290343432222e-05,
+      "loss": 0.6143,
+      "step": 5202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9322645571634787e-05,
+      "loss": 0.6058,
+      "step": 5203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9322387661576264e-05,
+      "loss": 0.5519,
+      "step": 5204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.932212970414796e-05,
+      "loss": 0.6436,
+      "step": 5205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9321871699351184e-05,
+      "loss": 0.5349,
+      "step": 5206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9321613647187247e-05,
+      "loss": 0.6445,
+      "step": 5207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9321355547657458e-05,
+      "loss": 0.6054,
+      "step": 5208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9321097400763135e-05,
+      "loss": 0.5834,
+      "step": 5209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9320839206505585e-05,
+      "loss": 0.6581,
+      "step": 5210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9320580964886124e-05,
+      "loss": 0.6165,
+      "step": 5211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9320322675906056e-05,
+      "loss": 0.6655,
+      "step": 5212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9320064339566706e-05,
+      "loss": 0.6247,
+      "step": 5213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9319805955869374e-05,
+      "loss": 0.639,
+      "step": 5214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9319547524815382e-05,
+      "loss": 0.6413,
+      "step": 5215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9319289046406038e-05,
+      "loss": 0.5871,
+      "step": 5216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9319030520642656e-05,
+      "loss": 0.6263,
+      "step": 5217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9318771947526555e-05,
+      "loss": 0.6025,
+      "step": 5218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9318513327059042e-05,
+      "loss": 0.6425,
+      "step": 5219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931825465924143e-05,
+      "loss": 0.6675,
+      "step": 5220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9317995944075042e-05,
+      "loss": 0.5875,
+      "step": 5221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9317737181561188e-05,
+      "loss": 0.6257,
+      "step": 5222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931747837170118e-05,
+      "loss": 0.6117,
+      "step": 5223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9317219514496334e-05,
+      "loss": 0.6042,
+      "step": 5224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931696060994797e-05,
+      "loss": 0.6304,
+      "step": 5225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.93167016580574e-05,
+      "loss": 0.5991,
+      "step": 5226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931644265882594e-05,
+      "loss": 0.6667,
+      "step": 5227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931618361225491e-05,
+      "loss": 0.6146,
+      "step": 5228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931592451834562e-05,
+      "loss": 0.5793,
+      "step": 5229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9315665377099386e-05,
+      "loss": 0.6166,
+      "step": 5230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931540618851753e-05,
+      "loss": 0.6303,
+      "step": 5231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9315146952601367e-05,
+      "loss": 0.5921,
+      "step": 5232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9314887669352216e-05,
+      "loss": 0.5978,
+      "step": 5233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9314628338771393e-05,
+      "loss": 0.6186,
+      "step": 5234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931436896086021e-05,
+      "loss": 0.6207,
+      "step": 5235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9314109535619997e-05,
+      "loss": 0.5468,
+      "step": 5236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9313850063052063e-05,
+      "loss": 0.6647,
+      "step": 5237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931359054315773e-05,
+      "loss": 0.5882,
+      "step": 5238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9313330975938313e-05,
+      "loss": 0.5843,
+      "step": 5239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931307136139514e-05,
+      "loss": 0.6371,
+      "step": 5240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9312811699529518e-05,
+      "loss": 0.5442,
+      "step": 5241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9312551990342775e-05,
+      "loss": 0.6308,
+      "step": 5242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9312292233836225e-05,
+      "loss": 0.6427,
+      "step": 5243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9312032430011195e-05,
+      "loss": 0.572,
+      "step": 5244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9311772578868998e-05,
+      "loss": 0.5944,
+      "step": 5245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931151268041096e-05,
+      "loss": 0.6181,
+      "step": 5246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.93112527346384e-05,
+      "loss": 0.6162,
+      "step": 5247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9310992741552636e-05,
+      "loss": 0.6169,
+      "step": 5248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9310732701154987e-05,
+      "loss": 0.5929,
+      "step": 5249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9310472613446784e-05,
+      "loss": 0.6434,
+      "step": 5250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.931021247842934e-05,
+      "loss": 0.5736,
+      "step": 5251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930995229610398e-05,
+      "loss": 0.6598,
+      "step": 5252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930969206647203e-05,
+      "loss": 0.5713,
+      "step": 5253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9309431789534802e-05,
+      "loss": 0.606,
+      "step": 5254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930917146529363e-05,
+      "loss": 0.6372,
+      "step": 5255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9308911093749828e-05,
+      "loss": 0.6644,
+      "step": 5256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9308650674904723e-05,
+      "loss": 0.6385,
+      "step": 5257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930839020875964e-05,
+      "loss": 0.626,
+      "step": 5258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9308129695315897e-05,
+      "loss": 0.5867,
+      "step": 5259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9307869134574823e-05,
+      "loss": 0.5982,
+      "step": 5260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930760852653774e-05,
+      "loss": 0.6982,
+      "step": 5261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930734787120597e-05,
+      "loss": 0.6143,
+      "step": 5262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9307087168580842e-05,
+      "loss": 0.6316,
+      "step": 5263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9306826418663677e-05,
+      "loss": 0.6439,
+      "step": 5264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9306565621455805e-05,
+      "loss": 0.5864,
+      "step": 5265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9306304776958545e-05,
+      "loss": 0.6126,
+      "step": 5266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9306043885173223e-05,
+      "loss": 0.5966,
+      "step": 5267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930578294610117e-05,
+      "loss": 0.6437,
+      "step": 5268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9305521959743706e-05,
+      "loss": 0.6071,
+      "step": 5269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9305260926102162e-05,
+      "loss": 0.6433,
+      "step": 5270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930499984517786e-05,
+      "loss": 0.6271,
+      "step": 5271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9304738716972128e-05,
+      "loss": 0.5988,
+      "step": 5272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9304477541486294e-05,
+      "loss": 0.5993,
+      "step": 5273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9304216318721687e-05,
+      "loss": 0.6025,
+      "step": 5274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930395504867963e-05,
+      "loss": 0.6562,
+      "step": 5275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9303693731361454e-05,
+      "loss": 0.4892,
+      "step": 5276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9303432366768485e-05,
+      "loss": 0.6241,
+      "step": 5277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9303170954902047e-05,
+      "loss": 0.573,
+      "step": 5278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.930290949576348e-05,
+      "loss": 0.5897,
+      "step": 5279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9302647989354102e-05,
+      "loss": 0.5663,
+      "step": 5280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9302386435675244e-05,
+      "loss": 0.5906,
+      "step": 5281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9302124834728237e-05,
+      "loss": 0.6652,
+      "step": 5282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9301863186514413e-05,
+      "loss": 0.5856,
+      "step": 5283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9301601491035094e-05,
+      "loss": 0.6173,
+      "step": 5284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9301339748291614e-05,
+      "loss": 0.6334,
+      "step": 5285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9301077958285305e-05,
+      "loss": 0.5845,
+      "step": 5286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9300816121017495e-05,
+      "loss": 0.5926,
+      "step": 5287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9300554236489512e-05,
+      "loss": 0.5433,
+      "step": 5288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9300292304702688e-05,
+      "loss": 0.548,
+      "step": 5289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9300030325658358e-05,
+      "loss": 0.6213,
+      "step": 5290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929976829935785e-05,
+      "loss": 0.5998,
+      "step": 5291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9299506225802494e-05,
+      "loss": 0.5946,
+      "step": 5292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9299244104993624e-05,
+      "loss": 0.6208,
+      "step": 5293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9298981936932573e-05,
+      "loss": 0.5687,
+      "step": 5294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929871972162067e-05,
+      "loss": 0.7203,
+      "step": 5295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9298457459059247e-05,
+      "loss": 0.6061,
+      "step": 5296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9298195149249642e-05,
+      "loss": 0.6444,
+      "step": 5297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9297932792193183e-05,
+      "loss": 0.6707,
+      "step": 5298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9297670387891204e-05,
+      "loss": 0.6375,
+      "step": 5299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9297407936345038e-05,
+      "loss": 0.5873,
+      "step": 5300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9297145437556023e-05,
+      "loss": 0.6554,
+      "step": 5301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9296882891525483e-05,
+      "loss": 0.6725,
+      "step": 5302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9296620298254764e-05,
+      "loss": 0.6246,
+      "step": 5303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9296357657745193e-05,
+      "loss": 0.659,
+      "step": 5304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9296094969998106e-05,
+      "loss": 0.6182,
+      "step": 5305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9295832235014838e-05,
+      "loss": 0.5894,
+      "step": 5306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9295569452796723e-05,
+      "loss": 0.6214,
+      "step": 5307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.92953066233451e-05,
+      "loss": 0.6139,
+      "step": 5308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9295043746661296e-05,
+      "loss": 0.6182,
+      "step": 5309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9294780822746658e-05,
+      "loss": 0.6362,
+      "step": 5310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9294517851602513e-05,
+      "loss": 0.5832,
+      "step": 5311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9294254833230207e-05,
+      "loss": 0.6353,
+      "step": 5312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9293991767631063e-05,
+      "loss": 0.6485,
+      "step": 5313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9293728654806427e-05,
+      "loss": 0.5828,
+      "step": 5314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9293465494757635e-05,
+      "loss": 0.6503,
+      "step": 5315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9293202287486027e-05,
+      "loss": 0.6202,
+      "step": 5316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929293903299293e-05,
+      "loss": 0.6442,
+      "step": 5317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9292675731279692e-05,
+      "loss": 0.5882,
+      "step": 5318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929241238234764e-05,
+      "loss": 0.6096,
+      "step": 5319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929214898619813e-05,
+      "loss": 0.6133,
+      "step": 5320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9291885542832482e-05,
+      "loss": 0.6033,
+      "step": 5321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9291622052252047e-05,
+      "loss": 0.5949,
+      "step": 5322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9291358514458157e-05,
+      "loss": 0.6666,
+      "step": 5323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9291094929452156e-05,
+      "loss": 0.5699,
+      "step": 5324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9290831297235374e-05,
+      "loss": 0.6453,
+      "step": 5325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9290567617809163e-05,
+      "loss": 0.5905,
+      "step": 5326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9290303891174855e-05,
+      "loss": 0.6723,
+      "step": 5327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.929004011733379e-05,
+      "loss": 0.6044,
+      "step": 5328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9289776296287318e-05,
+      "loss": 0.6296,
+      "step": 5329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9289512428036765e-05,
+      "loss": 0.6682,
+      "step": 5330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9289248512583482e-05,
+      "loss": 0.6759,
+      "step": 5331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9288984549928803e-05,
+      "loss": 0.5927,
+      "step": 5332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9288720540074077e-05,
+      "loss": 0.6436,
+      "step": 5333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928845648302064e-05,
+      "loss": 0.5787,
+      "step": 5334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9288192378769838e-05,
+      "loss": 0.5699,
+      "step": 5335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9287928227323008e-05,
+      "loss": 0.6201,
+      "step": 5336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928766402868149e-05,
+      "loss": 0.6146,
+      "step": 5337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928739978284664e-05,
+      "loss": 0.5731,
+      "step": 5338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9287135489819786e-05,
+      "loss": 0.5585,
+      "step": 5339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928687114960228e-05,
+      "loss": 0.5905,
+      "step": 5340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9286606762195458e-05,
+      "loss": 0.5968,
+      "step": 5341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9286342327600672e-05,
+      "loss": 0.626,
+      "step": 5342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9286077845819256e-05,
+      "loss": 0.6336,
+      "step": 5343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9285813316852562e-05,
+      "loss": 0.5931,
+      "step": 5344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9285548740701928e-05,
+      "loss": 0.6641,
+      "step": 5345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9285284117368706e-05,
+      "loss": 0.6258,
+      "step": 5346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9285019446854235e-05,
+      "loss": 0.6546,
+      "step": 5347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928475472915986e-05,
+      "loss": 0.606,
+      "step": 5348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9284489964286927e-05,
+      "loss": 0.5532,
+      "step": 5349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928422515223678e-05,
+      "loss": 0.6044,
+      "step": 5350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9283960293010768e-05,
+      "loss": 0.6268,
+      "step": 5351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9283695386610234e-05,
+      "loss": 0.5396,
+      "step": 5352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9283430433036522e-05,
+      "loss": 0.5596,
+      "step": 5353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9283165432290986e-05,
+      "loss": 0.6385,
+      "step": 5354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9282900384374964e-05,
+      "loss": 0.618,
+      "step": 5355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9282635289289812e-05,
+      "loss": 0.5992,
+      "step": 5356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9282370147036868e-05,
+      "loss": 0.6205,
+      "step": 5357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9282104957617483e-05,
+      "loss": 0.6137,
+      "step": 5358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9281839721033e-05,
+      "loss": 0.6109,
+      "step": 5359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9281574437284778e-05,
+      "loss": 0.6101,
+      "step": 5360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9281309106374154e-05,
+      "loss": 0.5526,
+      "step": 5361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928104372830248e-05,
+      "loss": 0.6095,
+      "step": 5362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.928077830307111e-05,
+      "loss": 0.6503,
+      "step": 5363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9280512830681385e-05,
+      "loss": 0.6646,
+      "step": 5364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9280247311134654e-05,
+      "loss": 0.6129,
+      "step": 5365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9279981744432266e-05,
+      "loss": 0.6265,
+      "step": 5366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927971613057558e-05,
+      "loss": 0.6108,
+      "step": 5367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9279450469565932e-05,
+      "loss": 0.6594,
+      "step": 5368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9279184761404682e-05,
+      "loss": 0.6234,
+      "step": 5369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9278919006093177e-05,
+      "loss": 0.6243,
+      "step": 5370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9278653203632765e-05,
+      "loss": 0.651,
+      "step": 5371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9278387354024802e-05,
+      "loss": 0.5992,
+      "step": 5372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9278121457270633e-05,
+      "loss": 0.6225,
+      "step": 5373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9277855513371617e-05,
+      "loss": 0.6257,
+      "step": 5374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9277589522329094e-05,
+      "loss": 0.5988,
+      "step": 5375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927732348414442e-05,
+      "loss": 0.6731,
+      "step": 5376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9277057398818956e-05,
+      "loss": 0.6109,
+      "step": 5377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927679126635404e-05,
+      "loss": 0.6115,
+      "step": 5378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9276525086751033e-05,
+      "loss": 0.6401,
+      "step": 5379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9276258860011287e-05,
+      "loss": 0.6352,
+      "step": 5380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927599258613615e-05,
+      "loss": 0.5849,
+      "step": 5381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9275726265126974e-05,
+      "loss": 0.5692,
+      "step": 5382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9275459896985126e-05,
+      "loss": 0.696,
+      "step": 5383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9275193481711946e-05,
+      "loss": 0.6353,
+      "step": 5384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9274927019308788e-05,
+      "loss": 0.6464,
+      "step": 5385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9274660509777012e-05,
+      "loss": 0.6433,
+      "step": 5386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927439395311797e-05,
+      "loss": 0.6742,
+      "step": 5387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9274127349333016e-05,
+      "loss": 0.5984,
+      "step": 5388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9273860698423506e-05,
+      "loss": 0.5864,
+      "step": 5389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9273594000390792e-05,
+      "loss": 0.6138,
+      "step": 5390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9273327255236232e-05,
+      "loss": 0.5129,
+      "step": 5391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927306046296118e-05,
+      "loss": 0.5842,
+      "step": 5392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927279362356699e-05,
+      "loss": 0.6141,
+      "step": 5393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9272526737055023e-05,
+      "loss": 0.5703,
+      "step": 5394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927225980342663e-05,
+      "loss": 0.5703,
+      "step": 5395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9271992822683175e-05,
+      "loss": 0.652,
+      "step": 5396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9271725794826003e-05,
+      "loss": 0.6213,
+      "step": 5397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9271458719856478e-05,
+      "loss": 0.5754,
+      "step": 5398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9271191597775957e-05,
+      "loss": 0.5889,
+      "step": 5399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9270924428585798e-05,
+      "loss": 0.6791,
+      "step": 5400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9270657212287355e-05,
+      "loss": 0.5486,
+      "step": 5401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9270389948881985e-05,
+      "loss": 0.6406,
+      "step": 5402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.927012263837105e-05,
+      "loss": 0.6438,
+      "step": 5403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926985528075591e-05,
+      "loss": 0.6205,
+      "step": 5404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9269587876037917e-05,
+      "loss": 0.6522,
+      "step": 5405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9269320424218437e-05,
+      "loss": 0.6024,
+      "step": 5406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9269052925298822e-05,
+      "loss": 0.6696,
+      "step": 5407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9268785379280435e-05,
+      "loss": 0.6355,
+      "step": 5408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9268517786164638e-05,
+      "loss": 0.5742,
+      "step": 5409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9268250145952786e-05,
+      "loss": 0.6251,
+      "step": 5410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926798245864624e-05,
+      "loss": 0.6243,
+      "step": 5411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926771472424636e-05,
+      "loss": 0.5962,
+      "step": 5412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9267446942754508e-05,
+      "loss": 0.6507,
+      "step": 5413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9267179114172045e-05,
+      "loss": 0.6109,
+      "step": 5414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926691123850033e-05,
+      "loss": 0.5973,
+      "step": 5415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9266643315740724e-05,
+      "loss": 0.5894,
+      "step": 5416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9266375345894592e-05,
+      "loss": 0.5983,
+      "step": 5417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926610732896329e-05,
+      "loss": 0.5876,
+      "step": 5418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9265839264948187e-05,
+      "loss": 0.6816,
+      "step": 5419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9265571153850638e-05,
+      "loss": 0.6296,
+      "step": 5420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9265302995672012e-05,
+      "loss": 0.5901,
+      "step": 5421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9265034790413662e-05,
+      "loss": 0.5874,
+      "step": 5422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926476653807696e-05,
+      "loss": 0.649,
+      "step": 5423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926449823866327e-05,
+      "loss": 0.6339,
+      "step": 5424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9264229892173946e-05,
+      "loss": 0.6163,
+      "step": 5425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9263961498610356e-05,
+      "loss": 0.5766,
+      "step": 5426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9263693057973866e-05,
+      "loss": 0.6621,
+      "step": 5427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9263424570265838e-05,
+      "loss": 0.5808,
+      "step": 5428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926315603548764e-05,
+      "loss": 0.5357,
+      "step": 5429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926288745364063e-05,
+      "loss": 0.5246,
+      "step": 5430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926261882472618e-05,
+      "loss": 0.6012,
+      "step": 5431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9262350148745647e-05,
+      "loss": 0.6809,
+      "step": 5432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9262081425700404e-05,
+      "loss": 0.566,
+      "step": 5433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9261812655591807e-05,
+      "loss": 0.6167,
+      "step": 5434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926154383842123e-05,
+      "loss": 0.5215,
+      "step": 5435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9261274974190042e-05,
+      "loss": 0.619,
+      "step": 5436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9261006062899597e-05,
+      "loss": 0.5501,
+      "step": 5437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926073710455127e-05,
+      "loss": 0.6387,
+      "step": 5438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9260468099146425e-05,
+      "loss": 0.5921,
+      "step": 5439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.926019904668643e-05,
+      "loss": 0.6256,
+      "step": 5440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925992994717265e-05,
+      "loss": 0.6076,
+      "step": 5441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9259660800606457e-05,
+      "loss": 0.6466,
+      "step": 5442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9259391606989213e-05,
+      "loss": 0.5687,
+      "step": 5443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9259122366322292e-05,
+      "loss": 0.5949,
+      "step": 5444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9258853078607057e-05,
+      "loss": 0.632,
+      "step": 5445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9258583743844875e-05,
+      "loss": 0.6213,
+      "step": 5446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925831436203712e-05,
+      "loss": 0.6259,
+      "step": 5447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9258044933185158e-05,
+      "loss": 0.577,
+      "step": 5448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925777545729036e-05,
+      "loss": 0.6577,
+      "step": 5449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9257505934354092e-05,
+      "loss": 0.6835,
+      "step": 5450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9257236364377725e-05,
+      "loss": 0.6499,
+      "step": 5451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925696674736263e-05,
+      "loss": 0.5737,
+      "step": 5452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9256697083310176e-05,
+      "loss": 0.5893,
+      "step": 5453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9256427372221732e-05,
+      "loss": 0.6216,
+      "step": 5454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925615761409867e-05,
+      "loss": 0.647,
+      "step": 5455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925588780894236e-05,
+      "loss": 0.5631,
+      "step": 5456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9255617956754172e-05,
+      "loss": 0.6546,
+      "step": 5457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9255348057535482e-05,
+      "loss": 0.6516,
+      "step": 5458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925507811128766e-05,
+      "loss": 0.581,
+      "step": 5459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925480811801207e-05,
+      "loss": 0.6488,
+      "step": 5460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925453807771009e-05,
+      "loss": 0.6011,
+      "step": 5461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925426799038309e-05,
+      "loss": 0.562,
+      "step": 5462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925399785603245e-05,
+      "loss": 0.6373,
+      "step": 5463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9253727674659536e-05,
+      "loss": 0.588,
+      "step": 5464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925345744626572e-05,
+      "loss": 0.6295,
+      "step": 5465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9253187170852375e-05,
+      "loss": 0.5944,
+      "step": 5466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925291684842088e-05,
+      "loss": 0.5861,
+      "step": 5467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.92526464789726e-05,
+      "loss": 0.5126,
+      "step": 5468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9252376062508914e-05,
+      "loss": 0.6066,
+      "step": 5469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9252105599031196e-05,
+      "loss": 0.5314,
+      "step": 5470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925183508854082e-05,
+      "loss": 0.6619,
+      "step": 5471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9251564531039163e-05,
+      "loss": 0.6125,
+      "step": 5472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9251293926527593e-05,
+      "loss": 0.6267,
+      "step": 5473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9251023275007494e-05,
+      "loss": 0.6157,
+      "step": 5474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9250752576480233e-05,
+      "loss": 0.5971,
+      "step": 5475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925048183094719e-05,
+      "loss": 0.6758,
+      "step": 5476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.925021103840974e-05,
+      "loss": 0.678,
+      "step": 5477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9249940198869258e-05,
+      "loss": 0.5732,
+      "step": 5478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9249669312327124e-05,
+      "loss": 0.6154,
+      "step": 5479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9249398378784708e-05,
+      "loss": 0.6009,
+      "step": 5480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9249127398243388e-05,
+      "loss": 0.5809,
+      "step": 5481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9248856370704546e-05,
+      "loss": 0.6051,
+      "step": 5482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924858529616956e-05,
+      "loss": 0.5888,
+      "step": 5483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9248314174639798e-05,
+      "loss": 0.5974,
+      "step": 5484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9248043006116642e-05,
+      "loss": 0.6437,
+      "step": 5485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9247771790601474e-05,
+      "loss": 0.6061,
+      "step": 5486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924750052809567e-05,
+      "loss": 0.6156,
+      "step": 5487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9247229218600607e-05,
+      "loss": 0.6273,
+      "step": 5488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9246957862117662e-05,
+      "loss": 0.5979,
+      "step": 5489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9246686458648217e-05,
+      "loss": 0.651,
+      "step": 5490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924641500819365e-05,
+      "loss": 0.603,
+      "step": 5491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924614351075534e-05,
+      "loss": 0.6388,
+      "step": 5492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9245871966334667e-05,
+      "loss": 0.7207,
+      "step": 5493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9245600374933013e-05,
+      "loss": 0.6088,
+      "step": 5494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9245328736551757e-05,
+      "loss": 0.6576,
+      "step": 5495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924505705119227e-05,
+      "loss": 0.5819,
+      "step": 5496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924478531885595e-05,
+      "loss": 0.6591,
+      "step": 5497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9244513539544166e-05,
+      "loss": 0.6721,
+      "step": 5498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.92442417132583e-05,
+      "loss": 0.6508,
+      "step": 5499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9243969839999734e-05,
+      "loss": 0.64,
+      "step": 5500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9243697919769847e-05,
+      "loss": 0.6133,
+      "step": 5501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9243425952570028e-05,
+      "loss": 0.634,
+      "step": 5502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9243153938401656e-05,
+      "loss": 0.6076,
+      "step": 5503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9242881877266108e-05,
+      "loss": 0.6037,
+      "step": 5504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924260976916477e-05,
+      "loss": 0.6217,
+      "step": 5505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924233761409903e-05,
+      "loss": 0.6449,
+      "step": 5506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.924206541207026e-05,
+      "loss": 0.5869,
+      "step": 5507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9241793163079853e-05,
+      "loss": 0.6239,
+      "step": 5508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9241520867129187e-05,
+      "loss": 0.6626,
+      "step": 5509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9241248524219647e-05,
+      "loss": 0.6571,
+      "step": 5510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9240976134352617e-05,
+      "loss": 0.6299,
+      "step": 5511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.924070369752948e-05,
+      "loss": 0.6626,
+      "step": 5512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.924043121375162e-05,
+      "loss": 0.5633,
+      "step": 5513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9240158683020427e-05,
+      "loss": 0.6033,
+      "step": 5514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923988610533728e-05,
+      "loss": 0.6283,
+      "step": 5515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9239613480703564e-05,
+      "loss": 0.6028,
+      "step": 5516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9239340809120667e-05,
+      "loss": 0.6289,
+      "step": 5517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9239068090589973e-05,
+      "loss": 0.6416,
+      "step": 5518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9238795325112867e-05,
+      "loss": 0.5853,
+      "step": 5519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923852251269074e-05,
+      "loss": 0.5645,
+      "step": 5520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9238249653324975e-05,
+      "loss": 0.5738,
+      "step": 5521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9237976747016952e-05,
+      "loss": 0.645,
+      "step": 5522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923770379376807e-05,
+      "loss": 0.574,
+      "step": 5523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9237430793579705e-05,
+      "loss": 0.6167,
+      "step": 5524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923715774645325e-05,
+      "loss": 0.6043,
+      "step": 5525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923688465239009e-05,
+      "loss": 0.6494,
+      "step": 5526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9236611511391617e-05,
+      "loss": 0.6055,
+      "step": 5527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9236338323459217e-05,
+      "loss": 0.5796,
+      "step": 5528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9236065088594274e-05,
+      "loss": 0.5667,
+      "step": 5529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923579180679818e-05,
+      "loss": 0.5932,
+      "step": 5530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9235518478072322e-05,
+      "loss": 0.5825,
+      "step": 5531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923524510241809e-05,
+      "loss": 0.6006,
+      "step": 5532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9234971679836872e-05,
+      "loss": 0.6365,
+      "step": 5533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9234698210330063e-05,
+      "loss": 0.623,
+      "step": 5534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9234424693899046e-05,
+      "loss": 0.6314,
+      "step": 5535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923415113054521e-05,
+      "loss": 0.6057,
+      "step": 5536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923387752026995e-05,
+      "loss": 0.6491,
+      "step": 5537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9233603863074653e-05,
+      "loss": 0.6163,
+      "step": 5538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923333015896071e-05,
+      "loss": 0.5933,
+      "step": 5539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9233056407929516e-05,
+      "loss": 0.6619,
+      "step": 5540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9232782609982456e-05,
+      "loss": 0.5981,
+      "step": 5541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9232508765120922e-05,
+      "loss": 0.5944,
+      "step": 5542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.923223487334631e-05,
+      "loss": 0.6065,
+      "step": 5543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9231960934660005e-05,
+      "loss": 0.6285,
+      "step": 5544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9231686949063405e-05,
+      "loss": 0.5726,
+      "step": 5545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9231412916557903e-05,
+      "loss": 0.6301,
+      "step": 5546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9231138837144885e-05,
+      "loss": 0.6271,
+      "step": 5547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9230864710825744e-05,
+      "loss": 0.6004,
+      "step": 5548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9230590537601882e-05,
+      "loss": 0.608,
+      "step": 5549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9230316317474686e-05,
+      "loss": 0.6132,
+      "step": 5550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9230042050445544e-05,
+      "loss": 0.6105,
+      "step": 5551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9229767736515862e-05,
+      "loss": 0.6243,
+      "step": 5552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922949337568702e-05,
+      "loss": 0.623,
+      "step": 5553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9229218967960427e-05,
+      "loss": 0.6157,
+      "step": 5554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9228944513337465e-05,
+      "loss": 0.6404,
+      "step": 5555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922867001181953e-05,
+      "loss": 0.61,
+      "step": 5556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9228395463408026e-05,
+      "loss": 0.6243,
+      "step": 5557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922812086810434e-05,
+      "loss": 0.6511,
+      "step": 5558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922784622590987e-05,
+      "loss": 0.6237,
+      "step": 5559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9227571536826005e-05,
+      "loss": 0.6365,
+      "step": 5560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922729680085415e-05,
+      "loss": 0.6465,
+      "step": 5561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9227022017995704e-05,
+      "loss": 0.673,
+      "step": 5562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922674718825205e-05,
+      "loss": 0.6115,
+      "step": 5563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922647231162459e-05,
+      "loss": 0.6604,
+      "step": 5564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9226197388114726e-05,
+      "loss": 0.6243,
+      "step": 5565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922592241772385e-05,
+      "loss": 0.6563,
+      "step": 5566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9225647400453362e-05,
+      "loss": 0.6461,
+      "step": 5567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9225372336304654e-05,
+      "loss": 0.6331,
+      "step": 5568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922509722527913e-05,
+      "loss": 0.6186,
+      "step": 5569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9224822067378187e-05,
+      "loss": 0.6058,
+      "step": 5570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9224546862603216e-05,
+      "loss": 0.5992,
+      "step": 5571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9224271610955628e-05,
+      "loss": 0.6343,
+      "step": 5572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922399631243681e-05,
+      "loss": 0.601,
+      "step": 5573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9223720967048166e-05,
+      "loss": 0.5926,
+      "step": 5574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9223445574791092e-05,
+      "loss": 0.5928,
+      "step": 5575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9223170135666994e-05,
+      "loss": 0.5622,
+      "step": 5576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9222894649677264e-05,
+      "loss": 0.6792,
+      "step": 5577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922261911682331e-05,
+      "loss": 0.6012,
+      "step": 5578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9222343537106524e-05,
+      "loss": 0.6655,
+      "step": 5579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922206791052831e-05,
+      "loss": 0.6111,
+      "step": 5580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9221792237090066e-05,
+      "loss": 0.5981,
+      "step": 5581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9221516516793198e-05,
+      "loss": 0.5337,
+      "step": 5582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92212407496391e-05,
+      "loss": 0.6267,
+      "step": 5583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922096493562918e-05,
+      "loss": 0.6265,
+      "step": 5584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9220689074764837e-05,
+      "loss": 0.6066,
+      "step": 5585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9220413167047473e-05,
+      "loss": 0.6427,
+      "step": 5586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.922013721247849e-05,
+      "loss": 0.6652,
+      "step": 5587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9219861211059284e-05,
+      "loss": 0.6193,
+      "step": 5588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921958516279127e-05,
+      "loss": 0.536,
+      "step": 5589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921930906767584e-05,
+      "loss": 0.6561,
+      "step": 5590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92190329257144e-05,
+      "loss": 0.6142,
+      "step": 5591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9218756736908357e-05,
+      "loss": 0.6258,
+      "step": 5592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921848050125911e-05,
+      "loss": 0.6183,
+      "step": 5593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921820421876806e-05,
+      "loss": 0.6693,
+      "step": 5594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921792788943662e-05,
+      "loss": 0.5936,
+      "step": 5595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9217651513266187e-05,
+      "loss": 0.5945,
+      "step": 5596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9217375090258166e-05,
+      "loss": 0.6234,
+      "step": 5597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9217098620413965e-05,
+      "loss": 0.5822,
+      "step": 5598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9216822103734986e-05,
+      "loss": 0.566,
+      "step": 5599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921654554022263e-05,
+      "loss": 0.6523,
+      "step": 5600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9216268929878313e-05,
+      "loss": 0.6156,
+      "step": 5601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921599227270343e-05,
+      "loss": 0.587,
+      "step": 5602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9215715568699392e-05,
+      "loss": 0.585,
+      "step": 5603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921543881786761e-05,
+      "loss": 0.609,
+      "step": 5604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9215162020209476e-05,
+      "loss": 0.6675,
+      "step": 5605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921488517572641e-05,
+      "loss": 0.6198,
+      "step": 5606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9214608284419812e-05,
+      "loss": 0.6281,
+      "step": 5607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9214331346291092e-05,
+      "loss": 0.6656,
+      "step": 5608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9214054361341657e-05,
+      "loss": 0.5793,
+      "step": 5609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9213777329572913e-05,
+      "loss": 0.5825,
+      "step": 5610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9213500250986262e-05,
+      "loss": 0.6165,
+      "step": 5611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9213223125583128e-05,
+      "loss": 0.5925,
+      "step": 5612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92129459533649e-05,
+      "loss": 0.6085,
+      "step": 5613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9212668734333e-05,
+      "loss": 0.5911,
+      "step": 5614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9212391468488832e-05,
+      "loss": 0.5978,
+      "step": 5615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92121141558338e-05,
+      "loss": 0.5631,
+      "step": 5616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9211836796369325e-05,
+      "loss": 0.6324,
+      "step": 5617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9211559390096808e-05,
+      "loss": 0.6213,
+      "step": 5618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921128193701766e-05,
+      "loss": 0.6322,
+      "step": 5619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9211004437133288e-05,
+      "loss": 0.6525,
+      "step": 5620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9210726890445107e-05,
+      "loss": 0.6196,
+      "step": 5621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9210449296954524e-05,
+      "loss": 0.6338,
+      "step": 5622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.921017165666295e-05,
+      "loss": 0.6144,
+      "step": 5623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92098939695718e-05,
+      "loss": 0.5714,
+      "step": 5624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920961623568248e-05,
+      "loss": 0.5687,
+      "step": 5625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.92093384549964e-05,
+      "loss": 0.5917,
+      "step": 5626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9209060627514977e-05,
+      "loss": 0.5545,
+      "step": 5627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9208782753239622e-05,
+      "loss": 0.6317,
+      "step": 5628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920850483217174e-05,
+      "loss": 0.5738,
+      "step": 5629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920822686431275e-05,
+      "loss": 0.6011,
+      "step": 5630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9207948849664063e-05,
+      "loss": 0.5763,
+      "step": 5631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9207670788227092e-05,
+      "loss": 0.6713,
+      "step": 5632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9207392680003244e-05,
+      "loss": 0.6009,
+      "step": 5633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9207114524993944e-05,
+      "loss": 0.6371,
+      "step": 5634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9206836323200595e-05,
+      "loss": 0.6451,
+      "step": 5635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9206558074624614e-05,
+      "loss": 0.6405,
+      "step": 5636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920627977926742e-05,
+      "loss": 0.587,
+      "step": 5637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920600143713042e-05,
+      "loss": 0.6065,
+      "step": 5638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9205723048215027e-05,
+      "loss": 0.5912,
+      "step": 5639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9205444612522665e-05,
+      "loss": 0.6233,
+      "step": 5640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9205166130054737e-05,
+      "loss": 0.6583,
+      "step": 5641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9204887600812668e-05,
+      "loss": 0.6361,
+      "step": 5642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9204609024797866e-05,
+      "loss": 0.5611,
+      "step": 5643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920433040201175e-05,
+      "loss": 0.5972,
+      "step": 5644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920405173245574e-05,
+      "loss": 0.5793,
+      "step": 5645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9203773016131248e-05,
+      "loss": 0.6335,
+      "step": 5646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9203494253039685e-05,
+      "loss": 0.6327,
+      "step": 5647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920321544318247e-05,
+      "loss": 0.5331,
+      "step": 5648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920293658656103e-05,
+      "loss": 0.617,
+      "step": 5649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920265768317677e-05,
+      "loss": 0.5645,
+      "step": 5650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9202378733031112e-05,
+      "loss": 0.6378,
+      "step": 5651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.920209973612547e-05,
+      "loss": 0.6211,
+      "step": 5652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9201820692461267e-05,
+      "loss": 0.6367,
+      "step": 5653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9201541602039917e-05,
+      "loss": 0.6058,
+      "step": 5654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9201262464862843e-05,
+      "loss": 0.5279,
+      "step": 5655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9200983280931455e-05,
+      "loss": 0.6368,
+      "step": 5656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9200704050247175e-05,
+      "loss": 0.5991,
+      "step": 5657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9200424772811427e-05,
+      "loss": 0.6338,
+      "step": 5658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9200145448625626e-05,
+      "loss": 0.6312,
+      "step": 5659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9199866077691193e-05,
+      "loss": 0.6085,
+      "step": 5660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9199586660009543e-05,
+      "loss": 0.6151,
+      "step": 5661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91993071955821e-05,
+      "loss": 0.6436,
+      "step": 5662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9199027684410283e-05,
+      "loss": 0.5751,
+      "step": 5663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9198748126495512e-05,
+      "loss": 0.6184,
+      "step": 5664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9198468521839206e-05,
+      "loss": 0.5967,
+      "step": 5665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9198188870442792e-05,
+      "loss": 0.6218,
+      "step": 5666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9197909172307683e-05,
+      "loss": 0.6159,
+      "step": 5667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9197629427435305e-05,
+      "loss": 0.5432,
+      "step": 5668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9197349635827078e-05,
+      "loss": 0.6277,
+      "step": 5669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9197069797484425e-05,
+      "loss": 0.5709,
+      "step": 5670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9196789912408765e-05,
+      "loss": 0.6587,
+      "step": 5671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9196509980601525e-05,
+      "loss": 0.575,
+      "step": 5672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919623000206412e-05,
+      "loss": 0.605,
+      "step": 5673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919594997679798e-05,
+      "loss": 0.5561,
+      "step": 5674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9195669904804526e-05,
+      "loss": 0.6018,
+      "step": 5675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9195389786085175e-05,
+      "loss": 0.6252,
+      "step": 5676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9195109620641362e-05,
+      "loss": 0.6333,
+      "step": 5677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9194829408474504e-05,
+      "loss": 0.6037,
+      "step": 5678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919454914958602e-05,
+      "loss": 0.6233,
+      "step": 5679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919426884397734e-05,
+      "loss": 0.6362,
+      "step": 5680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919398849164989e-05,
+      "loss": 0.6186,
+      "step": 5681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919370809260509e-05,
+      "loss": 0.6653,
+      "step": 5682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9193427646844366e-05,
+      "loss": 0.5848,
+      "step": 5683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919314715436915e-05,
+      "loss": 0.6017,
+      "step": 5684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919286661518085e-05,
+      "loss": 0.5997,
+      "step": 5685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919258602928091e-05,
+      "loss": 0.6216,
+      "step": 5686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9192305396670746e-05,
+      "loss": 0.6214,
+      "step": 5687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919202471735179e-05,
+      "loss": 0.6769,
+      "step": 5688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919174399132546e-05,
+      "loss": 0.6299,
+      "step": 5689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9191463218593186e-05,
+      "loss": 0.6659,
+      "step": 5690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91911823991564e-05,
+      "loss": 0.5424,
+      "step": 5691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919090153301652e-05,
+      "loss": 0.656,
+      "step": 5692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.919062062017498e-05,
+      "loss": 0.6222,
+      "step": 5693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9190339660633205e-05,
+      "loss": 0.6257,
+      "step": 5694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9190058654392623e-05,
+      "loss": 0.6163,
+      "step": 5695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918977760145466e-05,
+      "loss": 0.5871,
+      "step": 5696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9189496501820745e-05,
+      "loss": 0.6521,
+      "step": 5697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918921535549231e-05,
+      "loss": 0.5857,
+      "step": 5698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918893416247078e-05,
+      "loss": 0.6558,
+      "step": 5699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918865292275758e-05,
+      "loss": 0.636,
+      "step": 5700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918837163635415e-05,
+      "loss": 0.5825,
+      "step": 5701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918809030326191e-05,
+      "loss": 0.586,
+      "step": 5702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918780892348229e-05,
+      "loss": 0.662,
+      "step": 5703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9187527497016724e-05,
+      "loss": 0.555,
+      "step": 5704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9187246023866645e-05,
+      "loss": 0.5891,
+      "step": 5705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918696450403347e-05,
+      "loss": 0.5611,
+      "step": 5706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9186682937518642e-05,
+      "loss": 0.5997,
+      "step": 5707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918640132432359e-05,
+      "loss": 0.6516,
+      "step": 5708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9186119664449738e-05,
+      "loss": 0.5611,
+      "step": 5709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9185837957898525e-05,
+      "loss": 0.628,
+      "step": 5710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9185556204671374e-05,
+      "loss": 0.6126,
+      "step": 5711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9185274404769723e-05,
+      "loss": 0.6154,
+      "step": 5712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918499255819501e-05,
+      "loss": 0.6056,
+      "step": 5713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918471066494865e-05,
+      "loss": 0.6187,
+      "step": 5714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918442872503209e-05,
+      "loss": 0.5681,
+      "step": 5715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918414673844676e-05,
+      "loss": 0.5921,
+      "step": 5716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9183864705194088e-05,
+      "loss": 0.6041,
+      "step": 5717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918358262527551e-05,
+      "loss": 0.6086,
+      "step": 5718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918330049869246e-05,
+      "loss": 0.6684,
+      "step": 5719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918301832544637e-05,
+      "loss": 0.5848,
+      "step": 5720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9182736105538676e-05,
+      "loss": 0.5863,
+      "step": 5721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918245383897081e-05,
+      "loss": 0.6028,
+      "step": 5722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9182171525744207e-05,
+      "loss": 0.5324,
+      "step": 5723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91818891658603e-05,
+      "loss": 0.6469,
+      "step": 5724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9181606759320527e-05,
+      "loss": 0.6564,
+      "step": 5725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9181324306126318e-05,
+      "loss": 0.5773,
+      "step": 5726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9181041806279113e-05,
+      "loss": 0.5824,
+      "step": 5727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9180759259780345e-05,
+      "loss": 0.6239,
+      "step": 5728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.918047666663145e-05,
+      "loss": 0.6283,
+      "step": 5729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9180194026833868e-05,
+      "loss": 0.6664,
+      "step": 5730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917991134038903e-05,
+      "loss": 0.6505,
+      "step": 5731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917962860729837e-05,
+      "loss": 0.609,
+      "step": 5732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9179345827563333e-05,
+      "loss": 0.5996,
+      "step": 5733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917906300118535e-05,
+      "loss": 0.5945,
+      "step": 5734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9178780128165862e-05,
+      "loss": 0.6398,
+      "step": 5735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9178497208506302e-05,
+      "loss": 0.602,
+      "step": 5736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917821424220811e-05,
+      "loss": 0.5841,
+      "step": 5737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9177931229272723e-05,
+      "loss": 0.6174,
+      "step": 5738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917764816970158e-05,
+      "loss": 0.6311,
+      "step": 5739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917736506349612e-05,
+      "loss": 0.6593,
+      "step": 5740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9177081910657777e-05,
+      "loss": 0.6266,
+      "step": 5741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9176798711187996e-05,
+      "loss": 0.595,
+      "step": 5742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9176515465088213e-05,
+      "loss": 0.6136,
+      "step": 5743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9176232172359867e-05,
+      "loss": 0.6485,
+      "step": 5744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91759488330044e-05,
+      "loss": 0.6106,
+      "step": 5745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9175665447023245e-05,
+      "loss": 0.6576,
+      "step": 5746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9175382014417853e-05,
+      "loss": 0.6255,
+      "step": 5747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9175098535189653e-05,
+      "loss": 0.6054,
+      "step": 5748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9174815009340093e-05,
+      "loss": 0.5498,
+      "step": 5749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9174531436870608e-05,
+      "loss": 0.6253,
+      "step": 5750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9174247817782642e-05,
+      "loss": 0.6464,
+      "step": 5751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917396415207764e-05,
+      "loss": 0.6019,
+      "step": 5752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9173680439757034e-05,
+      "loss": 0.6439,
+      "step": 5753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9173396680822275e-05,
+      "loss": 0.607,
+      "step": 5754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91731128752748e-05,
+      "loss": 0.6344,
+      "step": 5755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917282902311605e-05,
+      "loss": 0.6191,
+      "step": 5756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917254512434747e-05,
+      "loss": 0.6556,
+      "step": 5757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9172261178970502e-05,
+      "loss": 0.5801,
+      "step": 5758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917197718698659e-05,
+      "loss": 0.6375,
+      "step": 5759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.917169314839718e-05,
+      "loss": 0.6114,
+      "step": 5760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9171409063203703e-05,
+      "loss": 0.6064,
+      "step": 5761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9171124931407616e-05,
+      "loss": 0.5885,
+      "step": 5762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9170840753010354e-05,
+      "loss": 0.5707,
+      "step": 5763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9170556528013365e-05,
+      "loss": 0.6478,
+      "step": 5764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9170272256418095e-05,
+      "loss": 0.6079,
+      "step": 5765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9169987938225984e-05,
+      "loss": 0.6176,
+      "step": 5766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916970357343848e-05,
+      "loss": 0.5956,
+      "step": 5767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9169419162057025e-05,
+      "loss": 0.6003,
+      "step": 5768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9169134704083068e-05,
+      "loss": 0.6151,
+      "step": 5769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916885019951805e-05,
+      "loss": 0.6354,
+      "step": 5770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9168565648363418e-05,
+      "loss": 0.5564,
+      "step": 5771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9168281050620622e-05,
+      "loss": 0.6129,
+      "step": 5772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9167996406291106e-05,
+      "loss": 0.6066,
+      "step": 5773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9167711715376312e-05,
+      "loss": 0.6466,
+      "step": 5774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916742697787769e-05,
+      "loss": 0.6152,
+      "step": 5775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9167142193796686e-05,
+      "loss": 0.6594,
+      "step": 5776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916685736313475e-05,
+      "loss": 0.6471,
+      "step": 5777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9166572485893325e-05,
+      "loss": 0.6263,
+      "step": 5778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9166287562073863e-05,
+      "loss": 0.5717,
+      "step": 5779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916600259167781e-05,
+      "loss": 0.6478,
+      "step": 5780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916571757470661e-05,
+      "loss": 0.6145,
+      "step": 5781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916543251116172e-05,
+      "loss": 0.5792,
+      "step": 5782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9165147401044577e-05,
+      "loss": 0.6139,
+      "step": 5783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9164862244356638e-05,
+      "loss": 0.5726,
+      "step": 5784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9164577041099353e-05,
+      "loss": 0.6621,
+      "step": 5785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9164291791274165e-05,
+      "loss": 0.576,
+      "step": 5786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9164006494882524e-05,
+      "loss": 0.5998,
+      "step": 5787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9163721151925888e-05,
+      "loss": 0.6142,
+      "step": 5788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.91634357624057e-05,
+      "loss": 0.6064,
+      "step": 5789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9163150326323407e-05,
+      "loss": 0.6213,
+      "step": 5790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9162864843680466e-05,
+      "loss": 0.6046,
+      "step": 5791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9162579314478327e-05,
+      "loss": 0.5958,
+      "step": 5792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916229373871844e-05,
+      "loss": 0.6428,
+      "step": 5793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9162008116402248e-05,
+      "loss": 0.5487,
+      "step": 5794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9161722447531216e-05,
+      "loss": 0.5684,
+      "step": 5795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.916143673210679e-05,
+      "loss": 0.6292,
+      "step": 5796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9161150970130418e-05,
+      "loss": 0.6155,
+      "step": 5797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9160865161603555e-05,
+      "loss": 0.6621,
+      "step": 5798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9160579306527654e-05,
+      "loss": 0.6592,
+      "step": 5799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9160293404904166e-05,
+      "loss": 0.6092,
+      "step": 5800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9160007456734543e-05,
+      "loss": 0.6932,
+      "step": 5801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9159721462020243e-05,
+      "loss": 0.6251,
+      "step": 5802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9159435420762716e-05,
+      "loss": 0.6212,
+      "step": 5803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9159149332963412e-05,
+      "loss": 0.6192,
+      "step": 5804
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915886319862379e-05,
+      "loss": 0.6217,
+      "step": 5805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.91585770177453e-05,
+      "loss": 0.6008,
+      "step": 5806
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.91582907903294e-05,
+      "loss": 0.6109,
+      "step": 5807
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915800451637754e-05,
+      "loss": 0.5808,
+      "step": 5808
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9157718195891176e-05,
+      "loss": 0.6309,
+      "step": 5809
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9157431828871767e-05,
+      "loss": 0.6002,
+      "step": 5810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9157145415320765e-05,
+      "loss": 0.5834,
+      "step": 5811
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9156858955239625e-05,
+      "loss": 0.5367,
+      "step": 5812
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9156572448629803e-05,
+      "loss": 0.5762,
+      "step": 5813
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9156285895492753e-05,
+      "loss": 0.5323,
+      "step": 5814
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915599929582993e-05,
+      "loss": 0.5235,
+      "step": 5815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9155712649642798e-05,
+      "loss": 0.5497,
+      "step": 5816
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915542595693281e-05,
+      "loss": 0.6075,
+      "step": 5817
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915513921770142e-05,
+      "loss": 0.645,
+      "step": 5818
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9154852431950086e-05,
+      "loss": 0.6513,
+      "step": 5819
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9154565599680267e-05,
+      "loss": 0.6715,
+      "step": 5820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9154278720893414e-05,
+      "loss": 0.5947,
+      "step": 5821
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9153991795590995e-05,
+      "loss": 0.5966,
+      "step": 5822
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915370482377446e-05,
+      "loss": 0.5296,
+      "step": 5823
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915341780544527e-05,
+      "loss": 0.6307,
+      "step": 5824
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9153130740604887e-05,
+      "loss": 0.575,
+      "step": 5825
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9152843629254764e-05,
+      "loss": 0.6424,
+      "step": 5826
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915255647139636e-05,
+      "loss": 0.6196,
+      "step": 5827
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915226926703114e-05,
+      "loss": 0.5917,
+      "step": 5828
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9151982016160555e-05,
+      "loss": 0.6034,
+      "step": 5829
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9151694718786072e-05,
+      "loss": 0.6244,
+      "step": 5830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.915140737490915e-05,
+      "loss": 0.6174,
+      "step": 5831
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9151119984531242e-05,
+      "loss": 0.5833,
+      "step": 5832
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9150832547653815e-05,
+      "loss": 0.6715,
+      "step": 5833
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9150545064278327e-05,
+      "loss": 0.5868,
+      "step": 5834
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9150257534406244e-05,
+      "loss": 0.6036,
+      "step": 5835
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9149969958039017e-05,
+      "loss": 0.6372,
+      "step": 5836
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9149682335178118e-05,
+      "loss": 0.5704,
+      "step": 5837
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9149394665824997e-05,
+      "loss": 0.6145,
+      "step": 5838
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9149106949981127e-05,
+      "loss": 0.5337,
+      "step": 5839
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9148819187647966e-05,
+      "loss": 0.5735,
+      "step": 5840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9148531378826974e-05,
+      "loss": 0.6648,
+      "step": 5841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9148243523519618e-05,
+      "loss": 0.6159,
+      "step": 5842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914795562172735e-05,
+      "loss": 0.6011,
+      "step": 5843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9147667673451647e-05,
+      "loss": 0.5684,
+      "step": 5844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9147379678693963e-05,
+      "loss": 0.6547,
+      "step": 5845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9147091637455763e-05,
+      "loss": 0.5993,
+      "step": 5846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9146803549738514e-05,
+      "loss": 0.6444,
+      "step": 5847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9146515415543672e-05,
+      "loss": 0.6849,
+      "step": 5848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9146227234872713e-05,
+      "loss": 0.5754,
+      "step": 5849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914593900772709e-05,
+      "loss": 0.5966,
+      "step": 5850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914565073410828e-05,
+      "loss": 0.5986,
+      "step": 5851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9145362414017732e-05,
+      "loss": 0.627,
+      "step": 5852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914507404745692e-05,
+      "loss": 0.6619,
+      "step": 5853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914478563442731e-05,
+      "loss": 0.5666,
+      "step": 5854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914449717493037e-05,
+      "loss": 0.6247,
+      "step": 5855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9144208668967557e-05,
+      "loss": 0.5923,
+      "step": 5856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9143920116540342e-05,
+      "loss": 0.6001,
+      "step": 5857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9143631517650195e-05,
+      "loss": 0.6083,
+      "step": 5858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9143342872298573e-05,
+      "loss": 0.6145,
+      "step": 5859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914305418048695e-05,
+      "loss": 0.5492,
+      "step": 5860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9142765442216796e-05,
+      "loss": 0.6215,
+      "step": 5861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9142476657489565e-05,
+      "loss": 0.6206,
+      "step": 5862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9142187826306738e-05,
+      "loss": 0.5962,
+      "step": 5863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9141898948669774e-05,
+      "loss": 0.6292,
+      "step": 5864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9141610024580148e-05,
+      "loss": 0.5846,
+      "step": 5865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914132105403932e-05,
+      "loss": 0.5837,
+      "step": 5866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9141032037048766e-05,
+      "loss": 0.6659,
+      "step": 5867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914074297360995e-05,
+      "loss": 0.5679,
+      "step": 5868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.914045386372434e-05,
+      "loss": 0.6579,
+      "step": 5869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9140164707393408e-05,
+      "loss": 0.6419,
+      "step": 5870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9139875504618626e-05,
+      "loss": 0.598,
+      "step": 5871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9139586255401458e-05,
+      "loss": 0.6203,
+      "step": 5872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9139296959743373e-05,
+      "loss": 0.637,
+      "step": 5873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9139007617645844e-05,
+      "loss": 0.5539,
+      "step": 5874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9138718229110343e-05,
+      "loss": 0.6619,
+      "step": 5875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9138428794138338e-05,
+      "loss": 0.59,
+      "step": 5876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9138139312731295e-05,
+      "loss": 0.5846,
+      "step": 5877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9137849784890696e-05,
+      "loss": 0.6254,
+      "step": 5878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9137560210618004e-05,
+      "loss": 0.6187,
+      "step": 5879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9137270589914695e-05,
+      "loss": 0.6295,
+      "step": 5880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9136980922782235e-05,
+      "loss": 0.6417,
+      "step": 5881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.91366912092221e-05,
+      "loss": 0.6586,
+      "step": 5882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.913640144923576e-05,
+      "loss": 0.6338,
+      "step": 5883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.913611164282469e-05,
+      "loss": 0.6105,
+      "step": 5884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9135821789990362e-05,
+      "loss": 0.6526,
+      "step": 5885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9135531890734245e-05,
+      "loss": 0.6427,
+      "step": 5886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9135241945057815e-05,
+      "loss": 0.6243,
+      "step": 5887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9134951952962547e-05,
+      "loss": 0.5736,
+      "step": 5888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9134661914449914e-05,
+      "loss": 0.6209,
+      "step": 5889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9134371829521386e-05,
+      "loss": 0.5728,
+      "step": 5890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.913408169817844e-05,
+      "loss": 0.6214,
+      "step": 5891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9133791520422554e-05,
+      "loss": 0.6024,
+      "step": 5892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9133501296255193e-05,
+      "loss": 0.6465,
+      "step": 5893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.913321102567784e-05,
+      "loss": 0.5899,
+      "step": 5894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9132920708691968e-05,
+      "loss": 0.6296,
+      "step": 5895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9132630345299045e-05,
+      "loss": 0.591,
+      "step": 5896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9132339935500558e-05,
+      "loss": 0.5854,
+      "step": 5897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9132049479297977e-05,
+      "loss": 0.5458,
+      "step": 5898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9131758976692777e-05,
+      "loss": 0.6905,
+      "step": 5899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9131468427686434e-05,
+      "loss": 0.6629,
+      "step": 5900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9131177832280426e-05,
+      "loss": 0.5907,
+      "step": 5901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9130887190476234e-05,
+      "loss": 0.6044,
+      "step": 5902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9130596502275323e-05,
+      "loss": 0.6452,
+      "step": 5903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9130305767679183e-05,
+      "loss": 0.6642,
+      "step": 5904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9130014986689283e-05,
+      "loss": 0.6323,
+      "step": 5905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9129724159307104e-05,
+      "loss": 0.6087,
+      "step": 5906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912943328553412e-05,
+      "loss": 0.6491,
+      "step": 5907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9129142365371813e-05,
+      "loss": 0.6227,
+      "step": 5908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912885139882166e-05,
+      "loss": 0.604,
+      "step": 5909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9128560385885143e-05,
+      "loss": 0.6184,
+      "step": 5910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9128269326563734e-05,
+      "loss": 0.6185,
+      "step": 5911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9127978220858915e-05,
+      "loss": 0.6278,
+      "step": 5912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9127687068772168e-05,
+      "loss": 0.6736,
+      "step": 5913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9127395870304966e-05,
+      "loss": 0.6504,
+      "step": 5914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9127104625458797e-05,
+      "loss": 0.6212,
+      "step": 5915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9126813334235132e-05,
+      "loss": 0.6138,
+      "step": 5916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9126521996635457e-05,
+      "loss": 0.6182,
+      "step": 5917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912623061266125e-05,
+      "loss": 0.5633,
+      "step": 5918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9125939182314e-05,
+      "loss": 0.5907,
+      "step": 5919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912564770559517e-05,
+      "loss": 0.5787,
+      "step": 5920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9125356182506257e-05,
+      "loss": 0.6398,
+      "step": 5921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9125064613048734e-05,
+      "loss": 0.6077,
+      "step": 5922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9124772997224086e-05,
+      "loss": 0.6421,
+      "step": 5923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9124481335033797e-05,
+      "loss": 0.5685,
+      "step": 5924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9124189626479343e-05,
+      "loss": 0.5521,
+      "step": 5925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912389787156221e-05,
+      "loss": 0.6206,
+      "step": 5926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912360607028388e-05,
+      "loss": 0.6659,
+      "step": 5927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9123314222645838e-05,
+      "loss": 0.5933,
+      "step": 5928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912302232864956e-05,
+      "loss": 0.6236,
+      "step": 5929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9122730388296537e-05,
+      "loss": 0.5974,
+      "step": 5930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9122438401588247e-05,
+      "loss": 0.6222,
+      "step": 5931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9122146368526177e-05,
+      "loss": 0.584,
+      "step": 5932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912185428911181e-05,
+      "loss": 0.6755,
+      "step": 5933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912156216334663e-05,
+      "loss": 0.6144,
+      "step": 5934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9121269991232122e-05,
+      "loss": 0.6388,
+      "step": 5935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912097777276977e-05,
+      "loss": 0.5596,
+      "step": 5936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9120685507961058e-05,
+      "loss": 0.5981,
+      "step": 5937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.912039319680747e-05,
+      "loss": 0.6499,
+      "step": 5938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9120100839310494e-05,
+      "loss": 0.6331,
+      "step": 5939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9119808435471613e-05,
+      "loss": 0.6,
+      "step": 5940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.911951598529232e-05,
+      "loss": 0.6588,
+      "step": 5941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.911922348877409e-05,
+      "loss": 0.5749,
+      "step": 5942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9118930945918416e-05,
+      "loss": 0.6616,
+      "step": 5943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9118638356726785e-05,
+      "loss": 0.6329,
+      "step": 5944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9118345721200683e-05,
+      "loss": 0.7189,
+      "step": 5945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9118053039341594e-05,
+      "loss": 0.6156,
+      "step": 5946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9117760311151006e-05,
+      "loss": 0.6339,
+      "step": 5947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.911746753663041e-05,
+      "loss": 0.6523,
+      "step": 5948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9117174715781292e-05,
+      "loss": 0.6684,
+      "step": 5949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9116881848605137e-05,
+      "loss": 0.5978,
+      "step": 5950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.9116588935103435e-05,
+      "loss": 0.6257,
+      "step": 5951
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9116295975277674e-05,
+      "loss": 0.6252,
+      "step": 5952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9116002969129345e-05,
+      "loss": 0.6398,
+      "step": 5953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.911570991665994e-05,
+      "loss": 0.5941,
+      "step": 5954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9115416817870935e-05,
+      "loss": 0.6274,
+      "step": 5955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.911512367276383e-05,
+      "loss": 0.5538,
+      "step": 5956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9114830481340114e-05,
+      "loss": 0.6349,
+      "step": 5957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9114537243601274e-05,
+      "loss": 0.6126,
+      "step": 5958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.91142439595488e-05,
+      "loss": 0.64,
+      "step": 5959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9113950629184185e-05,
+      "loss": 0.5725,
+      "step": 5960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9113657252508916e-05,
+      "loss": 0.5786,
+      "step": 5961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9113363829524486e-05,
+      "loss": 0.6384,
+      "step": 5962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9113070360232385e-05,
+      "loss": 0.5776,
+      "step": 5963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9112776844634108e-05,
+      "loss": 0.692,
+      "step": 5964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9112483282731138e-05,
+      "loss": 0.6106,
+      "step": 5965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9112189674524973e-05,
+      "loss": 0.6006,
+      "step": 5966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9111896020017104e-05,
+      "loss": 0.6065,
+      "step": 5967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9111602319209025e-05,
+      "loss": 0.6622,
+      "step": 5968
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9111308572102224e-05,
+      "loss": 0.5823,
+      "step": 5969
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9111014778698197e-05,
+      "loss": 0.6599,
+      "step": 5970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9110720938998436e-05,
+      "loss": 0.5885,
+      "step": 5971
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.911042705300443e-05,
+      "loss": 0.6077,
+      "step": 5972
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.911013312071768e-05,
+      "loss": 0.6485,
+      "step": 5973
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9109839142139672e-05,
+      "loss": 0.6303,
+      "step": 5974
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910954511727191e-05,
+      "loss": 0.5808,
+      "step": 5975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9109251046115875e-05,
+      "loss": 0.6148,
+      "step": 5976
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910895692867307e-05,
+      "loss": 0.6551,
+      "step": 5977
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9108662764944984e-05,
+      "loss": 0.6555,
+      "step": 5978
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910836855493312e-05,
+      "loss": 0.6256,
+      "step": 5979
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9108074298638965e-05,
+      "loss": 0.597,
+      "step": 5980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9107779996064016e-05,
+      "loss": 0.6426,
+      "step": 5981
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9107485647209773e-05,
+      "loss": 0.6072,
+      "step": 5982
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9107191252077724e-05,
+      "loss": 0.6824,
+      "step": 5983
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9106896810669373e-05,
+      "loss": 0.5948,
+      "step": 5984
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9106602322986212e-05,
+      "loss": 0.6124,
+      "step": 5985
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9106307789029736e-05,
+      "loss": 0.6537,
+      "step": 5986
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9106013208801443e-05,
+      "loss": 0.6039,
+      "step": 5987
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910571858230283e-05,
+      "loss": 0.6154,
+      "step": 5988
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9105423909535395e-05,
+      "loss": 0.5752,
+      "step": 5989
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9105129190500634e-05,
+      "loss": 0.6151,
+      "step": 5990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9104834425200047e-05,
+      "loss": 0.6099,
+      "step": 5991
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9104539613635126e-05,
+      "loss": 0.6205,
+      "step": 5992
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9104244755807376e-05,
+      "loss": 0.6572,
+      "step": 5993
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9103949851718292e-05,
+      "loss": 0.6209,
+      "step": 5994
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910365490136937e-05,
+      "loss": 0.6146,
+      "step": 5995
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9103359904762114e-05,
+      "loss": 0.5843,
+      "step": 5996
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9103064861898018e-05,
+      "loss": 0.6705,
+      "step": 5997
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9102769772778585e-05,
+      "loss": 0.6028,
+      "step": 5998
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910247463740531e-05,
+      "loss": 0.6478,
+      "step": 5999
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9102179455779702e-05,
+      "loss": 0.6221,
+      "step": 6000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9101884227903248e-05,
+      "loss": 0.5834,
+      "step": 6001
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910158895377746e-05,
+      "loss": 0.6148,
+      "step": 6002
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910129363340383e-05,
+      "loss": 0.638,
+      "step": 6003
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9100998266783862e-05,
+      "loss": 0.6254,
+      "step": 6004
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9100702853919054e-05,
+      "loss": 0.6673,
+      "step": 6005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.910040739481091e-05,
+      "loss": 0.6635,
+      "step": 6006
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9100111889460936e-05,
+      "loss": 0.5793,
+      "step": 6007
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9099816337870624e-05,
+      "loss": 0.6451,
+      "step": 6008
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9099520740041482e-05,
+      "loss": 0.6057,
+      "step": 6009
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909922509597501e-05,
+      "loss": 0.5938,
+      "step": 6010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9098929405672712e-05,
+      "loss": 0.6391,
+      "step": 6011
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9098633669136083e-05,
+      "loss": 0.5825,
+      "step": 6012
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9098337886366634e-05,
+      "loss": 0.6537,
+      "step": 6013
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909804205736587e-05,
+      "loss": 0.5581,
+      "step": 6014
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9097746182135287e-05,
+      "loss": 0.6068,
+      "step": 6015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909745026067639e-05,
+      "loss": 0.6564,
+      "step": 6016
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9097154292990687e-05,
+      "loss": 0.6031,
+      "step": 6017
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909685827907968e-05,
+      "loss": 0.5944,
+      "step": 6018
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9096562218944866e-05,
+      "loss": 0.643,
+      "step": 6019
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9096266112587757e-05,
+      "loss": 0.6362,
+      "step": 6020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9095969960009856e-05,
+      "loss": 0.5759,
+      "step": 6021
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909567376121267e-05,
+      "loss": 0.6128,
+      "step": 6022
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9095377516197698e-05,
+      "loss": 0.6045,
+      "step": 6023
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9095081224966453e-05,
+      "loss": 0.6481,
+      "step": 6024
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9094784887520436e-05,
+      "loss": 0.6158,
+      "step": 6025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9094488503861152e-05,
+      "loss": 0.6213,
+      "step": 6026
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9094192073990112e-05,
+      "loss": 0.6246,
+      "step": 6027
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9093895597908818e-05,
+      "loss": 0.6331,
+      "step": 6028
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9093599075618772e-05,
+      "loss": 0.572,
+      "step": 6029
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9093302507121493e-05,
+      "loss": 0.6323,
+      "step": 6030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9093005892418476e-05,
+      "loss": 0.6206,
+      "step": 6031
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9092709231511234e-05,
+      "loss": 0.5616,
+      "step": 6032
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9092412524401272e-05,
+      "loss": 0.6183,
+      "step": 6033
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9092115771090102e-05,
+      "loss": 0.5828,
+      "step": 6034
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9091818971579227e-05,
+      "loss": 0.57,
+      "step": 6035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909152212587016e-05,
+      "loss": 0.6139,
+      "step": 6036
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.90912252339644e-05,
+      "loss": 0.6353,
+      "step": 6037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909092829586347e-05,
+      "loss": 0.6585,
+      "step": 6038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9090631311568867e-05,
+      "loss": 0.6266,
+      "step": 6039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9090334281082102e-05,
+      "loss": 0.5624,
+      "step": 6040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.909003720440469e-05,
+      "loss": 0.6076,
+      "step": 6041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9089740081538137e-05,
+      "loss": 0.5439,
+      "step": 6042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9089442912483948e-05,
+      "loss": 0.5416,
+      "step": 6043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9089145697243644e-05,
+      "loss": 0.6431,
+      "step": 6044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9088848435818727e-05,
+      "loss": 0.5523,
+      "step": 6045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9088551128210703e-05,
+      "loss": 0.6254,
+      "step": 6046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9088253774421095e-05,
+      "loss": 0.6394,
+      "step": 6047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.908795637445141e-05,
+      "loss": 0.6428,
+      "step": 6048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9087658928303152e-05,
+      "loss": 0.6064,
+      "step": 6049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.908736143597784e-05,
+      "loss": 0.6207,
+      "step": 6050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9087063897476982e-05,
+      "loss": 0.6121,
+      "step": 6051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9086766312802093e-05,
+      "loss": 0.6525,
+      "step": 6052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9086468681954683e-05,
+      "loss": 0.5959,
+      "step": 6053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.908617100493626e-05,
+      "loss": 0.6335,
+      "step": 6054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9085873281748345e-05,
+      "loss": 0.6172,
+      "step": 6055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9085575512392446e-05,
+      "loss": 0.6265,
+      "step": 6056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9085277696870078e-05,
+      "loss": 0.6596,
+      "step": 6057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9084979835182757e-05,
+      "loss": 0.5859,
+      "step": 6058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9084681927331984e-05,
+      "loss": 0.6551,
+      "step": 6059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9084383973319288e-05,
+      "loss": 0.5733,
+      "step": 6060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9084085973146176e-05,
+      "loss": 0.6037,
+      "step": 6061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.908378792681416e-05,
+      "loss": 0.6016,
+      "step": 6062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9083489834324757e-05,
+      "loss": 0.6066,
+      "step": 6063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9083191695679485e-05,
+      "loss": 0.5762,
+      "step": 6064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9082893510879856e-05,
+      "loss": 0.5775,
+      "step": 6065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9082595279927383e-05,
+      "loss": 0.6174,
+      "step": 6066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9082297002823585e-05,
+      "loss": 0.603,
+      "step": 6067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9081998679569972e-05,
+      "loss": 0.592,
+      "step": 6068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9081700310168067e-05,
+      "loss": 0.5827,
+      "step": 6069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9081401894619383e-05,
+      "loss": 0.6103,
+      "step": 6070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9081103432925435e-05,
+      "loss": 0.6346,
+      "step": 6071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9080804925087744e-05,
+      "loss": 0.5922,
+      "step": 6072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9080506371107822e-05,
+      "loss": 0.6636,
+      "step": 6073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.908020777098719e-05,
+      "loss": 0.6381,
+      "step": 6074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907990912472736e-05,
+      "loss": 0.6067,
+      "step": 6075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907961043232985e-05,
+      "loss": 0.6238,
+      "step": 6076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9079311693796185e-05,
+      "loss": 0.6106,
+      "step": 6077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907901290912788e-05,
+      "loss": 0.6313,
+      "step": 6078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9078714078326447e-05,
+      "loss": 0.6426,
+      "step": 6079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907841520139341e-05,
+      "loss": 0.6377,
+      "step": 6080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9078116278330286e-05,
+      "loss": 0.6164,
+      "step": 6081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9077817309138598e-05,
+      "loss": 0.6612,
+      "step": 6082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9077518293819857e-05,
+      "loss": 0.6539,
+      "step": 6083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907721923237559e-05,
+      "loss": 0.5806,
+      "step": 6084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907692012480731e-05,
+      "loss": 0.6355,
+      "step": 6085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9076620971116546e-05,
+      "loss": 0.6435,
+      "step": 6086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907632177130481e-05,
+      "loss": 0.5832,
+      "step": 6087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9076022525373624e-05,
+      "loss": 0.608,
+      "step": 6088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9075723233324513e-05,
+      "loss": 0.6218,
+      "step": 6089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907542389515899e-05,
+      "loss": 0.6272,
+      "step": 6090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9075124510878583e-05,
+      "loss": 0.6352,
+      "step": 6091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907482508048481e-05,
+      "loss": 0.6462,
+      "step": 6092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9074525603979195e-05,
+      "loss": 0.6098,
+      "step": 6093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9074226081363253e-05,
+      "loss": 0.6251,
+      "step": 6094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9073926512638517e-05,
+      "loss": 0.6067,
+      "step": 6095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9073626897806495e-05,
+      "loss": 0.6082,
+      "step": 6096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.9073327236868725e-05,
+      "loss": 0.6769,
+      "step": 6097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.907302752982672e-05,
+      "loss": 0.6168,
+      "step": 6098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9072727776682004e-05,
+      "loss": 0.642,
+      "step": 6099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90724279774361e-05,
+      "loss": 0.6516,
+      "step": 6100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9072128132090536e-05,
+      "loss": 0.5571,
+      "step": 6101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9071828240646827e-05,
+      "loss": 0.6152,
+      "step": 6102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9071528303106507e-05,
+      "loss": 0.5589,
+      "step": 6103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9071228319471088e-05,
+      "loss": 0.6086,
+      "step": 6104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9070928289742108e-05,
+      "loss": 0.522,
+      "step": 6105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.907062821392108e-05,
+      "loss": 0.5937,
+      "step": 6106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.907032809200954e-05,
+      "loss": 0.558,
+      "step": 6107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9070027924009e-05,
+      "loss": 0.606,
+      "step": 6108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9069727709920995e-05,
+      "loss": 0.6516,
+      "step": 6109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9069427449747046e-05,
+      "loss": 0.5766,
+      "step": 6110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9069127143488683e-05,
+      "loss": 0.5918,
+      "step": 6111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9068826791147425e-05,
+      "loss": 0.6383,
+      "step": 6112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9068526392724807e-05,
+      "loss": 0.5925,
+      "step": 6113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9068225948222346e-05,
+      "loss": 0.5618,
+      "step": 6114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9067925457641574e-05,
+      "loss": 0.6045,
+      "step": 6115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906762492098402e-05,
+      "loss": 0.6276,
+      "step": 6116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90673243382512e-05,
+      "loss": 0.6186,
+      "step": 6117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906702370944466e-05,
+      "loss": 0.5582,
+      "step": 6118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906672303456591e-05,
+      "loss": 0.6378,
+      "step": 6119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906642231361649e-05,
+      "loss": 0.568,
+      "step": 6120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906612154659792e-05,
+      "loss": 0.5934,
+      "step": 6121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9065820733511733e-05,
+      "loss": 0.676,
+      "step": 6122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9065519874359454e-05,
+      "loss": 0.6571,
+      "step": 6123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9065218969142614e-05,
+      "loss": 0.687,
+      "step": 6124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9064918017862743e-05,
+      "loss": 0.6236,
+      "step": 6125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906461702052137e-05,
+      "loss": 0.6336,
+      "step": 6126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9064315977120017e-05,
+      "loss": 0.6813,
+      "step": 6127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9064014887660226e-05,
+      "loss": 0.5937,
+      "step": 6128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9063713752143516e-05,
+      "loss": 0.5305,
+      "step": 6129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9063412570571426e-05,
+      "loss": 0.6224,
+      "step": 6130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906311134294548e-05,
+      "loss": 0.6188,
+      "step": 6131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9062810069267214e-05,
+      "loss": 0.5991,
+      "step": 6132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906250874953815e-05,
+      "loss": 0.5374,
+      "step": 6133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906220738375983e-05,
+      "loss": 0.6604,
+      "step": 6134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.906190597193378e-05,
+      "loss": 0.5885,
+      "step": 6135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9061604514061533e-05,
+      "loss": 0.5839,
+      "step": 6136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9061303010144617e-05,
+      "loss": 0.6063,
+      "step": 6137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9061001460184566e-05,
+      "loss": 0.6324,
+      "step": 6138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9060699864182917e-05,
+      "loss": 0.5801,
+      "step": 6139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9060398222141193e-05,
+      "loss": 0.5728,
+      "step": 6140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9060096534060934e-05,
+      "loss": 0.6025,
+      "step": 6141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9059794799943676e-05,
+      "loss": 0.5809,
+      "step": 6142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905949301979094e-05,
+      "loss": 0.599,
+      "step": 6143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9059191193604272e-05,
+      "loss": 0.589,
+      "step": 6144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90588893213852e-05,
+      "loss": 0.6114,
+      "step": 6145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9058587403135257e-05,
+      "loss": 0.5906,
+      "step": 6146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905828543885598e-05,
+      "loss": 0.6164,
+      "step": 6147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9057983428548896e-05,
+      "loss": 0.6363,
+      "step": 6148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9057681372215547e-05,
+      "loss": 0.6117,
+      "step": 6149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905737926985747e-05,
+      "loss": 0.6744,
+      "step": 6150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9057077121476198e-05,
+      "loss": 0.6101,
+      "step": 6151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905677492707326e-05,
+      "loss": 0.674,
+      "step": 6152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90564726866502e-05,
+      "loss": 0.6208,
+      "step": 6153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9056170400208546e-05,
+      "loss": 0.5437,
+      "step": 6154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905586806774984e-05,
+      "loss": 0.583,
+      "step": 6155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905556568927562e-05,
+      "loss": 0.6699,
+      "step": 6156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905526326478741e-05,
+      "loss": 0.6313,
+      "step": 6157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9054960794286763e-05,
+      "loss": 0.671,
+      "step": 6158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9054658277775203e-05,
+      "loss": 0.569,
+      "step": 6159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9054355715254277e-05,
+      "loss": 0.5898,
+      "step": 6160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9054053106725517e-05,
+      "loss": 0.6236,
+      "step": 6161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905375045219046e-05,
+      "loss": 0.617,
+      "step": 6162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9053447751650645e-05,
+      "loss": 0.6274,
+      "step": 6163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9053145005107614e-05,
+      "loss": 0.6028,
+      "step": 6164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9052842212562898e-05,
+      "loss": 0.6334,
+      "step": 6165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905253937401804e-05,
+      "loss": 0.5665,
+      "step": 6166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905223648947458e-05,
+      "loss": 0.6113,
+      "step": 6167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9051933558934057e-05,
+      "loss": 0.6164,
+      "step": 6168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9051630582398006e-05,
+      "loss": 0.6409,
+      "step": 6169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.905132755986797e-05,
+      "loss": 0.639,
+      "step": 6170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9051024491345486e-05,
+      "loss": 0.5971,
+      "step": 6171
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9050721376832098e-05,
+      "loss": 0.5953,
+      "step": 6172
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9050418216329344e-05,
+      "loss": 0.5812,
+      "step": 6173
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9050115009838762e-05,
+      "loss": 0.6127,
+      "step": 6174
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9049811757361894e-05,
+      "loss": 0.5681,
+      "step": 6175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904950845890029e-05,
+      "loss": 0.5991,
+      "step": 6176
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9049205114455477e-05,
+      "loss": 0.6624,
+      "step": 6177
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9048901724029004e-05,
+      "loss": 0.6115,
+      "step": 6178
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9048598287622408e-05,
+      "loss": 0.5932,
+      "step": 6179
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904829480523724e-05,
+      "loss": 0.6104,
+      "step": 6180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904799127687503e-05,
+      "loss": 0.5862,
+      "step": 6181
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904768770253733e-05,
+      "loss": 0.6493,
+      "step": 6182
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904738408222568e-05,
+      "loss": 0.6124,
+      "step": 6183
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9047080415941622e-05,
+      "loss": 0.5946,
+      "step": 6184
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9046776703686695e-05,
+      "loss": 0.5541,
+      "step": 6185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904647294546245e-05,
+      "loss": 0.5814,
+      "step": 6186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9046169141270424e-05,
+      "loss": 0.5835,
+      "step": 6187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9045865291112167e-05,
+      "loss": 0.628,
+      "step": 6188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9045561394989216e-05,
+      "loss": 0.6518,
+      "step": 6189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9045257452903117e-05,
+      "loss": 0.6237,
+      "step": 6190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904495346485542e-05,
+      "loss": 0.5837,
+      "step": 6191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9044649430847664e-05,
+      "loss": 0.5886,
+      "step": 6192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9044345350881396e-05,
+      "loss": 0.5995,
+      "step": 6193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904404122495816e-05,
+      "loss": 0.6434,
+      "step": 6194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90437370530795e-05,
+      "loss": 0.5671,
+      "step": 6195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9043432835246964e-05,
+      "loss": 0.5813,
+      "step": 6196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9043128571462098e-05,
+      "loss": 0.6663,
+      "step": 6197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9042824261726447e-05,
+      "loss": 0.6354,
+      "step": 6198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904251990604156e-05,
+      "loss": 0.6012,
+      "step": 6199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9042215504408977e-05,
+      "loss": 0.5984,
+      "step": 6200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904191105683025e-05,
+      "loss": 0.6721,
+      "step": 6201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9041606563306926e-05,
+      "loss": 0.5872,
+      "step": 6202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9041302023840554e-05,
+      "loss": 0.5975,
+      "step": 6203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9040997438432674e-05,
+      "loss": 0.6599,
+      "step": 6204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.904069280708484e-05,
+      "loss": 0.6276,
+      "step": 6205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90403881297986e-05,
+      "loss": 0.6139,
+      "step": 6206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9040083406575497e-05,
+      "loss": 0.5911,
+      "step": 6207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9039778637417084e-05,
+      "loss": 0.5904,
+      "step": 6208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9039473822324907e-05,
+      "loss": 0.6009,
+      "step": 6209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903916896130052e-05,
+      "loss": 0.594,
+      "step": 6210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9038864054345464e-05,
+      "loss": 0.6158,
+      "step": 6211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9038559101461295e-05,
+      "loss": 0.6392,
+      "step": 6212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903825410264956e-05,
+      "loss": 0.6608,
+      "step": 6213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903794905791181e-05,
+      "loss": 0.5449,
+      "step": 6214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903764396724959e-05,
+      "loss": 0.5881,
+      "step": 6215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9037338830664458e-05,
+      "loss": 0.6366,
+      "step": 6216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9037033648157957e-05,
+      "loss": 0.6623,
+      "step": 6217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9036728419731646e-05,
+      "loss": 0.6529,
+      "step": 6218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903642314538707e-05,
+      "loss": 0.5991,
+      "step": 6219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9036117825125777e-05,
+      "loss": 0.5778,
+      "step": 6220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9035812458949327e-05,
+      "loss": 0.6423,
+      "step": 6221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903550704685927e-05,
+      "loss": 0.6127,
+      "step": 6222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9035201588857152e-05,
+      "loss": 0.5889,
+      "step": 6223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903489608494453e-05,
+      "loss": 0.5753,
+      "step": 6224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9034590535122953e-05,
+      "loss": 0.6668,
+      "step": 6225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9034284939393977e-05,
+      "loss": 0.5878,
+      "step": 6226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9033979297759152e-05,
+      "loss": 0.5797,
+      "step": 6227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9033673610220035e-05,
+      "loss": 0.6679,
+      "step": 6228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9033367876778176e-05,
+      "loss": 0.6063,
+      "step": 6229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9033062097435127e-05,
+      "loss": 0.6505,
+      "step": 6230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9032756272192447e-05,
+      "loss": 0.6072,
+      "step": 6231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9032450401051684e-05,
+      "loss": 0.5649,
+      "step": 6232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.90321444840144e-05,
+      "loss": 0.6475,
+      "step": 6233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903183852108214e-05,
+      "loss": 0.6054,
+      "step": 6234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9031532512256466e-05,
+      "loss": 0.5817,
+      "step": 6235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.903122645753893e-05,
+      "loss": 0.6422,
+      "step": 6236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9030920356931087e-05,
+      "loss": 0.5774,
+      "step": 6237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9030614210434492e-05,
+      "loss": 0.5958,
+      "step": 6238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9030308018050704e-05,
+      "loss": 0.6862,
+      "step": 6239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9030001779781274e-05,
+      "loss": 0.5924,
+      "step": 6240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.902969549562776e-05,
+      "loss": 0.6147,
+      "step": 6241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.902938916559172e-05,
+      "loss": 0.63,
+      "step": 6242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9029082789674706e-05,
+      "loss": 0.6148,
+      "step": 6243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9028776367878282e-05,
+      "loss": 0.5729,
+      "step": 6244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9028469900204e-05,
+      "loss": 0.5793,
+      "step": 6245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9028163386653417e-05,
+      "loss": 0.5873,
+      "step": 6246
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9027856827228098e-05,
+      "loss": 0.6427,
+      "step": 6247
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902755022192959e-05,
+      "loss": 0.6248,
+      "step": 6248
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9027243570759454e-05,
+      "loss": 0.6458,
+      "step": 6249
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902693687371925e-05,
+      "loss": 0.6153,
+      "step": 6250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9026630130810538e-05,
+      "loss": 0.5771,
+      "step": 6251
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9026323342034873e-05,
+      "loss": 0.6307,
+      "step": 6252
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9026016507393817e-05,
+      "loss": 0.6579,
+      "step": 6253
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9025709626888926e-05,
+      "loss": 0.6506,
+      "step": 6254
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902540270052176e-05,
+      "loss": 0.5402,
+      "step": 6255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902509572829388e-05,
+      "loss": 0.6167,
+      "step": 6256
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9024788710206848e-05,
+      "loss": 0.6019,
+      "step": 6257
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9024481646262224e-05,
+      "loss": 0.561,
+      "step": 6258
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902417453646156e-05,
+      "loss": 0.6267,
+      "step": 6259
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902386738080643e-05,
+      "loss": 0.591,
+      "step": 6260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9023560179298378e-05,
+      "loss": 0.5986,
+      "step": 6261
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902325293193898e-05,
+      "loss": 0.5934,
+      "step": 6262
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902294563872979e-05,
+      "loss": 0.5806,
+      "step": 6263
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902263829967237e-05,
+      "loss": 0.5746,
+      "step": 6264
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9022330914768283e-05,
+      "loss": 0.6078,
+      "step": 6265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9022023484019086e-05,
+      "loss": 0.5957,
+      "step": 6266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9021716007426353e-05,
+      "loss": 0.6796,
+      "step": 6267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9021408484991634e-05,
+      "loss": 0.6281,
+      "step": 6268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9021100916716498e-05,
+      "loss": 0.6263,
+      "step": 6269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9020793302602504e-05,
+      "loss": 0.6314,
+      "step": 6270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.902048564265122e-05,
+      "loss": 0.5558,
+      "step": 6271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9020177936864205e-05,
+      "loss": 0.6492,
+      "step": 6272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9019870185243024e-05,
+      "loss": 0.5322,
+      "step": 6273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9019562387789247e-05,
+      "loss": 0.6463,
+      "step": 6274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9019254544504424e-05,
+      "loss": 0.6081,
+      "step": 6275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9018946655390128e-05,
+      "loss": 0.604,
+      "step": 6276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9018638720447927e-05,
+      "loss": 0.5967,
+      "step": 6277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901833073967938e-05,
+      "loss": 0.5723,
+      "step": 6278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9018022713086054e-05,
+      "loss": 0.6625,
+      "step": 6279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901771464066951e-05,
+      "loss": 0.6393,
+      "step": 6280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901740652243132e-05,
+      "loss": 0.6553,
+      "step": 6281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9017098358373046e-05,
+      "loss": 0.6086,
+      "step": 6282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9016790148496253e-05,
+      "loss": 0.6379,
+      "step": 6283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9016481892802513e-05,
+      "loss": 0.606,
+      "step": 6284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9016173591293384e-05,
+      "loss": 0.5561,
+      "step": 6285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901586524397044e-05,
+      "loss": 0.5581,
+      "step": 6286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901555685083524e-05,
+      "loss": 0.6147,
+      "step": 6287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9015248411889356e-05,
+      "loss": 0.6054,
+      "step": 6288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9014939927134356e-05,
+      "loss": 0.6077,
+      "step": 6289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9014631396571804e-05,
+      "loss": 0.632,
+      "step": 6290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9014322820203272e-05,
+      "loss": 0.6933,
+      "step": 6291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9014014198030323e-05,
+      "loss": 0.6011,
+      "step": 6292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901370553005453e-05,
+      "loss": 0.5313,
+      "step": 6293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.901339681627746e-05,
+      "loss": 0.6191,
+      "step": 6294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9013088056700683e-05,
+      "loss": 0.6068,
+      "step": 6295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9012779251325763e-05,
+      "loss": 0.6062,
+      "step": 6296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9012470400154274e-05,
+      "loss": 0.6944,
+      "step": 6297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9012161503187783e-05,
+      "loss": 0.6351,
+      "step": 6298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9011852560427856e-05,
+      "loss": 0.6701,
+      "step": 6299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9011543571876073e-05,
+      "loss": 0.6157,
+      "step": 6300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9011234537533995e-05,
+      "loss": 0.533,
+      "step": 6301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9010925457403195e-05,
+      "loss": 0.6301,
+      "step": 6302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9010616331485244e-05,
+      "loss": 0.6124,
+      "step": 6303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9010307159781712e-05,
+      "loss": 0.6517,
+      "step": 6304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900999794229417e-05,
+      "loss": 0.6122,
+      "step": 6305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900968867902419e-05,
+      "loss": 0.6031,
+      "step": 6306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9009379369973346e-05,
+      "loss": 0.5991,
+      "step": 6307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9009070015143206e-05,
+      "loss": 0.6156,
+      "step": 6308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9008760614535343e-05,
+      "loss": 0.6043,
+      "step": 6309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9008451168151326e-05,
+      "loss": 0.6275,
+      "step": 6310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9008141675992733e-05,
+      "loss": 0.6395,
+      "step": 6311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9007832138061134e-05,
+      "loss": 0.6063,
+      "step": 6312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.90075225543581e-05,
+      "loss": 0.6255,
+      "step": 6313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900721292488521e-05,
+      "loss": 0.687,
+      "step": 6314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900690324964403e-05,
+      "loss": 0.6317,
+      "step": 6315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900659352863614e-05,
+      "loss": 0.6574,
+      "step": 6316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9006283761863105e-05,
+      "loss": 0.5864,
+      "step": 6317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9005973949326508e-05,
+      "loss": 0.5732,
+      "step": 6318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900566409102792e-05,
+      "loss": 0.5769,
+      "step": 6319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9005354186968913e-05,
+      "loss": 0.6471,
+      "step": 6320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9005044237151064e-05,
+      "loss": 0.5437,
+      "step": 6321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900473424157595e-05,
+      "loss": 0.6548,
+      "step": 6322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900442420024514e-05,
+      "loss": 0.628,
+      "step": 6323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900411411316022e-05,
+      "loss": 0.621,
+      "step": 6324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9003803980322754e-05,
+      "loss": 0.6523,
+      "step": 6325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9003493801734326e-05,
+      "loss": 0.6351,
+      "step": 6326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900318357739651e-05,
+      "loss": 0.7169,
+      "step": 6327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9002873307310882e-05,
+      "loss": 0.5685,
+      "step": 6328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9002562991479013e-05,
+      "loss": 0.667,
+      "step": 6329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900225262990249e-05,
+      "loss": 0.5997,
+      "step": 6330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9001942222582884e-05,
+      "loss": 0.5952,
+      "step": 6331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900163176952177e-05,
+      "loss": 0.5791,
+      "step": 6332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900132127072073e-05,
+      "loss": 0.6834,
+      "step": 6333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9001010726181343e-05,
+      "loss": 0.6243,
+      "step": 6334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9000700135905185e-05,
+      "loss": 0.5841,
+      "step": 6335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.9000389499893833e-05,
+      "loss": 0.5923,
+      "step": 6336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.900007881814886e-05,
+      "loss": 0.6571,
+      "step": 6337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8999768090671858e-05,
+      "loss": 0.6014,
+      "step": 6338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8999457317464397e-05,
+      "loss": 0.5626,
+      "step": 6339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8999146498528056e-05,
+      "loss": 0.6141,
+      "step": 6340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8998835633864415e-05,
+      "loss": 0.6038,
+      "step": 6341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899852472347506e-05,
+      "loss": 0.6375,
+      "step": 6342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899821376736156e-05,
+      "loss": 0.6688,
+      "step": 6343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8997902765525503e-05,
+      "loss": 0.5799,
+      "step": 6344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8997591717968468e-05,
+      "loss": 0.6227,
+      "step": 6345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8997280624692034e-05,
+      "loss": 0.6409,
+      "step": 6346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8996969485697782e-05,
+      "loss": 0.6791,
+      "step": 6347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8996658300987292e-05,
+      "loss": 0.6206,
+      "step": 6348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8996347070562146e-05,
+      "loss": 0.5972,
+      "step": 6349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8996035794423926e-05,
+      "loss": 0.6248,
+      "step": 6350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8995724472574214e-05,
+      "loss": 0.6263,
+      "step": 6351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899541310501459e-05,
+      "loss": 0.6442,
+      "step": 6352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899510169174664e-05,
+      "loss": 0.6796,
+      "step": 6353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8994790232771944e-05,
+      "loss": 0.6337,
+      "step": 6354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8994478728092085e-05,
+      "loss": 0.6522,
+      "step": 6355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8994167177708642e-05,
+      "loss": 0.6459,
+      "step": 6356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8993855581623205e-05,
+      "loss": 0.6377,
+      "step": 6357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8993543939837354e-05,
+      "loss": 0.6403,
+      "step": 6358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8993232252352667e-05,
+      "loss": 0.6231,
+      "step": 6359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899292051917074e-05,
+      "loss": 0.5801,
+      "step": 6360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899260874029315e-05,
+      "loss": 0.6704,
+      "step": 6361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8992296915721475e-05,
+      "loss": 0.6371,
+      "step": 6362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.899198504545731e-05,
+      "loss": 0.6549,
+      "step": 6363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8991673129502236e-05,
+      "loss": 0.6202,
+      "step": 6364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8991361167857833e-05,
+      "loss": 0.5763,
+      "step": 6365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8991049160525695e-05,
+      "loss": 0.6482,
+      "step": 6366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8990737107507403e-05,
+      "loss": 0.5684,
+      "step": 6367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8990425008804544e-05,
+      "loss": 0.6639,
+      "step": 6368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8990112864418697e-05,
+      "loss": 0.6242,
+      "step": 6369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898980067435146e-05,
+      "loss": 0.6463,
+      "step": 6370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898948843860441e-05,
+      "loss": 0.6264,
+      "step": 6371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8989176157179136e-05,
+      "loss": 0.6756,
+      "step": 6372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8988863830077227e-05,
+      "loss": 0.5691,
+      "step": 6373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8988551457300267e-05,
+      "loss": 0.62,
+      "step": 6374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8988239038849846e-05,
+      "loss": 0.5628,
+      "step": 6375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8987926574727548e-05,
+      "loss": 0.6375,
+      "step": 6376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898761406493496e-05,
+      "loss": 0.6058,
+      "step": 6377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898730150947368e-05,
+      "loss": 0.649,
+      "step": 6378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8986988908345285e-05,
+      "loss": 0.6006,
+      "step": 6379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898667626155137e-05,
+      "loss": 0.6327,
+      "step": 6380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8986363569093516e-05,
+      "loss": 0.5946,
+      "step": 6381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8986050830973322e-05,
+      "loss": 0.541,
+      "step": 6382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898573804719237e-05,
+      "loss": 0.6424,
+      "step": 6383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898542521775225e-05,
+      "loss": 0.5763,
+      "step": 6384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8985112342654556e-05,
+      "loss": 0.5847,
+      "step": 6385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8984799421900872e-05,
+      "loss": 0.6183,
+      "step": 6386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8984486455492793e-05,
+      "loss": 0.6251,
+      "step": 6387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8984173443431905e-05,
+      "loss": 0.6242,
+      "step": 6388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.8983860385719804e-05,
+      "loss": 0.6111,
+      "step": 6389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898354728235807e-05,
+      "loss": 0.6341,
+      "step": 6390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.898323413334831e-05,
+      "loss": 0.5812,
+      "step": 6391
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8982920938692102e-05,
+      "loss": 0.6684,
+      "step": 6392
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.898260769839104e-05,
+      "loss": 0.6156,
+      "step": 6393
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8982294412446722e-05,
+      "loss": 0.6359,
+      "step": 6394
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8981981080860733e-05,
+      "loss": 0.6317,
+      "step": 6395
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8981667703634672e-05,
+      "loss": 0.671,
+      "step": 6396
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8981354280770123e-05,
+      "loss": 0.5668,
+      "step": 6397
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.898104081226868e-05,
+      "loss": 0.6536,
+      "step": 6398
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8980727298131945e-05,
+      "loss": 0.5805,
+      "step": 6399
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.89804137383615e-05,
+      "loss": 0.6482,
+      "step": 6400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8980100132958946e-05,
+      "loss": 0.6326,
+      "step": 6401
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897978648192587e-05,
+      "loss": 0.5844,
+      "step": 6402
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897947278526387e-05,
+      "loss": 0.6562,
+      "step": 6403
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897915904297454e-05,
+      "loss": 0.6224,
+      "step": 6404
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8978845255059474e-05,
+      "loss": 0.6693,
+      "step": 6405
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897853142152026e-05,
+      "loss": 0.6355,
+      "step": 6406
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8978217542358503e-05,
+      "loss": 0.6009,
+      "step": 6407
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8977903617575793e-05,
+      "loss": 0.6511,
+      "step": 6408
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8977589647173724e-05,
+      "loss": 0.6129,
+      "step": 6409
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8977275631153895e-05,
+      "loss": 0.6705,
+      "step": 6410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8976961569517898e-05,
+      "loss": 0.6002,
+      "step": 6411
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8976647462267332e-05,
+      "loss": 0.5883,
+      "step": 6412
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897633330940379e-05,
+      "loss": 0.5763,
+      "step": 6413
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897601911092887e-05,
+      "loss": 0.5822,
+      "step": 6414
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8975704866844163e-05,
+      "loss": 0.6667,
+      "step": 6415
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897539057715128e-05,
+      "loss": 0.5915,
+      "step": 6416
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.89750762418518e-05,
+      "loss": 0.6536,
+      "step": 6417
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8974761860947337e-05,
+      "loss": 0.6033,
+      "step": 6418
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8974447434439476e-05,
+      "loss": 0.6249,
+      "step": 6419
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897413296232982e-05,
+      "loss": 0.577,
+      "step": 6420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8973818444619968e-05,
+      "loss": 0.6537,
+      "step": 6421
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8973503881311516e-05,
+      "loss": 0.5768,
+      "step": 6422
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897318927240606e-05,
+      "loss": 0.5957,
+      "step": 6423
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8972874617905204e-05,
+      "loss": 0.6384,
+      "step": 6424
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8972559917810542e-05,
+      "loss": 0.6373,
+      "step": 6425
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897224517212368e-05,
+      "loss": 0.5907,
+      "step": 6426
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8971930380846205e-05,
+      "loss": 0.5965,
+      "step": 6427
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897161554397973e-05,
+      "loss": 0.6802,
+      "step": 6428
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897130066152585e-05,
+      "loss": 0.6519,
+      "step": 6429
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.897098573348616e-05,
+      "loss": 0.6507,
+      "step": 6430
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8970670759862265e-05,
+      "loss": 0.5959,
+      "step": 6431
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8970355740655767e-05,
+      "loss": 0.6597,
+      "step": 6432
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8970040675868262e-05,
+      "loss": 0.6023,
+      "step": 6433
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8969725565501355e-05,
+      "loss": 0.6101,
+      "step": 6434
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8969410409556645e-05,
+      "loss": 0.5826,
+      "step": 6435
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896909520803573e-05,
+      "loss": 0.6485,
+      "step": 6436
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896877996094022e-05,
+      "loss": 0.6926,
+      "step": 6437
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8968464668271714e-05,
+      "loss": 0.5723,
+      "step": 6438
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896814933003181e-05,
+      "loss": 0.6097,
+      "step": 6439
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8967833946222113e-05,
+      "loss": 0.6465,
+      "step": 6440
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8967518516844224e-05,
+      "loss": 0.6608,
+      "step": 6441
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896720304189975e-05,
+      "loss": 0.5736,
+      "step": 6442
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8966887521390288e-05,
+      "loss": 0.5581,
+      "step": 6443
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8966571955317446e-05,
+      "loss": 0.639,
+      "step": 6444
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8966256343682823e-05,
+      "loss": 0.5794,
+      "step": 6445
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896594068648803e-05,
+      "loss": 0.6298,
+      "step": 6446
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8965624983734665e-05,
+      "loss": 0.625,
+      "step": 6447
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8965309235424335e-05,
+      "loss": 0.5599,
+      "step": 6448
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896499344155864e-05,
+      "loss": 0.6703,
+      "step": 6449
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896467760213919e-05,
+      "loss": 0.6351,
+      "step": 6450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8964361717167588e-05,
+      "loss": 0.6074,
+      "step": 6451
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896404578664544e-05,
+      "loss": 0.681,
+      "step": 6452
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8963729810574344e-05,
+      "loss": 0.6234,
+      "step": 6453
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8963413788955917e-05,
+      "loss": 0.5983,
+      "step": 6454
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896309772179176e-05,
+      "loss": 0.5539,
+      "step": 6455
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8962781609083478e-05,
+      "loss": 0.6506,
+      "step": 6456
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8962465450832677e-05,
+      "loss": 0.6342,
+      "step": 6457
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896214924704096e-05,
+      "loss": 0.5848,
+      "step": 6458
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8961832997709947e-05,
+      "loss": 0.6217,
+      "step": 6459
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8961516702841226e-05,
+      "loss": 0.6142,
+      "step": 6460
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8961200362436424e-05,
+      "loss": 0.604,
+      "step": 6461
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8960883976497132e-05,
+      "loss": 0.62,
+      "step": 6462
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8960567545024965e-05,
+      "loss": 0.5898,
+      "step": 6463
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.896025106802153e-05,
+      "loss": 0.6312,
+      "step": 6464
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8959934545488436e-05,
+      "loss": 0.5716,
+      "step": 6465
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895961797742729e-05,
+      "loss": 0.5871,
+      "step": 6466
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8959301363839705e-05,
+      "loss": 0.5939,
+      "step": 6467
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8958984704727277e-05,
+      "loss": 0.5903,
+      "step": 6468
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8958668000091633e-05,
+      "loss": 0.6257,
+      "step": 6469
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895835124993437e-05,
+      "loss": 0.6354,
+      "step": 6470
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8958034454257102e-05,
+      "loss": 0.575,
+      "step": 6471
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8957717613061434e-05,
+      "loss": 0.6757,
+      "step": 6472
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8957400726348983e-05,
+      "loss": 0.6075,
+      "step": 6473
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8957083794121352e-05,
+      "loss": 0.5715,
+      "step": 6474
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8956766816380158e-05,
+      "loss": 0.6445,
+      "step": 6475
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895644979312701e-05,
+      "loss": 0.6288,
+      "step": 6476
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8956132724363515e-05,
+      "loss": 0.6572,
+      "step": 6477
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8955815610091286e-05,
+      "loss": 0.6049,
+      "step": 6478
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8955498450311937e-05,
+      "loss": 0.6569,
+      "step": 6479
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8955181245027074e-05,
+      "loss": 0.5872,
+      "step": 6480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8954863994238314e-05,
+      "loss": 0.6237,
+      "step": 6481
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895454669794727e-05,
+      "loss": 0.6199,
+      "step": 6482
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895422935615555e-05,
+      "loss": 0.6335,
+      "step": 6483
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895391196886477e-05,
+      "loss": 0.6061,
+      "step": 6484
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895359453607654e-05,
+      "loss": 0.5974,
+      "step": 6485
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8953277057792473e-05,
+      "loss": 0.6423,
+      "step": 6486
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8952959534014182e-05,
+      "loss": 0.6434,
+      "step": 6487
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8952641964743287e-05,
+      "loss": 0.5845,
+      "step": 6488
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895232434998139e-05,
+      "loss": 0.619,
+      "step": 6489
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8952006689730117e-05,
+      "loss": 0.5946,
+      "step": 6490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895168898399107e-05,
+      "loss": 0.5435,
+      "step": 6491
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8951371232765874e-05,
+      "loss": 0.5611,
+      "step": 6492
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895105343605614e-05,
+      "loss": 0.6138,
+      "step": 6493
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895073559386348e-05,
+      "loss": 0.6316,
+      "step": 6494
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.895041770618951e-05,
+      "loss": 0.652,
+      "step": 6495
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8950099773035848e-05,
+      "loss": 0.6401,
+      "step": 6496
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8949781794404108e-05,
+      "loss": 0.5492,
+      "step": 6497
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8949463770295903e-05,
+      "loss": 0.6203,
+      "step": 6498
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8949145700712855e-05,
+      "loss": 0.6226,
+      "step": 6499
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8948827585656575e-05,
+      "loss": 0.6177,
+      "step": 6500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8948509425128683e-05,
+      "loss": 0.6178,
+      "step": 6501
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8948191219130793e-05,
+      "loss": 0.6479,
+      "step": 6502
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8947872967664522e-05,
+      "loss": 0.6122,
+      "step": 6503
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8947554670731488e-05,
+      "loss": 0.6228,
+      "step": 6504
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894723632833331e-05,
+      "loss": 0.6146,
+      "step": 6505
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.89469179404716e-05,
+      "loss": 0.5997,
+      "step": 6506
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8946599507147983e-05,
+      "loss": 0.5719,
+      "step": 6507
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8946281028364074e-05,
+      "loss": 0.6216,
+      "step": 6508
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894596250412149e-05,
+      "loss": 0.6091,
+      "step": 6509
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8945643934421853e-05,
+      "loss": 0.5712,
+      "step": 6510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8945325319266777e-05,
+      "loss": 0.6845,
+      "step": 6511
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8945006658657884e-05,
+      "loss": 0.589,
+      "step": 6512
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894468795259679e-05,
+      "loss": 0.6199,
+      "step": 6513
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894436920108512e-05,
+      "loss": 0.6153,
+      "step": 6514
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894405040412449e-05,
+      "loss": 0.6148,
+      "step": 6515
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894373156171652e-05,
+      "loss": 0.6356,
+      "step": 6516
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894341267386283e-05,
+      "loss": 0.6522,
+      "step": 6517
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8943093740565042e-05,
+      "loss": 0.6224,
+      "step": 6518
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8942774761824773e-05,
+      "loss": 0.5436,
+      "step": 6519
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894245573764365e-05,
+      "loss": 0.5841,
+      "step": 6520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894213666802329e-05,
+      "loss": 0.652,
+      "step": 6521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8941817552965312e-05,
+      "loss": 0.5994,
+      "step": 6522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8941498392471343e-05,
+      "loss": 0.5964,
+      "step": 6523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8941179186543e-05,
+      "loss": 0.6424,
+      "step": 6524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894085993518191e-05,
+      "loss": 0.5801,
+      "step": 6525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.894054063838969e-05,
+      "loss": 0.5539,
+      "step": 6526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8940221296167966e-05,
+      "loss": 0.6928,
+      "step": 6527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8939901908518358e-05,
+      "loss": 0.6141,
+      "step": 6528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.893958247544249e-05,
+      "loss": 0.6382,
+      "step": 6529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8939262996941986e-05,
+      "loss": 0.5435,
+      "step": 6530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.893894347301847e-05,
+      "loss": 0.5995,
+      "step": 6531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8938623903673564e-05,
+      "loss": 0.6059,
+      "step": 6532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.893830428890889e-05,
+      "loss": 0.5623,
+      "step": 6533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8937984628726075e-05,
+      "loss": 0.595,
+      "step": 6534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8937664923126743e-05,
+      "loss": 0.6882,
+      "step": 6535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.893734517211252e-05,
+      "loss": 0.6372,
+      "step": 6536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8937025375685025e-05,
+      "loss": 0.5469,
+      "step": 6537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.8936705533845888e-05,
+      "loss": 0.6151,
+      "step": 6538
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8936385646596735e-05,
+      "loss": 0.6299,
+      "step": 6539
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8936065713939186e-05,
+      "loss": 0.708,
+      "step": 6540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893574573587487e-05,
+      "loss": 0.6509,
+      "step": 6541
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893542571240542e-05,
+      "loss": 0.6502,
+      "step": 6542
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8935105643532446e-05,
+      "loss": 0.5906,
+      "step": 6543
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893478552925759e-05,
+      "loss": 0.7033,
+      "step": 6544
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8934465369582468e-05,
+      "loss": 0.6181,
+      "step": 6545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893414516450871e-05,
+      "loss": 0.6333,
+      "step": 6546
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893382491403795e-05,
+      "loss": 0.6276,
+      "step": 6547
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8933504618171806e-05,
+      "loss": 0.6071,
+      "step": 6548
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8933184276911906e-05,
+      "loss": 0.5968,
+      "step": 6549
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8932863890259883e-05,
+      "loss": 0.5892,
+      "step": 6550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893254345821736e-05,
+      "loss": 0.6053,
+      "step": 6551
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8932222980785973e-05,
+      "loss": 0.6071,
+      "step": 6552
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893190245796734e-05,
+      "loss": 0.6485,
+      "step": 6553
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8931581889763098e-05,
+      "loss": 0.599,
+      "step": 6554
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893126127617487e-05,
+      "loss": 0.5698,
+      "step": 6555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.893094061720429e-05,
+      "loss": 0.6373,
+      "step": 6556
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8930619912852984e-05,
+      "loss": 0.586,
+      "step": 6557
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8930299163122582e-05,
+      "loss": 0.6302,
+      "step": 6558
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8929978368014718e-05,
+      "loss": 0.6718,
+      "step": 6559
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8929657527531017e-05,
+      "loss": 0.6183,
+      "step": 6560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892933664167311e-05,
+      "loss": 0.6215,
+      "step": 6561
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892901571044263e-05,
+      "loss": 0.6334,
+      "step": 6562
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.89286947338412e-05,
+      "loss": 0.568,
+      "step": 6563
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8928373711870466e-05,
+      "loss": 0.5892,
+      "step": 6564
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8928052644532043e-05,
+      "loss": 0.6241,
+      "step": 6565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8927731531827576e-05,
+      "loss": 0.628,
+      "step": 6566
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8927410373758687e-05,
+      "loss": 0.5923,
+      "step": 6567
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8927089170327013e-05,
+      "loss": 0.7058,
+      "step": 6568
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8926767921534184e-05,
+      "loss": 0.6583,
+      "step": 6569
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8926446627381832e-05,
+      "loss": 0.6234,
+      "step": 6570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892612528787159e-05,
+      "loss": 0.6098,
+      "step": 6571
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892580390300509e-05,
+      "loss": 0.5955,
+      "step": 6572
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892548247278397e-05,
+      "loss": 0.6151,
+      "step": 6573
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892516099720986e-05,
+      "loss": 0.6462,
+      "step": 6574
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8924839476284392e-05,
+      "loss": 0.6411,
+      "step": 6575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8924517910009198e-05,
+      "loss": 0.5804,
+      "step": 6576
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8924196298385916e-05,
+      "loss": 0.6304,
+      "step": 6577
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8923874641416183e-05,
+      "loss": 0.6054,
+      "step": 6578
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8923552939101625e-05,
+      "loss": 0.6178,
+      "step": 6579
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8923231191443883e-05,
+      "loss": 0.6751,
+      "step": 6580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892290939844459e-05,
+      "loss": 0.5873,
+      "step": 6581
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892258756010538e-05,
+      "loss": 0.615,
+      "step": 6582
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892226567642789e-05,
+      "loss": 0.6024,
+      "step": 6583
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8921943747413755e-05,
+      "loss": 0.6564,
+      "step": 6584
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8921621773064612e-05,
+      "loss": 0.6069,
+      "step": 6585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8921299753382096e-05,
+      "loss": 0.6492,
+      "step": 6586
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8920977688367845e-05,
+      "loss": 0.6036,
+      "step": 6587
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892065557802349e-05,
+      "loss": 0.5949,
+      "step": 6588
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.892033342235067e-05,
+      "loss": 0.6288,
+      "step": 6589
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8920011221351026e-05,
+      "loss": 0.5951,
+      "step": 6590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8919688975026193e-05,
+      "loss": 0.5715,
+      "step": 6591
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891936668337781e-05,
+      "loss": 0.5921,
+      "step": 6592
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891904434640751e-05,
+      "loss": 0.5665,
+      "step": 6593
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8918721964116936e-05,
+      "loss": 0.6134,
+      "step": 6594
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891839953650772e-05,
+      "loss": 0.5769,
+      "step": 6595
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8918077063581506e-05,
+      "loss": 0.6194,
+      "step": 6596
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891775454533993e-05,
+      "loss": 0.6066,
+      "step": 6597
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891743198178463e-05,
+      "loss": 0.6047,
+      "step": 6598
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891710937291725e-05,
+      "loss": 0.5828,
+      "step": 6599
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8916786718739426e-05,
+      "loss": 0.6535,
+      "step": 6600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8916464019252797e-05,
+      "loss": 0.5994,
+      "step": 6601
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8916141274459e-05,
+      "loss": 0.6184,
+      "step": 6602
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891581848435968e-05,
+      "loss": 0.6021,
+      "step": 6603
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8915495648956473e-05,
+      "loss": 0.5959,
+      "step": 6604
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8915172768251024e-05,
+      "loss": 0.6128,
+      "step": 6605
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891484984224497e-05,
+      "loss": 0.6179,
+      "step": 6606
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8914526870939955e-05,
+      "loss": 0.557,
+      "step": 6607
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8914203854337615e-05,
+      "loss": 0.5993,
+      "step": 6608
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8913880792439597e-05,
+      "loss": 0.5837,
+      "step": 6609
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891355768524754e-05,
+      "loss": 0.5908,
+      "step": 6610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8913234532763087e-05,
+      "loss": 0.6234,
+      "step": 6611
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8912911334987876e-05,
+      "loss": 0.6471,
+      "step": 6612
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8912588091923556e-05,
+      "loss": 0.5919,
+      "step": 6613
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8912264803571765e-05,
+      "loss": 0.5638,
+      "step": 6614
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8911941469934143e-05,
+      "loss": 0.5856,
+      "step": 6615
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891161809101234e-05,
+      "loss": 0.6611,
+      "step": 6616
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8911294666807992e-05,
+      "loss": 0.5787,
+      "step": 6617
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891097119732275e-05,
+      "loss": 0.6159,
+      "step": 6618
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.891064768255825e-05,
+      "loss": 0.5755,
+      "step": 6619
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8910324122516143e-05,
+      "loss": 0.5557,
+      "step": 6620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8910000517198067e-05,
+      "loss": 0.6135,
+      "step": 6621
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890967686660567e-05,
+      "loss": 0.5773,
+      "step": 6622
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8909353170740596e-05,
+      "loss": 0.6386,
+      "step": 6623
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8909029429604486e-05,
+      "loss": 0.6093,
+      "step": 6624
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8908705643198993e-05,
+      "loss": 0.5818,
+      "step": 6625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8908381811525754e-05,
+      "loss": 0.5726,
+      "step": 6626
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890805793458642e-05,
+      "loss": 0.6573,
+      "step": 6627
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8907734012382634e-05,
+      "loss": 0.6197,
+      "step": 6628
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8907410044916045e-05,
+      "loss": 0.6484,
+      "step": 6629
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8907086032188295e-05,
+      "loss": 0.6488,
+      "step": 6630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8906761974201033e-05,
+      "loss": 0.6067,
+      "step": 6631
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8906437870955906e-05,
+      "loss": 0.5939,
+      "step": 6632
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8906113722454554e-05,
+      "loss": 0.6146,
+      "step": 6633
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8905789528698637e-05,
+      "loss": 0.6254,
+      "step": 6634
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890546528968979e-05,
+      "loss": 0.5883,
+      "step": 6635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890514100542967e-05,
+      "loss": 0.6141,
+      "step": 6636
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8904816675919918e-05,
+      "loss": 0.5806,
+      "step": 6637
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8904492301162185e-05,
+      "loss": 0.5413,
+      "step": 6638
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890416788115812e-05,
+      "loss": 0.5979,
+      "step": 6639
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890384341590937e-05,
+      "loss": 0.5858,
+      "step": 6640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8903518905417582e-05,
+      "loss": 0.6929,
+      "step": 6641
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890319434968441e-05,
+      "loss": 0.6708,
+      "step": 6642
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8902869748711494e-05,
+      "loss": 0.5938,
+      "step": 6643
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8902545102500494e-05,
+      "loss": 0.5913,
+      "step": 6644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8902220411053056e-05,
+      "loss": 0.6394,
+      "step": 6645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8901895674370826e-05,
+      "loss": 0.6146,
+      "step": 6646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8901570892455462e-05,
+      "loss": 0.6386,
+      "step": 6647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8901246065308606e-05,
+      "loss": 0.6436,
+      "step": 6648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890092119293191e-05,
+      "loss": 0.6363,
+      "step": 6649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890059627532703e-05,
+      "loss": 0.6411,
+      "step": 6650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.890027131249561e-05,
+      "loss": 0.604,
+      "step": 6651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889994630443931e-05,
+      "loss": 0.6263,
+      "step": 6652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8899621251159775e-05,
+      "loss": 0.6222,
+      "step": 6653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8899296152658654e-05,
+      "loss": 0.5862,
+      "step": 6654
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8898971008937604e-05,
+      "loss": 0.6382,
+      "step": 6655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889864581999828e-05,
+      "loss": 0.6065,
+      "step": 6656
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8898320585842328e-05,
+      "loss": 0.6903,
+      "step": 6657
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8897995306471405e-05,
+      "loss": 0.6541,
+      "step": 6658
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889766998188716e-05,
+      "loss": 0.5792,
+      "step": 6659
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889734461209125e-05,
+      "loss": 0.6079,
+      "step": 6660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8897019197085323e-05,
+      "loss": 0.5621,
+      "step": 6661
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8896693736871036e-05,
+      "loss": 0.5779,
+      "step": 6662
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8896368231450045e-05,
+      "loss": 0.6536,
+      "step": 6663
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8896042680824004e-05,
+      "loss": 0.5863,
+      "step": 6664
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8895717084994558e-05,
+      "loss": 0.6305,
+      "step": 6665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8895391443963372e-05,
+      "loss": 0.5879,
+      "step": 6666
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8895065757732096e-05,
+      "loss": 0.5968,
+      "step": 6667
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889474002630239e-05,
+      "loss": 0.5899,
+      "step": 6668
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.88944142496759e-05,
+      "loss": 0.6239,
+      "step": 6669
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889408842785429e-05,
+      "loss": 0.6021,
+      "step": 6670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8893762560839208e-05,
+      "loss": 0.6538,
+      "step": 6671
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8893436648632315e-05,
+      "loss": 0.6602,
+      "step": 6672
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8893110691235268e-05,
+      "loss": 0.6596,
+      "step": 6673
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889278468864972e-05,
+      "loss": 0.6216,
+      "step": 6674
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8892458640877326e-05,
+      "loss": 0.606,
+      "step": 6675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8892132547919748e-05,
+      "loss": 0.5947,
+      "step": 6676
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889180640977864e-05,
+      "loss": 0.6364,
+      "step": 6677
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889148022645566e-05,
+      "loss": 0.651,
+      "step": 6678
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8891153997952464e-05,
+      "loss": 0.6229,
+      "step": 6679
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.889082772427071e-05,
+      "loss": 0.6273,
+      "step": 6680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8890501405412057e-05,
+      "loss": 0.6258,
+      "step": 6681
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.8890175041378165e-05,
+      "loss": 0.6017,
+      "step": 6682
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.888984863217069e-05,
+      "loss": 0.6664,
+      "step": 6683
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.888952217779129e-05,
+      "loss": 0.5698,
+      "step": 6684
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.888919567824162e-05,
+      "loss": 0.6262,
+      "step": 6685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8888869133523346e-05,
+      "loss": 0.6125,
+      "step": 6686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8888542543638128e-05,
+      "loss": 0.5936,
+      "step": 6687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888821590858762e-05,
+      "loss": 0.5971,
+      "step": 6688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8887889228373482e-05,
+      "loss": 0.5759,
+      "step": 6689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8887562502997378e-05,
+      "loss": 0.5888,
+      "step": 6690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8887235732460964e-05,
+      "loss": 0.5807,
+      "step": 6691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8886908916765906e-05,
+      "loss": 0.6344,
+      "step": 6692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8886582055913858e-05,
+      "loss": 0.6177,
+      "step": 6693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8886255149906484e-05,
+      "loss": 0.6081,
+      "step": 6694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8885928198745447e-05,
+      "loss": 0.5983,
+      "step": 6695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8885601202432408e-05,
+      "loss": 0.6374,
+      "step": 6696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8885274160969022e-05,
+      "loss": 0.5717,
+      "step": 6697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888494707435696e-05,
+      "loss": 0.6615,
+      "step": 6698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888461994259788e-05,
+      "loss": 0.6269,
+      "step": 6699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888429276569344e-05,
+      "loss": 0.6276,
+      "step": 6700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888396554364531e-05,
+      "loss": 0.6469,
+      "step": 6701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8883638276455146e-05,
+      "loss": 0.6423,
+      "step": 6702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8883310964124617e-05,
+      "loss": 0.5676,
+      "step": 6703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8882983606655384e-05,
+      "loss": 0.6773,
+      "step": 6704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8882656204049106e-05,
+      "loss": 0.5913,
+      "step": 6705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.888232875630746e-05,
+      "loss": 0.5661,
+      "step": 6706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8882001263432087e-05,
+      "loss": 0.5934,
+      "step": 6707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8881673725424673e-05,
+      "loss": 0.6572,
+      "step": 6708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8881346142286872e-05,
+      "loss": 0.5573,
+      "step": 6709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8881018514020352e-05,
+      "loss": 0.6094,
+      "step": 6710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8880690840626772e-05,
+      "loss": 0.6252,
+      "step": 6711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88803631221078e-05,
+      "loss": 0.6259,
+      "step": 6712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8880035358465107e-05,
+      "loss": 0.6051,
+      "step": 6713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887970754970035e-05,
+      "loss": 0.6197,
+      "step": 6714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88793796958152e-05,
+      "loss": 0.6098,
+      "step": 6715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8879051796811323e-05,
+      "loss": 0.5644,
+      "step": 6716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8878723852690384e-05,
+      "loss": 0.6267,
+      "step": 6717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8878395863454045e-05,
+      "loss": 0.593,
+      "step": 6718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8878067829103978e-05,
+      "loss": 0.6123,
+      "step": 6719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887773974964185e-05,
+      "loss": 0.6059,
+      "step": 6720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8877411625069325e-05,
+      "loss": 0.6547,
+      "step": 6721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887708345538807e-05,
+      "loss": 0.6198,
+      "step": 6722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8876755240599758e-05,
+      "loss": 0.5497,
+      "step": 6723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8876426980706048e-05,
+      "loss": 0.6484,
+      "step": 6724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8876098675708616e-05,
+      "loss": 0.5819,
+      "step": 6725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8875770325609126e-05,
+      "loss": 0.6017,
+      "step": 6726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8875441930409247e-05,
+      "loss": 0.6049,
+      "step": 6727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887511349011065e-05,
+      "loss": 0.6974,
+      "step": 6728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8874785004715005e-05,
+      "loss": 0.5943,
+      "step": 6729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8874456474223972e-05,
+      "loss": 0.6073,
+      "step": 6730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8874127898639233e-05,
+      "loss": 0.6262,
+      "step": 6731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8873799277962447e-05,
+      "loss": 0.6374,
+      "step": 6732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8873470612195288e-05,
+      "loss": 0.6066,
+      "step": 6733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887314190133943e-05,
+      "loss": 0.6675,
+      "step": 6734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8872813145396535e-05,
+      "loss": 0.626,
+      "step": 6735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8872484344368277e-05,
+      "loss": 0.5907,
+      "step": 6736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887215549825633e-05,
+      "loss": 0.6514,
+      "step": 6737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8871826607062363e-05,
+      "loss": 0.6451,
+      "step": 6738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8871497670788046e-05,
+      "loss": 0.5806,
+      "step": 6739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8871168689435053e-05,
+      "loss": 0.5951,
+      "step": 6740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887083966300505e-05,
+      "loss": 0.5728,
+      "step": 6741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8870510591499718e-05,
+      "loss": 0.6024,
+      "step": 6742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.887018147492072e-05,
+      "loss": 0.606,
+      "step": 6743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8869852313269734e-05,
+      "loss": 0.6168,
+      "step": 6744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886952310654843e-05,
+      "loss": 0.6276,
+      "step": 6745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886919385475848e-05,
+      "loss": 0.6514,
+      "step": 6746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886886455790156e-05,
+      "loss": 0.5817,
+      "step": 6747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886853521597934e-05,
+      "loss": 0.5591,
+      "step": 6748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88682058289935e-05,
+      "loss": 0.6294,
+      "step": 6749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8867876396945704e-05,
+      "loss": 0.5981,
+      "step": 6750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8867546919837632e-05,
+      "loss": 0.6419,
+      "step": 6751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8867217397670962e-05,
+      "loss": 0.5986,
+      "step": 6752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886688783044736e-05,
+      "loss": 0.6152,
+      "step": 6753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88665582181685e-05,
+      "loss": 0.6207,
+      "step": 6754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8866228560836065e-05,
+      "loss": 0.621,
+      "step": 6755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8865898858451727e-05,
+      "loss": 0.7035,
+      "step": 6756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886556911101716e-05,
+      "loss": 0.5722,
+      "step": 6757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8865239318534038e-05,
+      "loss": 0.5329,
+      "step": 6758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886490948100404e-05,
+      "loss": 0.5631,
+      "step": 6759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886457959842884e-05,
+      "loss": 0.6369,
+      "step": 6760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8864249670810116e-05,
+      "loss": 0.5487,
+      "step": 6761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8863919698149545e-05,
+      "loss": 0.6929,
+      "step": 6762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88635896804488e-05,
+      "loss": 0.6359,
+      "step": 6763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886325961770956e-05,
+      "loss": 0.6256,
+      "step": 6764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88629295099335e-05,
+      "loss": 0.5827,
+      "step": 6765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8862599357122303e-05,
+      "loss": 0.6741,
+      "step": 6766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8862269159277643e-05,
+      "loss": 0.6243,
+      "step": 6767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8861938916401195e-05,
+      "loss": 0.5954,
+      "step": 6768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8861608628494643e-05,
+      "loss": 0.6502,
+      "step": 6769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886127829555966e-05,
+      "loss": 0.6202,
+      "step": 6770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.886094791759793e-05,
+      "loss": 0.6472,
+      "step": 6771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8860617494611127e-05,
+      "loss": 0.6482,
+      "step": 6772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8860287026600928e-05,
+      "loss": 0.5991,
+      "step": 6773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885995651356902e-05,
+      "loss": 0.6558,
+      "step": 6774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8859625955517078e-05,
+      "loss": 0.6247,
+      "step": 6775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885929535244678e-05,
+      "loss": 0.596,
+      "step": 6776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885896470435981e-05,
+      "loss": 0.5932,
+      "step": 6777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8858634011257843e-05,
+      "loss": 0.5699,
+      "step": 6778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8858303273142566e-05,
+      "loss": 0.5559,
+      "step": 6779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8857972490015653e-05,
+      "loss": 0.5879,
+      "step": 6780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8857641661878788e-05,
+      "loss": 0.5982,
+      "step": 6781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8857310788733652e-05,
+      "loss": 0.6506,
+      "step": 6782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8856979870581925e-05,
+      "loss": 0.6991,
+      "step": 6783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885664890742529e-05,
+      "loss": 0.6324,
+      "step": 6784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885631789926543e-05,
+      "loss": 0.6239,
+      "step": 6785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8855986846104022e-05,
+      "loss": 0.5925,
+      "step": 6786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885565574794275e-05,
+      "loss": 0.5341,
+      "step": 6787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8855324604783304e-05,
+      "loss": 0.6877,
+      "step": 6788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8854993416627357e-05,
+      "loss": 0.5987,
+      "step": 6789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8854662183476593e-05,
+      "loss": 0.576,
+      "step": 6790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88543309053327e-05,
+      "loss": 0.6521,
+      "step": 6791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8853999582197356e-05,
+      "loss": 0.6639,
+      "step": 6792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885366821407225e-05,
+      "loss": 0.5805,
+      "step": 6793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885333680095906e-05,
+      "loss": 0.6083,
+      "step": 6794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8853005342859477e-05,
+      "loss": 0.6145,
+      "step": 6795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885267383977518e-05,
+      "loss": 0.5997,
+      "step": 6796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885234229170785e-05,
+      "loss": 0.6428,
+      "step": 6797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885201069865918e-05,
+      "loss": 0.5809,
+      "step": 6798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885167906063085e-05,
+      "loss": 0.6509,
+      "step": 6799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8851347377624546e-05,
+      "loss": 0.5904,
+      "step": 6800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8851015649641955e-05,
+      "loss": 0.6626,
+      "step": 6801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885068387668476e-05,
+      "loss": 0.5616,
+      "step": 6802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.885035205875465e-05,
+      "loss": 0.6289,
+      "step": 6803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8850020195853308e-05,
+      "loss": 0.6182,
+      "step": 6804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8849688287982424e-05,
+      "loss": 0.6319,
+      "step": 6805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884935633514368e-05,
+      "loss": 0.6385,
+      "step": 6806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8849024337338763e-05,
+      "loss": 0.6129,
+      "step": 6807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8848692294569364e-05,
+      "loss": 0.6861,
+      "step": 6808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884836020683717e-05,
+      "loss": 0.624,
+      "step": 6809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8848028074143863e-05,
+      "loss": 0.6055,
+      "step": 6810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8847695896491135e-05,
+      "loss": 0.5952,
+      "step": 6811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884736367388067e-05,
+      "loss": 0.5831,
+      "step": 6812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8847031406314165e-05,
+      "loss": 0.6557,
+      "step": 6813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88466990937933e-05,
+      "loss": 0.6257,
+      "step": 6814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8846366736319764e-05,
+      "loss": 0.5955,
+      "step": 6815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884603433389525e-05,
+      "loss": 0.617,
+      "step": 6816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8845701886521445e-05,
+      "loss": 0.6098,
+      "step": 6817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8845369394200036e-05,
+      "loss": 0.584,
+      "step": 6818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8845036856932718e-05,
+      "loss": 0.6244,
+      "step": 6819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8844704274721174e-05,
+      "loss": 0.6136,
+      "step": 6820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.88443716475671e-05,
+      "loss": 0.6515,
+      "step": 6821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8844038975472184e-05,
+      "loss": 0.6461,
+      "step": 6822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884370625843811e-05,
+      "loss": 0.6237,
+      "step": 6823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884337349646658e-05,
+      "loss": 0.6114,
+      "step": 6824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884304068955928e-05,
+      "loss": 0.6388,
+      "step": 6825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8842707837717896e-05,
+      "loss": 0.603,
+      "step": 6826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8842374940944127e-05,
+      "loss": 0.5881,
+      "step": 6827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884204199923966e-05,
+      "loss": 0.6271,
+      "step": 6828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.884170901260619e-05,
+      "loss": 0.6409,
+      "step": 6829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8841375981045405e-05,
+      "loss": 0.5961,
+      "step": 6830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8841042904559003e-05,
+      "loss": 0.5715,
+      "step": 6831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.8840709783148667e-05,
+      "loss": 0.5573,
+      "step": 6832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.88403766168161e-05,
+      "loss": 0.6009,
+      "step": 6833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8840043405562987e-05,
+      "loss": 0.6389,
+      "step": 6834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8839710149391027e-05,
+      "loss": 0.6054,
+      "step": 6835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883937684830191e-05,
+      "loss": 0.6195,
+      "step": 6836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883904350229733e-05,
+      "loss": 0.6183,
+      "step": 6837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883871011137898e-05,
+      "loss": 0.6374,
+      "step": 6838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883837667554856e-05,
+      "loss": 0.6009,
+      "step": 6839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8838043194807755e-05,
+      "loss": 0.5829,
+      "step": 6840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8837709669158267e-05,
+      "loss": 0.5726,
+      "step": 6841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883737609860179e-05,
+      "loss": 0.6455,
+      "step": 6842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883704248314001e-05,
+      "loss": 0.59,
+      "step": 6843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8836708822774636e-05,
+      "loss": 0.5543,
+      "step": 6844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883637511750735e-05,
+      "loss": 0.5996,
+      "step": 6845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8836041367339862e-05,
+      "loss": 0.6274,
+      "step": 6846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8835707572273852e-05,
+      "loss": 0.6443,
+      "step": 6847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8835373732311027e-05,
+      "loss": 0.614,
+      "step": 6848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8835039847453084e-05,
+      "loss": 0.5966,
+      "step": 6849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883470591770171e-05,
+      "loss": 0.6146,
+      "step": 6850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8834371943058617e-05,
+      "loss": 0.5849,
+      "step": 6851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8834037923525485e-05,
+      "loss": 0.5747,
+      "step": 6852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883370385910402e-05,
+      "loss": 0.573,
+      "step": 6853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8833369749795923e-05,
+      "loss": 0.6094,
+      "step": 6854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8833035595602882e-05,
+      "loss": 0.5838,
+      "step": 6855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8832701396526602e-05,
+      "loss": 0.6796,
+      "step": 6856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8832367152568783e-05,
+      "loss": 0.5742,
+      "step": 6857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8832032863731117e-05,
+      "loss": 0.6361,
+      "step": 6858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8831698530015304e-05,
+      "loss": 0.5931,
+      "step": 6859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8831364151423045e-05,
+      "loss": 0.5747,
+      "step": 6860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.883102972795604e-05,
+      "loss": 0.6363,
+      "step": 6861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8830695259615987e-05,
+      "loss": 0.634,
+      "step": 6862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8830360746404582e-05,
+      "loss": 0.5945,
+      "step": 6863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8830026188323527e-05,
+      "loss": 0.6242,
+      "step": 6864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882969158537453e-05,
+      "loss": 0.6258,
+      "step": 6865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882935693755928e-05,
+      "loss": 0.6627,
+      "step": 6866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882902224487948e-05,
+      "loss": 0.6105,
+      "step": 6867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8828687507336837e-05,
+      "loss": 0.6467,
+      "step": 6868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8828352724933042e-05,
+      "loss": 0.5835,
+      "step": 6869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.88280178976698e-05,
+      "loss": 0.6214,
+      "step": 6870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882768302554882e-05,
+      "loss": 0.629,
+      "step": 6871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8827348108571795e-05,
+      "loss": 0.5916,
+      "step": 6872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8827013146740428e-05,
+      "loss": 0.6358,
+      "step": 6873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8826678140056424e-05,
+      "loss": 0.639,
+      "step": 6874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882634308852148e-05,
+      "loss": 0.5651,
+      "step": 6875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8826007992137302e-05,
+      "loss": 0.641,
+      "step": 6876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8825672850905597e-05,
+      "loss": 0.6871,
+      "step": 6877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8825337664828058e-05,
+      "loss": 0.6372,
+      "step": 6878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8825002433906396e-05,
+      "loss": 0.589,
+      "step": 6879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8824667158142314e-05,
+      "loss": 0.5843,
+      "step": 6880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882433183753751e-05,
+      "loss": 0.6467,
+      "step": 6881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8823996472093695e-05,
+      "loss": 0.619,
+      "step": 6882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882366106181257e-05,
+      "loss": 0.6044,
+      "step": 6883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8823325606695837e-05,
+      "loss": 0.6291,
+      "step": 6884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.88229901067452e-05,
+      "loss": 0.6491,
+      "step": 6885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8822654561962372e-05,
+      "loss": 0.5945,
+      "step": 6886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882231897234905e-05,
+      "loss": 0.6395,
+      "step": 6887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8821983337906942e-05,
+      "loss": 0.6446,
+      "step": 6888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.882164765863775e-05,
+      "loss": 0.5883,
+      "step": 6889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8821311934543186e-05,
+      "loss": 0.6295,
+      "step": 6890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8820976165624952e-05,
+      "loss": 0.5849,
+      "step": 6891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8820640351884757e-05,
+      "loss": 0.6066,
+      "step": 6892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.88203044933243e-05,
+      "loss": 0.6487,
+      "step": 6893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8819968589945294e-05,
+      "loss": 0.6183,
+      "step": 6894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8819632641749447e-05,
+      "loss": 0.6735,
+      "step": 6895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8819296648738462e-05,
+      "loss": 0.633,
+      "step": 6896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.881896061091405e-05,
+      "loss": 0.5866,
+      "step": 6897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8818624528277914e-05,
+      "loss": 0.6157,
+      "step": 6898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8818288400831765e-05,
+      "loss": 0.6113,
+      "step": 6899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.881795222857731e-05,
+      "loss": 0.6398,
+      "step": 6900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8817616011516258e-05,
+      "loss": 0.5928,
+      "step": 6901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8817279749650314e-05,
+      "loss": 0.6078,
+      "step": 6902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.881694344298119e-05,
+      "loss": 0.5595,
+      "step": 6903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8816607091510598e-05,
+      "loss": 0.645,
+      "step": 6904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.881627069524024e-05,
+      "loss": 0.576,
+      "step": 6905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.881593425417183e-05,
+      "loss": 0.5658,
+      "step": 6906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8815597768307074e-05,
+      "loss": 0.6722,
+      "step": 6907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8815261237647686e-05,
+      "loss": 0.6394,
+      "step": 6908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8814924662195377e-05,
+      "loss": 0.6109,
+      "step": 6909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8814588041951846e-05,
+      "loss": 0.6189,
+      "step": 6910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8814251376918816e-05,
+      "loss": 0.5792,
+      "step": 6911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8813914667097997e-05,
+      "loss": 0.5626,
+      "step": 6912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8813577912491094e-05,
+      "loss": 0.6461,
+      "step": 6913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8813241113099817e-05,
+      "loss": 0.6356,
+      "step": 6914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8812904268925885e-05,
+      "loss": 0.5582,
+      "step": 6915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8812567379971004e-05,
+      "loss": 0.6403,
+      "step": 6916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8812230446236887e-05,
+      "loss": 0.6618,
+      "step": 6917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8811893467725247e-05,
+      "loss": 0.5779,
+      "step": 6918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8811556444437792e-05,
+      "loss": 0.5772,
+      "step": 6919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8811219376376243e-05,
+      "loss": 0.581,
+      "step": 6920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8810882263542305e-05,
+      "loss": 0.6572,
+      "step": 6921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8810545105937694e-05,
+      "loss": 0.5748,
+      "step": 6922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8810207903564123e-05,
+      "loss": 0.6529,
+      "step": 6923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880987065642331e-05,
+      "loss": 0.5497,
+      "step": 6924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880953336451696e-05,
+      "loss": 0.5977,
+      "step": 6925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880919602784679e-05,
+      "loss": 0.6423,
+      "step": 6926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8808858646414516e-05,
+      "loss": 0.6184,
+      "step": 6927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8808521220221853e-05,
+      "loss": 0.6837,
+      "step": 6928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8808183749270513e-05,
+      "loss": 0.6308,
+      "step": 6929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880784623356221e-05,
+      "loss": 0.6129,
+      "step": 6930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8807508673098663e-05,
+      "loss": 0.5967,
+      "step": 6931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8807171067881582e-05,
+      "loss": 0.6039,
+      "step": 6932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880683341791269e-05,
+      "loss": 0.6406,
+      "step": 6933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8806495723193695e-05,
+      "loss": 0.6067,
+      "step": 6934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880615798372632e-05,
+      "loss": 0.7188,
+      "step": 6935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880582019951227e-05,
+      "loss": 0.6141,
+      "step": 6936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8805482370553278e-05,
+      "loss": 0.5638,
+      "step": 6937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8805144496851046e-05,
+      "loss": 0.5715,
+      "step": 6938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8804806578407297e-05,
+      "loss": 0.5672,
+      "step": 6939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8804468615223745e-05,
+      "loss": 0.5962,
+      "step": 6940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8804130607302112e-05,
+      "loss": 0.6216,
+      "step": 6941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8803792554644112e-05,
+      "loss": 0.6209,
+      "step": 6942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8803454457251463e-05,
+      "loss": 0.64,
+      "step": 6943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8803116315125886e-05,
+      "loss": 0.6421,
+      "step": 6944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8802778128269095e-05,
+      "loss": 0.5886,
+      "step": 6945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8802439896682808e-05,
+      "loss": 0.6363,
+      "step": 6946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880210162036875e-05,
+      "loss": 0.6414,
+      "step": 6947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8801763299328635e-05,
+      "loss": 0.5742,
+      "step": 6948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.880142493356418e-05,
+      "loss": 0.6691,
+      "step": 6949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8801086523077113e-05,
+      "loss": 0.5447,
+      "step": 6950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8800748067869144e-05,
+      "loss": 0.6488,
+      "step": 6951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8800409567941996e-05,
+      "loss": 0.5791,
+      "step": 6952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8800071023297394e-05,
+      "loss": 0.6633,
+      "step": 6953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879973243393705e-05,
+      "loss": 0.5564,
+      "step": 6954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879939379986269e-05,
+      "loss": 0.5875,
+      "step": 6955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879905512107603e-05,
+      "loss": 0.6072,
+      "step": 6956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8798716397578797e-05,
+      "loss": 0.6436,
+      "step": 6957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879837762937271e-05,
+      "loss": 0.6506,
+      "step": 6958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879803881645949e-05,
+      "loss": 0.6233,
+      "step": 6959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8797699958840854e-05,
+      "loss": 0.6251,
+      "step": 6960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879736105651853e-05,
+      "loss": 0.6253,
+      "step": 6961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8797022109494238e-05,
+      "loss": 0.6599,
+      "step": 6962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8796683117769702e-05,
+      "loss": 0.5814,
+      "step": 6963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8796344081346644e-05,
+      "loss": 0.6265,
+      "step": 6964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879600500022678e-05,
+      "loss": 0.5982,
+      "step": 6965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8795665874411845e-05,
+      "loss": 0.6116,
+      "step": 6966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879532670390355e-05,
+      "loss": 0.6399,
+      "step": 6967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8794987488703628e-05,
+      "loss": 0.5935,
+      "step": 6968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8794648228813797e-05,
+      "loss": 0.6575,
+      "step": 6969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879430892423578e-05,
+      "loss": 0.5825,
+      "step": 6970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879396957497131e-05,
+      "loss": 0.6096,
+      "step": 6971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8793630181022103e-05,
+      "loss": 0.5803,
+      "step": 6972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879329074238988e-05,
+      "loss": 0.6421,
+      "step": 6973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879295125907638e-05,
+      "loss": 0.5469,
+      "step": 6974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8792611731083316e-05,
+      "loss": 0.6753,
+      "step": 6975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8792272158412417e-05,
+      "loss": 0.5616,
+      "step": 6976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.879193254106541e-05,
+      "loss": 0.6384,
+      "step": 6977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8791592879044013e-05,
+      "loss": 0.6345,
+      "step": 6978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.8791253172349963e-05,
+      "loss": 0.6401,
+      "step": 6979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.879091342098498e-05,
+      "loss": 0.6405,
+      "step": 6980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.879057362495079e-05,
+      "loss": 0.5921,
+      "step": 6981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8790233784249122e-05,
+      "loss": 0.5968,
+      "step": 6982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8789893898881703e-05,
+      "loss": 0.6971,
+      "step": 6983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878955396885026e-05,
+      "loss": 0.6368,
+      "step": 6984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8789213994156514e-05,
+      "loss": 0.6041,
+      "step": 6985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87888739748022e-05,
+      "loss": 0.6591,
+      "step": 6986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8788533910789044e-05,
+      "loss": 0.5862,
+      "step": 6987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8788193802118773e-05,
+      "loss": 0.5925,
+      "step": 6988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8787853648793114e-05,
+      "loss": 0.5838,
+      "step": 6989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87875134508138e-05,
+      "loss": 0.6102,
+      "step": 6990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8787173208182556e-05,
+      "loss": 0.6181,
+      "step": 6991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878683292090111e-05,
+      "loss": 0.6116,
+      "step": 6992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8786492588971196e-05,
+      "loss": 0.6535,
+      "step": 6993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8786152212394537e-05,
+      "loss": 0.6203,
+      "step": 6994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8785811791172865e-05,
+      "loss": 0.6152,
+      "step": 6995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8785471325307912e-05,
+      "loss": 0.6203,
+      "step": 6996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8785130814801405e-05,
+      "loss": 0.5981,
+      "step": 6997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8784790259655073e-05,
+      "loss": 0.7109,
+      "step": 6998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8784449659870653e-05,
+      "loss": 0.6532,
+      "step": 6999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878410901544987e-05,
+      "loss": 0.6502,
+      "step": 7000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8783768326394457e-05,
+      "loss": 0.6349,
+      "step": 7001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8783427592706146e-05,
+      "loss": 0.637,
+      "step": 7002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8783086814386663e-05,
+      "loss": 0.6935,
+      "step": 7003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878274599143775e-05,
+      "loss": 0.6753,
+      "step": 7004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878240512386113e-05,
+      "loss": 0.5947,
+      "step": 7005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8782064211658533e-05,
+      "loss": 0.6252,
+      "step": 7006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8781723254831698e-05,
+      "loss": 0.573,
+      "step": 7007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8781382253382358e-05,
+      "loss": 0.5966,
+      "step": 7008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878104120731224e-05,
+      "loss": 0.6315,
+      "step": 7009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8780700116623082e-05,
+      "loss": 0.5869,
+      "step": 7010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8780358981316613e-05,
+      "loss": 0.6385,
+      "step": 7011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.878001780139457e-05,
+      "loss": 0.6385,
+      "step": 7012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8779676576858687e-05,
+      "loss": 0.5973,
+      "step": 7013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8779335307710692e-05,
+      "loss": 0.5997,
+      "step": 7014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8778993993952328e-05,
+      "loss": 0.5734,
+      "step": 7015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8778652635585323e-05,
+      "loss": 0.6129,
+      "step": 7016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877831123261141e-05,
+      "loss": 0.6494,
+      "step": 7017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8777969785032328e-05,
+      "loss": 0.61,
+      "step": 7018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8777628292849815e-05,
+      "loss": 0.6136,
+      "step": 7019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8777286756065596e-05,
+      "loss": 0.6058,
+      "step": 7020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8776945174681412e-05,
+      "loss": 0.7137,
+      "step": 7021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8776603548699003e-05,
+      "loss": 0.6403,
+      "step": 7022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8776261878120102e-05,
+      "loss": 0.6641,
+      "step": 7023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8775920162946444e-05,
+      "loss": 0.5964,
+      "step": 7024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877557840317976e-05,
+      "loss": 0.6478,
+      "step": 7025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87752365988218e-05,
+      "loss": 0.6193,
+      "step": 7026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877489474987429e-05,
+      "loss": 0.6563,
+      "step": 7027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8774552856338966e-05,
+      "loss": 0.6142,
+      "step": 7028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877421091821757e-05,
+      "loss": 0.5839,
+      "step": 7029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8773868935511843e-05,
+      "loss": 0.6077,
+      "step": 7030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8773526908223516e-05,
+      "loss": 0.6836,
+      "step": 7031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8773184836354326e-05,
+      "loss": 0.5803,
+      "step": 7032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877284271990602e-05,
+      "loss": 0.6139,
+      "step": 7033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8772500558880328e-05,
+      "loss": 0.5639,
+      "step": 7034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8772158353278992e-05,
+      "loss": 0.6838,
+      "step": 7035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877181610310375e-05,
+      "loss": 0.5522,
+      "step": 7036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8771473808356343e-05,
+      "loss": 0.6377,
+      "step": 7037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.877113146903851e-05,
+      "loss": 0.6254,
+      "step": 7038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8770789085151985e-05,
+      "loss": 0.5497,
+      "step": 7039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8770446656698516e-05,
+      "loss": 0.6133,
+      "step": 7040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8770104183679833e-05,
+      "loss": 0.5788,
+      "step": 7041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8769761666097688e-05,
+      "loss": 0.5978,
+      "step": 7042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8769419103953815e-05,
+      "loss": 0.5343,
+      "step": 7043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8769076497249954e-05,
+      "loss": 0.6244,
+      "step": 7044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8768733845987846e-05,
+      "loss": 0.6385,
+      "step": 7045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8768391150169235e-05,
+      "loss": 0.5777,
+      "step": 7046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876804840979586e-05,
+      "loss": 0.5654,
+      "step": 7047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8767705624869463e-05,
+      "loss": 0.6582,
+      "step": 7048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8767362795391787e-05,
+      "loss": 0.591,
+      "step": 7049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8767019921364575e-05,
+      "loss": 0.6103,
+      "step": 7050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8766677002789562e-05,
+      "loss": 0.6536,
+      "step": 7051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8766334039668498e-05,
+      "loss": 0.6185,
+      "step": 7052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8765991032003122e-05,
+      "loss": 0.6887,
+      "step": 7053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8765647979795184e-05,
+      "loss": 0.6105,
+      "step": 7054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8765304883046415e-05,
+      "loss": 0.6203,
+      "step": 7055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876496174175857e-05,
+      "loss": 0.6714,
+      "step": 7056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8764618555933384e-05,
+      "loss": 0.6077,
+      "step": 7057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87642753255726e-05,
+      "loss": 0.5977,
+      "step": 7058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8763932050677977e-05,
+      "loss": 0.5924,
+      "step": 7059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876358873125124e-05,
+      "loss": 0.6352,
+      "step": 7060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8763245367294147e-05,
+      "loss": 0.6441,
+      "step": 7061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8762901958808434e-05,
+      "loss": 0.668,
+      "step": 7062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8762558505795852e-05,
+      "loss": 0.5938,
+      "step": 7063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8762215008258143e-05,
+      "loss": 0.6254,
+      "step": 7064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8761871466197056e-05,
+      "loss": 0.6354,
+      "step": 7065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876152787961433e-05,
+      "loss": 0.6452,
+      "step": 7066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876118424851172e-05,
+      "loss": 0.5582,
+      "step": 7067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8760840572890965e-05,
+      "loss": 0.6197,
+      "step": 7068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876049685275381e-05,
+      "loss": 0.6124,
+      "step": 7069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.876015308810201e-05,
+      "loss": 0.618,
+      "step": 7070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87598092789373e-05,
+      "loss": 0.5195,
+      "step": 7071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875946542526144e-05,
+      "loss": 0.5743,
+      "step": 7072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875912152707617e-05,
+      "loss": 0.6017,
+      "step": 7073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8758777584383234e-05,
+      "loss": 0.6327,
+      "step": 7074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8758433597184388e-05,
+      "loss": 0.6086,
+      "step": 7075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8758089565481373e-05,
+      "loss": 0.5907,
+      "step": 7076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8757745489275946e-05,
+      "loss": 0.563,
+      "step": 7077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875740136856984e-05,
+      "loss": 0.5941,
+      "step": 7078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875705720336482e-05,
+      "loss": 0.6565,
+      "step": 7079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8756712993662627e-05,
+      "loss": 0.603,
+      "step": 7080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8756368739465006e-05,
+      "loss": 0.5838,
+      "step": 7081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8756024440773717e-05,
+      "loss": 0.6472,
+      "step": 7082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87556800975905e-05,
+      "loss": 0.5987,
+      "step": 7083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875533570991711e-05,
+      "loss": 0.6887,
+      "step": 7084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8754991277755295e-05,
+      "loss": 0.6823,
+      "step": 7085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8754646801106805e-05,
+      "loss": 0.6093,
+      "step": 7086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8754302279973388e-05,
+      "loss": 0.5606,
+      "step": 7087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87539577143568e-05,
+      "loss": 0.6831,
+      "step": 7088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8753613104258787e-05,
+      "loss": 0.5999,
+      "step": 7089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8753268449681106e-05,
+      "loss": 0.6299,
+      "step": 7090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87529237506255e-05,
+      "loss": 0.6228,
+      "step": 7091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8752579007093728e-05,
+      "loss": 0.6387,
+      "step": 7092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8752234219087538e-05,
+      "loss": 0.5875,
+      "step": 7093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8751889386608682e-05,
+      "loss": 0.6622,
+      "step": 7094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8751544509658917e-05,
+      "loss": 0.5906,
+      "step": 7095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.875119958823999e-05,
+      "loss": 0.5733,
+      "step": 7096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8750854622353648e-05,
+      "loss": 0.6501,
+      "step": 7097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8750509612001658e-05,
+      "loss": 0.5724,
+      "step": 7098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8750164557185764e-05,
+      "loss": 0.6142,
+      "step": 7099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8749819457907722e-05,
+      "loss": 0.6491,
+      "step": 7100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8749474314169283e-05,
+      "loss": 0.6416,
+      "step": 7101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8749129125972203e-05,
+      "loss": 0.6094,
+      "step": 7102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8748783893318235e-05,
+      "loss": 0.6009,
+      "step": 7103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8748438616209136e-05,
+      "loss": 0.5896,
+      "step": 7104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8748093294646657e-05,
+      "loss": 0.6279,
+      "step": 7105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8747747928632556e-05,
+      "loss": 0.6077,
+      "step": 7106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8747402518168582e-05,
+      "loss": 0.5798,
+      "step": 7107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8747057063256496e-05,
+      "loss": 0.6039,
+      "step": 7108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.874671156389805e-05,
+      "loss": 0.6681,
+      "step": 7109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8746366020095004e-05,
+      "loss": 0.6214,
+      "step": 7110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.874602043184911e-05,
+      "loss": 0.6486,
+      "step": 7111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8745674799162123e-05,
+      "loss": 0.5963,
+      "step": 7112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8745329122035802e-05,
+      "loss": 0.665,
+      "step": 7113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8744983400471903e-05,
+      "loss": 0.6176,
+      "step": 7114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8744637634472182e-05,
+      "loss": 0.5579,
+      "step": 7115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8744291824038396e-05,
+      "loss": 0.5846,
+      "step": 7116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.87439459691723e-05,
+      "loss": 0.5418,
+      "step": 7117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.874360006987566e-05,
+      "loss": 0.59,
+      "step": 7118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8743254126150224e-05,
+      "loss": 0.6073,
+      "step": 7119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8742908137997752e-05,
+      "loss": 0.5985,
+      "step": 7120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8742562105420005e-05,
+      "loss": 0.6159,
+      "step": 7121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8742216028418737e-05,
+      "loss": 0.5954,
+      "step": 7122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.874186990699571e-05,
+      "loss": 0.641,
+      "step": 7123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8741523741152683e-05,
+      "loss": 0.629,
+      "step": 7124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8741177530891415e-05,
+      "loss": 0.6196,
+      "step": 7125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.8740831276213663e-05,
+      "loss": 0.6325,
+      "step": 7126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8740484977121186e-05,
+      "loss": 0.6047,
+      "step": 7127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8740138633615747e-05,
+      "loss": 0.6254,
+      "step": 7128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.87397922456991e-05,
+      "loss": 0.5591,
+      "step": 7129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8739445813373012e-05,
+      "loss": 0.5938,
+      "step": 7130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.873909933663924e-05,
+      "loss": 0.6797,
+      "step": 7131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8738752815499547e-05,
+      "loss": 0.5872,
+      "step": 7132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.873840624995569e-05,
+      "loss": 0.6246,
+      "step": 7133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.873805964000943e-05,
+      "loss": 0.5686,
+      "step": 7134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8737712985662528e-05,
+      "loss": 0.6001,
+      "step": 7135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8737366286916752e-05,
+      "loss": 0.6602,
+      "step": 7136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8737019543773857e-05,
+      "loss": 0.6601,
+      "step": 7137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8736672756235606e-05,
+      "loss": 0.6117,
+      "step": 7138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8736325924303766e-05,
+      "loss": 0.574,
+      "step": 7139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8735979047980094e-05,
+      "loss": 0.5929,
+      "step": 7140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8735632127266353e-05,
+      "loss": 0.6242,
+      "step": 7141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8735285162164304e-05,
+      "loss": 0.593,
+      "step": 7142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8734938152675717e-05,
+      "loss": 0.6158,
+      "step": 7143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8734591098802352e-05,
+      "loss": 0.6782,
+      "step": 7144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8734244000545968e-05,
+      "loss": 0.6323,
+      "step": 7145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8733896857908335e-05,
+      "loss": 0.6305,
+      "step": 7146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8733549670891213e-05,
+      "loss": 0.6065,
+      "step": 7147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8733202439496368e-05,
+      "loss": 0.5359,
+      "step": 7148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.873285516372556e-05,
+      "loss": 0.6059,
+      "step": 7149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8732507843580563e-05,
+      "loss": 0.5962,
+      "step": 7150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8732160479063133e-05,
+      "loss": 0.6354,
+      "step": 7151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8731813070175038e-05,
+      "loss": 0.5876,
+      "step": 7152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8731465616918046e-05,
+      "loss": 0.5841,
+      "step": 7153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8731118119293918e-05,
+      "loss": 0.5995,
+      "step": 7154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.873077057730442e-05,
+      "loss": 0.6091,
+      "step": 7155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8730422990951326e-05,
+      "loss": 0.6416,
+      "step": 7156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8730075360236394e-05,
+      "loss": 0.641,
+      "step": 7157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872972768516139e-05,
+      "loss": 0.6355,
+      "step": 7158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872937996572808e-05,
+      "loss": 0.6242,
+      "step": 7159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872903220193824e-05,
+      "loss": 0.5668,
+      "step": 7160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872868439379363e-05,
+      "loss": 0.6391,
+      "step": 7161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8728336541296012e-05,
+      "loss": 0.6288,
+      "step": 7162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8727988644447167e-05,
+      "loss": 0.6318,
+      "step": 7163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8727640703248853e-05,
+      "loss": 0.6007,
+      "step": 7164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8727292717702838e-05,
+      "loss": 0.5358,
+      "step": 7165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8726944687810898e-05,
+      "loss": 0.6306,
+      "step": 7166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8726596613574794e-05,
+      "loss": 0.6614,
+      "step": 7167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8726248494996295e-05,
+      "loss": 0.5746,
+      "step": 7168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8725900332077173e-05,
+      "loss": 0.6295,
+      "step": 7169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8725552124819196e-05,
+      "loss": 0.6106,
+      "step": 7170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8725203873224137e-05,
+      "loss": 0.6454,
+      "step": 7171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8724855577293758e-05,
+      "loss": 0.605,
+      "step": 7172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8724507237029833e-05,
+      "loss": 0.5874,
+      "step": 7173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8724158852434132e-05,
+      "loss": 0.5802,
+      "step": 7174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8723810423508423e-05,
+      "loss": 0.6179,
+      "step": 7175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8723461950254477e-05,
+      "loss": 0.602,
+      "step": 7176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872311343267407e-05,
+      "loss": 0.6293,
+      "step": 7177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872276487076897e-05,
+      "loss": 0.6653,
+      "step": 7178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8722416264540945e-05,
+      "loss": 0.5684,
+      "step": 7179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872206761399177e-05,
+      "loss": 0.6963,
+      "step": 7180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872171891912321e-05,
+      "loss": 0.6306,
+      "step": 7181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872137017993705e-05,
+      "loss": 0.6426,
+      "step": 7182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8721021396435045e-05,
+      "loss": 0.5923,
+      "step": 7183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.872067256861898e-05,
+      "loss": 0.6106,
+      "step": 7184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8720323696490626e-05,
+      "loss": 0.5932,
+      "step": 7185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871997478005175e-05,
+      "loss": 0.602,
+      "step": 7186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8719625819304132e-05,
+      "loss": 0.6818,
+      "step": 7187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871927681424954e-05,
+      "loss": 0.6369,
+      "step": 7188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871892776488975e-05,
+      "loss": 0.6175,
+      "step": 7189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8718578671226533e-05,
+      "loss": 0.5674,
+      "step": 7190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8718229533261666e-05,
+      "loss": 0.6537,
+      "step": 7191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871788035099692e-05,
+      "loss": 0.5857,
+      "step": 7192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8717531124434066e-05,
+      "loss": 0.6337,
+      "step": 7193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871718185357489e-05,
+      "loss": 0.5915,
+      "step": 7194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871683253842116e-05,
+      "loss": 0.6287,
+      "step": 7195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871648317897465e-05,
+      "loss": 0.6421,
+      "step": 7196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8716133775237132e-05,
+      "loss": 0.6662,
+      "step": 7197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871578432721039e-05,
+      "loss": 0.6316,
+      "step": 7198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8715434834896195e-05,
+      "loss": 0.6566,
+      "step": 7199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8715085298296322e-05,
+      "loss": 0.601,
+      "step": 7200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871473571741255e-05,
+      "loss": 0.6249,
+      "step": 7201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871438609224665e-05,
+      "loss": 0.6119,
+      "step": 7202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8714036422800408e-05,
+      "loss": 0.6389,
+      "step": 7203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871368670907559e-05,
+      "loss": 0.6594,
+      "step": 7204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8713336951073978e-05,
+      "loss": 0.6167,
+      "step": 7205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8712987148797353e-05,
+      "loss": 0.6362,
+      "step": 7206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8712637302247487e-05,
+      "loss": 0.616,
+      "step": 7207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8712287411426157e-05,
+      "loss": 0.6066,
+      "step": 7208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8711937476335145e-05,
+      "loss": 0.6085,
+      "step": 7209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8711587496976226e-05,
+      "loss": 0.6269,
+      "step": 7210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8711237473351182e-05,
+      "loss": 0.6039,
+      "step": 7211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8710887405461785e-05,
+      "loss": 0.61,
+      "step": 7212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.871053729330982e-05,
+      "loss": 0.6056,
+      "step": 7213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8710187136897065e-05,
+      "loss": 0.6589,
+      "step": 7214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.87098369362253e-05,
+      "loss": 0.6248,
+      "step": 7215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.87094866912963e-05,
+      "loss": 0.5975,
+      "step": 7216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8709136402111848e-05,
+      "loss": 0.6944,
+      "step": 7217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8708786068673723e-05,
+      "loss": 0.6395,
+      "step": 7218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8708435690983706e-05,
+      "loss": 0.6827,
+      "step": 7219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870808526904358e-05,
+      "loss": 0.5725,
+      "step": 7220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8707734802855117e-05,
+      "loss": 0.5604,
+      "step": 7221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8707384292420106e-05,
+      "loss": 0.5662,
+      "step": 7222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8707033737740326e-05,
+      "loss": 0.6811,
+      "step": 7223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8706683138817555e-05,
+      "loss": 0.6492,
+      "step": 7224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870633249565358e-05,
+      "loss": 0.5991,
+      "step": 7225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8705981808250176e-05,
+      "loss": 0.5969,
+      "step": 7226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8705631076609133e-05,
+      "loss": 0.6172,
+      "step": 7227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8705280300732225e-05,
+      "loss": 0.6135,
+      "step": 7228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870492948062124e-05,
+      "loss": 0.694,
+      "step": 7229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870457861627796e-05,
+      "loss": 0.572,
+      "step": 7230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870422770770417e-05,
+      "loss": 0.5848,
+      "step": 7231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8703876754901644e-05,
+      "loss": 0.6785,
+      "step": 7232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8703525757872172e-05,
+      "loss": 0.5945,
+      "step": 7233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8703174716617537e-05,
+      "loss": 0.6212,
+      "step": 7234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8702823631139526e-05,
+      "loss": 0.6353,
+      "step": 7235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8702472501439916e-05,
+      "loss": 0.6255,
+      "step": 7236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8702121327520494e-05,
+      "loss": 0.6213,
+      "step": 7237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870177010938305e-05,
+      "loss": 0.6154,
+      "step": 7238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870141884702936e-05,
+      "loss": 0.6078,
+      "step": 7239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870106754046121e-05,
+      "loss": 0.637,
+      "step": 7240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.870071618968039e-05,
+      "loss": 0.5804,
+      "step": 7241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8700364794688682e-05,
+      "loss": 0.6218,
+      "step": 7242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8700013355487872e-05,
+      "loss": 0.6322,
+      "step": 7243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8699661872079752e-05,
+      "loss": 0.6448,
+      "step": 7244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8699310344466094e-05,
+      "loss": 0.5753,
+      "step": 7245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.86989587726487e-05,
+      "loss": 0.6236,
+      "step": 7246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8698607156629345e-05,
+      "loss": 0.6062,
+      "step": 7247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869825549640982e-05,
+      "loss": 0.6326,
+      "step": 7248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869790379199191e-05,
+      "loss": 0.5768,
+      "step": 7249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.86975520433774e-05,
+      "loss": 0.5597,
+      "step": 7250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869720025056809e-05,
+      "loss": 0.7032,
+      "step": 7251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8696848413565752e-05,
+      "loss": 0.639,
+      "step": 7252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8696496532372186e-05,
+      "loss": 0.6233,
+      "step": 7253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869614460698917e-05,
+      "loss": 0.5847,
+      "step": 7254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8695792637418496e-05,
+      "loss": 0.6906,
+      "step": 7255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8695440623661953e-05,
+      "loss": 0.5726,
+      "step": 7256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8695088565721332e-05,
+      "loss": 0.654,
+      "step": 7257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8694736463598416e-05,
+      "loss": 0.6057,
+      "step": 7258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8694384317295e-05,
+      "loss": 0.639,
+      "step": 7259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8694032126812872e-05,
+      "loss": 0.5745,
+      "step": 7260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869367989215382e-05,
+      "loss": 0.6724,
+      "step": 7261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8693327613319633e-05,
+      "loss": 0.6224,
+      "step": 7262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.86929752903121e-05,
+      "loss": 0.545,
+      "step": 7263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8692622923133018e-05,
+      "loss": 0.6127,
+      "step": 7264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8692270511784173e-05,
+      "loss": 0.5933,
+      "step": 7265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8691918056267354e-05,
+      "loss": 0.6424,
+      "step": 7266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8691565556584354e-05,
+      "loss": 0.6143,
+      "step": 7267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8691213012736967e-05,
+      "loss": 0.6057,
+      "step": 7268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8690860424726977e-05,
+      "loss": 0.6342,
+      "step": 7269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.869050779255618e-05,
+      "loss": 0.5961,
+      "step": 7270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8690155116226374e-05,
+      "loss": 0.65,
+      "step": 7271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8689802395739337e-05,
+      "loss": 0.571,
+      "step": 7272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.8689449631096875e-05,
+      "loss": 0.672,
+      "step": 7273
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.868909682230077e-05,
+      "loss": 0.6687,
+      "step": 7274
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8688743969352823e-05,
+      "loss": 0.598,
+      "step": 7275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8688391072254822e-05,
+      "loss": 0.5461,
+      "step": 7276
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.868803813100856e-05,
+      "loss": 0.5975,
+      "step": 7277
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8687685145615836e-05,
+      "loss": 0.6006,
+      "step": 7278
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8687332116078436e-05,
+      "loss": 0.5888,
+      "step": 7279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8686979042398158e-05,
+      "loss": 0.5723,
+      "step": 7280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8686625924576794e-05,
+      "loss": 0.6269,
+      "step": 7281
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8686272762616144e-05,
+      "loss": 0.6628,
+      "step": 7282
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8685919556517992e-05,
+      "loss": 0.5889,
+      "step": 7283
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8685566306284142e-05,
+      "loss": 0.6206,
+      "step": 7284
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8685213011916386e-05,
+      "loss": 0.6688,
+      "step": 7285
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.868485967341652e-05,
+      "loss": 0.6392,
+      "step": 7286
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8684506290786336e-05,
+      "loss": 0.6655,
+      "step": 7287
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8684152864027635e-05,
+      "loss": 0.5678,
+      "step": 7288
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8683799393142207e-05,
+      "loss": 0.577,
+      "step": 7289
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8683445878131855e-05,
+      "loss": 0.592,
+      "step": 7290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8683092318998367e-05,
+      "loss": 0.6381,
+      "step": 7291
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.868273871574355e-05,
+      "loss": 0.7002,
+      "step": 7292
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8682385068369187e-05,
+      "loss": 0.6369,
+      "step": 7293
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8682031376877088e-05,
+      "loss": 0.6399,
+      "step": 7294
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8681677641269042e-05,
+      "loss": 0.5871,
+      "step": 7295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8681323861546853e-05,
+      "loss": 0.6248,
+      "step": 7296
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8680970037712313e-05,
+      "loss": 0.5816,
+      "step": 7297
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.868061616976722e-05,
+      "loss": 0.6136,
+      "step": 7298
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8680262257713377e-05,
+      "loss": 0.6082,
+      "step": 7299
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8679908301552578e-05,
+      "loss": 0.6021,
+      "step": 7300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8679554301286625e-05,
+      "loss": 0.588,
+      "step": 7301
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8679200256917313e-05,
+      "loss": 0.5892,
+      "step": 7302
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867884616844644e-05,
+      "loss": 0.6225,
+      "step": 7303
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8678492035875812e-05,
+      "loss": 0.6173,
+      "step": 7304
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8678137859207224e-05,
+      "loss": 0.6344,
+      "step": 7305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8677783638442474e-05,
+      "loss": 0.6635,
+      "step": 7306
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8677429373583364e-05,
+      "loss": 0.6864,
+      "step": 7307
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8677075064631696e-05,
+      "loss": 0.6278,
+      "step": 7308
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8676720711589267e-05,
+      "loss": 0.6231,
+      "step": 7309
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8676366314457878e-05,
+      "loss": 0.6333,
+      "step": 7310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8676011873239332e-05,
+      "loss": 0.6125,
+      "step": 7311
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867565738793543e-05,
+      "loss": 0.5605,
+      "step": 7312
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867530285854797e-05,
+      "loss": 0.6383,
+      "step": 7313
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8674948285078755e-05,
+      "loss": 0.6028,
+      "step": 7314
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867459366752959e-05,
+      "loss": 0.6103,
+      "step": 7315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867423900590227e-05,
+      "loss": 0.5946,
+      "step": 7316
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8673884300198607e-05,
+      "loss": 0.6219,
+      "step": 7317
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8673529550420392e-05,
+      "loss": 0.6481,
+      "step": 7318
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8673174756569437e-05,
+      "loss": 0.6079,
+      "step": 7319
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8672819918647537e-05,
+      "loss": 0.614,
+      "step": 7320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8672465036656505e-05,
+      "loss": 0.644,
+      "step": 7321
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8672110110598133e-05,
+      "loss": 0.625,
+      "step": 7322
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867175514047423e-05,
+      "loss": 0.5586,
+      "step": 7323
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8671400126286604e-05,
+      "loss": 0.6802,
+      "step": 7324
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8671045068037052e-05,
+      "loss": 0.6095,
+      "step": 7325
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867068996572738e-05,
+      "loss": 0.6249,
+      "step": 7326
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.867033481935939e-05,
+      "loss": 0.6048,
+      "step": 7327
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8669979628934894e-05,
+      "loss": 0.6305,
+      "step": 7328
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8669624394455692e-05,
+      "loss": 0.5741,
+      "step": 7329
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866926911592359e-05,
+      "loss": 0.692,
+      "step": 7330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8668913793340388e-05,
+      "loss": 0.5944,
+      "step": 7331
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8668558426707902e-05,
+      "loss": 0.609,
+      "step": 7332
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8668203016027927e-05,
+      "loss": 0.561,
+      "step": 7333
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866784756130228e-05,
+      "loss": 0.6285,
+      "step": 7334
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8667492062532755e-05,
+      "loss": 0.6132,
+      "step": 7335
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8667136519721166e-05,
+      "loss": 0.6158,
+      "step": 7336
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866678093286932e-05,
+      "loss": 0.6009,
+      "step": 7337
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866642530197902e-05,
+      "loss": 0.6414,
+      "step": 7338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8666069627052076e-05,
+      "loss": 0.6429,
+      "step": 7339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8665713908090293e-05,
+      "loss": 0.5913,
+      "step": 7340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866535814509548e-05,
+      "loss": 0.5871,
+      "step": 7341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8665002338069444e-05,
+      "loss": 0.6672,
+      "step": 7342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8664646487013997e-05,
+      "loss": 0.596,
+      "step": 7343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866429059193094e-05,
+      "loss": 0.6475,
+      "step": 7344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8663934652822088e-05,
+      "loss": 0.6224,
+      "step": 7345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8663578669689244e-05,
+      "loss": 0.5963,
+      "step": 7346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866322264253422e-05,
+      "loss": 0.5888,
+      "step": 7347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8662866571358825e-05,
+      "loss": 0.6473,
+      "step": 7348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8662510456164868e-05,
+      "loss": 0.6103,
+      "step": 7349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866215429695416e-05,
+      "loss": 0.6299,
+      "step": 7350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8661798093728505e-05,
+      "loss": 0.6142,
+      "step": 7351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866144184648972e-05,
+      "loss": 0.6634,
+      "step": 7352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8661085555239613e-05,
+      "loss": 0.5726,
+      "step": 7353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8660729219979994e-05,
+      "loss": 0.6028,
+      "step": 7354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866037284071267e-05,
+      "loss": 0.6319,
+      "step": 7355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.866001641743946e-05,
+      "loss": 0.6608,
+      "step": 7356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8659659950162168e-05,
+      "loss": 0.5988,
+      "step": 7357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8659303438882608e-05,
+      "loss": 0.6262,
+      "step": 7358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865894688360259e-05,
+      "loss": 0.5609,
+      "step": 7359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8658590284323926e-05,
+      "loss": 0.5805,
+      "step": 7360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8658233641048433e-05,
+      "loss": 0.579,
+      "step": 7361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8657876953777917e-05,
+      "loss": 0.596,
+      "step": 7362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865752022251419e-05,
+      "loss": 0.5839,
+      "step": 7363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8657163447259072e-05,
+      "loss": 0.6409,
+      "step": 7364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8656806628014368e-05,
+      "loss": 0.6058,
+      "step": 7365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8656449764781898e-05,
+      "loss": 0.5763,
+      "step": 7366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8656092857563468e-05,
+      "loss": 0.6682,
+      "step": 7367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8655735906360897e-05,
+      "loss": 0.613,
+      "step": 7368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8655378911175995e-05,
+      "loss": 0.7161,
+      "step": 7369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8655021872010578e-05,
+      "loss": 0.622,
+      "step": 7370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865466478886646e-05,
+      "loss": 0.5376,
+      "step": 7371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865430766174546e-05,
+      "loss": 0.6164,
+      "step": 7372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8653950490649386e-05,
+      "loss": 0.6184,
+      "step": 7373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865359327558005e-05,
+      "loss": 0.6418,
+      "step": 7374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865323601653928e-05,
+      "loss": 0.6232,
+      "step": 7375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8652878713528878e-05,
+      "loss": 0.5751,
+      "step": 7376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8652521366550667e-05,
+      "loss": 0.6265,
+      "step": 7377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865216397560646e-05,
+      "loss": 0.6009,
+      "step": 7378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8651806540698074e-05,
+      "loss": 0.5748,
+      "step": 7379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865144906182733e-05,
+      "loss": 0.5886,
+      "step": 7380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8651091538996033e-05,
+      "loss": 0.6223,
+      "step": 7381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8650733972206007e-05,
+      "loss": 0.5912,
+      "step": 7382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8650376361459074e-05,
+      "loss": 0.6488,
+      "step": 7383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.865001870675704e-05,
+      "loss": 0.6302,
+      "step": 7384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8649661008101726e-05,
+      "loss": 0.5941,
+      "step": 7385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864930326549496e-05,
+      "loss": 0.5988,
+      "step": 7386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8648945478938544e-05,
+      "loss": 0.6193,
+      "step": 7387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8648587648434303e-05,
+      "loss": 0.6097,
+      "step": 7388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8648229773984058e-05,
+      "loss": 0.5748,
+      "step": 7389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8647871855589622e-05,
+      "loss": 0.589,
+      "step": 7390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864751389325282e-05,
+      "loss": 0.5911,
+      "step": 7391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8647155886975464e-05,
+      "loss": 0.6893,
+      "step": 7392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8646797836759377e-05,
+      "loss": 0.6517,
+      "step": 7393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864643974260638e-05,
+      "loss": 0.5559,
+      "step": 7394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864608160451829e-05,
+      "loss": 0.6436,
+      "step": 7395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8645723422496927e-05,
+      "loss": 0.6018,
+      "step": 7396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864536519654411e-05,
+      "loss": 0.6313,
+      "step": 7397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864500692666166e-05,
+      "loss": 0.5902,
+      "step": 7398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.86446486128514e-05,
+      "loss": 0.6389,
+      "step": 7399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8644290255115146e-05,
+      "loss": 0.6133,
+      "step": 7400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8643931853454725e-05,
+      "loss": 0.6716,
+      "step": 7401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864357340787195e-05,
+      "loss": 0.6333,
+      "step": 7402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8643214918368652e-05,
+      "loss": 0.6008,
+      "step": 7403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8642856384946646e-05,
+      "loss": 0.6045,
+      "step": 7404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8642497807607756e-05,
+      "loss": 0.608,
+      "step": 7405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8642139186353803e-05,
+      "loss": 0.5957,
+      "step": 7406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864178052118661e-05,
+      "loss": 0.6474,
+      "step": 7407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8641421812108e-05,
+      "loss": 0.5473,
+      "step": 7408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8641063059119795e-05,
+      "loss": 0.5552,
+      "step": 7409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8640704262223816e-05,
+      "loss": 0.5649,
+      "step": 7410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.864034542142189e-05,
+      "loss": 0.6627,
+      "step": 7411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8639986536715837e-05,
+      "loss": 0.5926,
+      "step": 7412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8639627608107485e-05,
+      "loss": 0.5957,
+      "step": 7413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8639268635598652e-05,
+      "loss": 0.6166,
+      "step": 7414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8638909619191164e-05,
+      "loss": 0.5793,
+      "step": 7415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.863855055888685e-05,
+      "loss": 0.6743,
+      "step": 7416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.863819145468753e-05,
+      "loss": 0.6375,
+      "step": 7417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8637832306595028e-05,
+      "loss": 0.6546,
+      "step": 7418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.863747311461117e-05,
+      "loss": 0.6051,
+      "step": 7419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.863711387873778e-05,
+      "loss": 0.6028,
+      "step": 7420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8636754598976687e-05,
+      "loss": 0.6099,
+      "step": 7421
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8636395275329716e-05,
+      "loss": 0.5641,
+      "step": 7422
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863603590779869e-05,
+      "loss": 0.5863,
+      "step": 7423
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8635676496385433e-05,
+      "loss": 0.6175,
+      "step": 7424
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8635317041091778e-05,
+      "loss": 0.6594,
+      "step": 7425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863495754191955e-05,
+      "loss": 0.5751,
+      "step": 7426
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863459799887057e-05,
+      "loss": 0.57,
+      "step": 7427
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863423841194667e-05,
+      "loss": 0.6555,
+      "step": 7428
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863387878114968e-05,
+      "loss": 0.5859,
+      "step": 7429
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8633519106481415e-05,
+      "loss": 0.6485,
+      "step": 7430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8633159387943717e-05,
+      "loss": 0.6261,
+      "step": 7431
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863279962553841e-05,
+      "loss": 0.6321,
+      "step": 7432
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8632439819267314e-05,
+      "loss": 0.5972,
+      "step": 7433
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8632079969132265e-05,
+      "loss": 0.5822,
+      "step": 7434
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863172007513509e-05,
+      "loss": 0.5937,
+      "step": 7435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863136013727762e-05,
+      "loss": 0.5635,
+      "step": 7436
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.863100015556168e-05,
+      "loss": 0.5981,
+      "step": 7437
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.86306401299891e-05,
+      "loss": 0.6068,
+      "step": 7438
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8630280060561706e-05,
+      "loss": 0.6046,
+      "step": 7439
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8629919947281334e-05,
+      "loss": 0.6695,
+      "step": 7440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8629559790149813e-05,
+      "loss": 0.625,
+      "step": 7441
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862919958916897e-05,
+      "loss": 0.645,
+      "step": 7442
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862883934434064e-05,
+      "loss": 0.5707,
+      "step": 7443
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8628479055666645e-05,
+      "loss": 0.5636,
+      "step": 7444
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8628118723148824e-05,
+      "loss": 0.6149,
+      "step": 7445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8627758346789004e-05,
+      "loss": 0.6068,
+      "step": 7446
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8627397926589018e-05,
+      "loss": 0.6481,
+      "step": 7447
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8627037462550697e-05,
+      "loss": 0.5824,
+      "step": 7448
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8626676954675873e-05,
+      "loss": 0.5856,
+      "step": 7449
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8626316402966372e-05,
+      "loss": 0.5612,
+      "step": 7450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8625955807424034e-05,
+      "loss": 0.604,
+      "step": 7451
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862559516805069e-05,
+      "loss": 0.5802,
+      "step": 7452
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862523448484817e-05,
+      "loss": 0.6044,
+      "step": 7453
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8624873757818306e-05,
+      "loss": 0.5574,
+      "step": 7454
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8624512986962933e-05,
+      "loss": 0.6417,
+      "step": 7455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8624152172283885e-05,
+      "loss": 0.6163,
+      "step": 7456
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8623791313782994e-05,
+      "loss": 0.5971,
+      "step": 7457
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862343041146209e-05,
+      "loss": 0.637,
+      "step": 7458
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8623069465323014e-05,
+      "loss": 0.6471,
+      "step": 7459
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8622708475367598e-05,
+      "loss": 0.6238,
+      "step": 7460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862234744159767e-05,
+      "loss": 0.5804,
+      "step": 7461
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8621986364015073e-05,
+      "loss": 0.6151,
+      "step": 7462
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.862162524262164e-05,
+      "loss": 0.5955,
+      "step": 7463
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.86212640774192e-05,
+      "loss": 0.5983,
+      "step": 7464
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8620902868409598e-05,
+      "loss": 0.6152,
+      "step": 7465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8620541615594657e-05,
+      "loss": 0.5906,
+      "step": 7466
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8620180318976223e-05,
+      "loss": 0.6059,
+      "step": 7467
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8619818978556127e-05,
+      "loss": 0.6854,
+      "step": 7468
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861945759433621e-05,
+      "loss": 0.6302,
+      "step": 7469
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.86190961663183e-05,
+      "loss": 0.6539,
+      "step": 7470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8618734694504243e-05,
+      "loss": 0.6314,
+      "step": 7471
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8618373178895866e-05,
+      "loss": 0.6387,
+      "step": 7472
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8618011619495012e-05,
+      "loss": 0.5749,
+      "step": 7473
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8617650016303517e-05,
+      "loss": 0.5787,
+      "step": 7474
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861728836932322e-05,
+      "loss": 0.6312,
+      "step": 7475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8616926678555955e-05,
+      "loss": 0.5864,
+      "step": 7476
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8616564944003565e-05,
+      "loss": 0.6084,
+      "step": 7477
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8616203165667882e-05,
+      "loss": 0.6458,
+      "step": 7478
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861584134355075e-05,
+      "loss": 0.6032,
+      "step": 7479
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8615479477654002e-05,
+      "loss": 0.5559,
+      "step": 7480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861511756797948e-05,
+      "loss": 0.6213,
+      "step": 7481
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8614755614529024e-05,
+      "loss": 0.5901,
+      "step": 7482
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8614393617304473e-05,
+      "loss": 0.6033,
+      "step": 7483
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8614031576307665e-05,
+      "loss": 0.6228,
+      "step": 7484
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861366949154044e-05,
+      "loss": 0.6033,
+      "step": 7485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8613307363004636e-05,
+      "loss": 0.6121,
+      "step": 7486
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8612945190702095e-05,
+      "loss": 0.6156,
+      "step": 7487
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8612582974634655e-05,
+      "loss": 0.61,
+      "step": 7488
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861222071480416e-05,
+      "loss": 0.6167,
+      "step": 7489
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861185841121245e-05,
+      "loss": 0.592,
+      "step": 7490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8611496063861367e-05,
+      "loss": 0.5685,
+      "step": 7491
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.861113367275275e-05,
+      "loss": 0.6334,
+      "step": 7492
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8610771237888436e-05,
+      "loss": 0.6653,
+      "step": 7493
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8610408759270277e-05,
+      "loss": 0.5832,
+      "step": 7494
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8610046236900108e-05,
+      "loss": 0.6296,
+      "step": 7495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8609683670779773e-05,
+      "loss": 0.6242,
+      "step": 7496
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860932106091111e-05,
+      "loss": 0.5579,
+      "step": 7497
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8608958407295968e-05,
+      "loss": 0.6268,
+      "step": 7498
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860859570993619e-05,
+      "loss": 0.6199,
+      "step": 7499
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8608232968833615e-05,
+      "loss": 0.5968,
+      "step": 7500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8607870183990084e-05,
+      "loss": 0.5887,
+      "step": 7501
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8607507355407446e-05,
+      "loss": 0.6113,
+      "step": 7502
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860714448308754e-05,
+      "loss": 0.6513,
+      "step": 7503
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8606781567032218e-05,
+      "loss": 0.5944,
+      "step": 7504
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860641860724331e-05,
+      "loss": 0.6627,
+      "step": 7505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8606055603722674e-05,
+      "loss": 0.6653,
+      "step": 7506
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8605692556472147e-05,
+      "loss": 0.6638,
+      "step": 7507
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860532946549358e-05,
+      "loss": 0.5875,
+      "step": 7508
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8604966330788808e-05,
+      "loss": 0.5506,
+      "step": 7509
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8604603152359685e-05,
+      "loss": 0.6296,
+      "step": 7510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8604239930208053e-05,
+      "loss": 0.611,
+      "step": 7511
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8603876664335758e-05,
+      "loss": 0.6563,
+      "step": 7512
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8603513354744643e-05,
+      "loss": 0.594,
+      "step": 7513
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8603150001436563e-05,
+      "loss": 0.6125,
+      "step": 7514
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8602786604413352e-05,
+      "loss": 0.5712,
+      "step": 7515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860242316367687e-05,
+      "loss": 0.6543,
+      "step": 7516
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860205967922895e-05,
+      "loss": 0.6382,
+      "step": 7517
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8601696151071447e-05,
+      "loss": 0.6332,
+      "step": 7518
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8601332579206207e-05,
+      "loss": 0.5795,
+      "step": 7519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860096896363508e-05,
+      "loss": 0.6284,
+      "step": 7520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8600605304359907e-05,
+      "loss": 0.6018,
+      "step": 7521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.860024160138254e-05,
+      "loss": 0.5519,
+      "step": 7522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8599877854704828e-05,
+      "loss": 0.583,
+      "step": 7523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8599514064328617e-05,
+      "loss": 0.6325,
+      "step": 7524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8599150230255762e-05,
+      "loss": 0.6461,
+      "step": 7525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85987863524881e-05,
+      "loss": 0.5448,
+      "step": 7526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859842243102749e-05,
+      "loss": 0.6042,
+      "step": 7527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8598058465875774e-05,
+      "loss": 0.6113,
+      "step": 7528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8597694457034807e-05,
+      "loss": 0.6153,
+      "step": 7529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8597330404506436e-05,
+      "loss": 0.555,
+      "step": 7530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8596966308292514e-05,
+      "loss": 0.6109,
+      "step": 7531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8596602168394883e-05,
+      "loss": 0.5603,
+      "step": 7532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85962379848154e-05,
+      "loss": 0.6337,
+      "step": 7533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8595873757555916e-05,
+      "loss": 0.5896,
+      "step": 7534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859550948661828e-05,
+      "loss": 0.6023,
+      "step": 7535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8595145172004345e-05,
+      "loss": 0.6302,
+      "step": 7536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8594780813715955e-05,
+      "loss": 0.5775,
+      "step": 7537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859441641175497e-05,
+      "loss": 0.6466,
+      "step": 7538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859405196612324e-05,
+      "loss": 0.6215,
+      "step": 7539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859368747682261e-05,
+      "loss": 0.6305,
+      "step": 7540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8593322943854943e-05,
+      "loss": 0.5793,
+      "step": 7541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859295836722208e-05,
+      "loss": 0.6762,
+      "step": 7542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8592593746925884e-05,
+      "loss": 0.5922,
+      "step": 7543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85922290829682e-05,
+      "loss": 0.5774,
+      "step": 7544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8591864375350882e-05,
+      "loss": 0.5741,
+      "step": 7545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8591499624075787e-05,
+      "loss": 0.6102,
+      "step": 7546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.859113482914477e-05,
+      "loss": 0.5896,
+      "step": 7547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8590769990559675e-05,
+      "loss": 0.6473,
+      "step": 7548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8590405108322362e-05,
+      "loss": 0.5981,
+      "step": 7549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8590040182434687e-05,
+      "loss": 0.6317,
+      "step": 7550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85896752128985e-05,
+      "loss": 0.6335,
+      "step": 7551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.858931019971566e-05,
+      "loss": 0.6008,
+      "step": 7552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8588945142888018e-05,
+      "loss": 0.5905,
+      "step": 7553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8588580042417434e-05,
+      "loss": 0.6525,
+      "step": 7554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8588214898305757e-05,
+      "loss": 0.6412,
+      "step": 7555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.858784971055484e-05,
+      "loss": 0.6617,
+      "step": 7556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.858748447916655e-05,
+      "loss": 0.5642,
+      "step": 7557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8587119204142735e-05,
+      "loss": 0.6011,
+      "step": 7558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8586753885485252e-05,
+      "loss": 0.6296,
+      "step": 7559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8586388523195957e-05,
+      "loss": 0.6207,
+      "step": 7560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8586023117276707e-05,
+      "loss": 0.6085,
+      "step": 7561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.858565766772936e-05,
+      "loss": 0.6281,
+      "step": 7562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8585292174555772e-05,
+      "loss": 0.6564,
+      "step": 7563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85849266377578e-05,
+      "loss": 0.5923,
+      "step": 7564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.85845610573373e-05,
+      "loss": 0.6013,
+      "step": 7565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8584195433296133e-05,
+      "loss": 0.5422,
+      "step": 7566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8583829765636157e-05,
+      "loss": 0.5835,
+      "step": 7567
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8583464054359224e-05,
+      "loss": 0.6374,
+      "step": 7568
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.85830982994672e-05,
+      "loss": 0.6866,
+      "step": 7569
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8582732500961938e-05,
+      "loss": 0.6469,
+      "step": 7570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8582366658845297e-05,
+      "loss": 0.5855,
+      "step": 7571
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.858200077311914e-05,
+      "loss": 0.6419,
+      "step": 7572
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8581634843785323e-05,
+      "loss": 0.685,
+      "step": 7573
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8581268870845707e-05,
+      "loss": 0.6018,
+      "step": 7574
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.858090285430215e-05,
+      "loss": 0.5837,
+      "step": 7575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.858053679415651e-05,
+      "loss": 0.5824,
+      "step": 7576
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8580170690410653e-05,
+      "loss": 0.5704,
+      "step": 7577
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8579804543066435e-05,
+      "loss": 0.6269,
+      "step": 7578
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857943835212572e-05,
+      "loss": 0.6284,
+      "step": 7579
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857907211759036e-05,
+      "loss": 0.6433,
+      "step": 7580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8578705839462227e-05,
+      "loss": 0.591,
+      "step": 7581
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8578339517743175e-05,
+      "loss": 0.6169,
+      "step": 7582
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8577973152435068e-05,
+      "loss": 0.5656,
+      "step": 7583
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8577606743539768e-05,
+      "loss": 0.6227,
+      "step": 7584
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8577240291059137e-05,
+      "loss": 0.6245,
+      "step": 7585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8576873794995032e-05,
+      "loss": 0.632,
+      "step": 7586
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8576507255349324e-05,
+      "loss": 0.5941,
+      "step": 7587
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857614067212387e-05,
+      "loss": 0.6229,
+      "step": 7588
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857577404532053e-05,
+      "loss": 0.5865,
+      "step": 7589
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857540737494117e-05,
+      "loss": 0.5645,
+      "step": 7590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8575040660987656e-05,
+      "loss": 0.6635,
+      "step": 7591
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.857467390346185e-05,
+      "loss": 0.6115,
+      "step": 7592
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8574307102365617e-05,
+      "loss": 0.5804,
+      "step": 7593
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8573940257700814e-05,
+      "loss": 0.6084,
+      "step": 7594
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8573573369469314e-05,
+      "loss": 0.5695,
+      "step": 7595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8573206437672976e-05,
+      "loss": 0.592,
+      "step": 7596
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8572839462313662e-05,
+      "loss": 0.6333,
+      "step": 7597
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8572472443393242e-05,
+      "loss": 0.6107,
+      "step": 7598
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8572105380913578e-05,
+      "loss": 0.6035,
+      "step": 7599
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8571738274876538e-05,
+      "loss": 0.6373,
+      "step": 7600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8571371125283983e-05,
+      "loss": 0.6233,
+      "step": 7601
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8571003932137783e-05,
+      "loss": 0.5933,
+      "step": 7602
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.85706366954398e-05,
+      "loss": 0.5796,
+      "step": 7603
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8570269415191908e-05,
+      "loss": 0.5831,
+      "step": 7604
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8569902091395958e-05,
+      "loss": 0.5963,
+      "step": 7605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856953472405383e-05,
+      "loss": 0.701,
+      "step": 7606
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8569167313167388e-05,
+      "loss": 0.625,
+      "step": 7607
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8568799858738497e-05,
+      "loss": 0.6439,
+      "step": 7608
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8568432360769023e-05,
+      "loss": 0.6463,
+      "step": 7609
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8568064819260835e-05,
+      "loss": 0.6022,
+      "step": 7610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8567697234215798e-05,
+      "loss": 0.5801,
+      "step": 7611
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8567329605635788e-05,
+      "loss": 0.6086,
+      "step": 7612
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8566961933522662e-05,
+      "loss": 0.5536,
+      "step": 7613
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8566594217878293e-05,
+      "loss": 0.6381,
+      "step": 7614
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8566226458704553e-05,
+      "loss": 0.5857,
+      "step": 7615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8565858656003307e-05,
+      "loss": 0.6329,
+      "step": 7616
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8565490809776423e-05,
+      "loss": 0.6513,
+      "step": 7617
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8565122920025774e-05,
+      "loss": 0.6433,
+      "step": 7618
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8564754986753227e-05,
+      "loss": 0.6026,
+      "step": 7619
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856438700996065e-05,
+      "loss": 0.6489,
+      "step": 7620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8564018989649914e-05,
+      "loss": 0.6384,
+      "step": 7621
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8563650925822888e-05,
+      "loss": 0.5789,
+      "step": 7622
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856328281848145e-05,
+      "loss": 0.6278,
+      "step": 7623
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8562914667627462e-05,
+      "loss": 0.5874,
+      "step": 7624
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8562546473262793e-05,
+      "loss": 0.5841,
+      "step": 7625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8562178235389317e-05,
+      "loss": 0.6589,
+      "step": 7626
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856180995400891e-05,
+      "loss": 0.598,
+      "step": 7627
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856144162912344e-05,
+      "loss": 0.6622,
+      "step": 7628
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8561073260734773e-05,
+      "loss": 0.5802,
+      "step": 7629
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.856070484884479e-05,
+      "loss": 0.6197,
+      "step": 7630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8560336393455353e-05,
+      "loss": 0.5681,
+      "step": 7631
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8559967894568346e-05,
+      "loss": 0.6127,
+      "step": 7632
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8559599352185632e-05,
+      "loss": 0.672,
+      "step": 7633
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8559230766309087e-05,
+      "loss": 0.5568,
+      "step": 7634
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8558862136940585e-05,
+      "loss": 0.6335,
+      "step": 7635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8558493464081996e-05,
+      "loss": 0.6535,
+      "step": 7636
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8558124747735197e-05,
+      "loss": 0.6082,
+      "step": 7637
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855775598790206e-05,
+      "loss": 0.6275,
+      "step": 7638
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855738718458446e-05,
+      "loss": 0.6206,
+      "step": 7639
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8557018337784262e-05,
+      "loss": 0.6038,
+      "step": 7640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8556649447503354e-05,
+      "loss": 0.6655,
+      "step": 7641
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.85562805137436e-05,
+      "loss": 0.577,
+      "step": 7642
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8555911536506887e-05,
+      "loss": 0.5887,
+      "step": 7643
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855554251579508e-05,
+      "loss": 0.6008,
+      "step": 7644
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855517345161005e-05,
+      "loss": 0.5984,
+      "step": 7645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855480434395368e-05,
+      "loss": 0.6587,
+      "step": 7646
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8554435192827845e-05,
+      "loss": 0.6068,
+      "step": 7647
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855406599823442e-05,
+      "loss": 0.5966,
+      "step": 7648
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855369676017528e-05,
+      "loss": 0.5734,
+      "step": 7649
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8553327478652302e-05,
+      "loss": 0.6077,
+      "step": 7650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855295815366736e-05,
+      "loss": 0.6423,
+      "step": 7651
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8552588785222334e-05,
+      "loss": 0.6565,
+      "step": 7652
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.85522193733191e-05,
+      "loss": 0.5967,
+      "step": 7653
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8551849917959534e-05,
+      "loss": 0.5999,
+      "step": 7654
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8551480419145518e-05,
+      "loss": 0.6313,
+      "step": 7655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8551110876878922e-05,
+      "loss": 0.5421,
+      "step": 7656
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8550741291161623e-05,
+      "loss": 0.6136,
+      "step": 7657
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.855037166199551e-05,
+      "loss": 0.6213,
+      "step": 7658
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8550001989382455e-05,
+      "loss": 0.619,
+      "step": 7659
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8549632273324332e-05,
+      "loss": 0.5711,
+      "step": 7660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8549262513823025e-05,
+      "loss": 0.692,
+      "step": 7661
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854889271088041e-05,
+      "loss": 0.6464,
+      "step": 7662
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854852286449837e-05,
+      "loss": 0.5263,
+      "step": 7663
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8548152974678778e-05,
+      "loss": 0.5818,
+      "step": 7664
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8547783041423522e-05,
+      "loss": 0.5632,
+      "step": 7665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8547413064734473e-05,
+      "loss": 0.6235,
+      "step": 7666
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8547043044613518e-05,
+      "loss": 0.6643,
+      "step": 7667
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854667298106253e-05,
+      "loss": 0.6321,
+      "step": 7668
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8546302874083403e-05,
+      "loss": 0.6609,
+      "step": 7669
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8545932723678e-05,
+      "loss": 0.6108,
+      "step": 7670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8545562529848213e-05,
+      "loss": 0.659,
+      "step": 7671
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854519229259592e-05,
+      "loss": 0.5749,
+      "step": 7672
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8544822011922998e-05,
+      "loss": 0.5722,
+      "step": 7673
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8544451687831337e-05,
+      "loss": 0.609,
+      "step": 7674
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8544081320322814e-05,
+      "loss": 0.6413,
+      "step": 7675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854371090939931e-05,
+      "loss": 0.6879,
+      "step": 7676
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.854334045506271e-05,
+      "loss": 0.6048,
+      "step": 7677
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8542969957314896e-05,
+      "loss": 0.649,
+      "step": 7678
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8542599416157748e-05,
+      "loss": 0.6647,
+      "step": 7679
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8542228831593153e-05,
+      "loss": 0.6205,
+      "step": 7680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8541858203622988e-05,
+      "loss": 0.6145,
+      "step": 7681
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8541487532249144e-05,
+      "loss": 0.5919,
+      "step": 7682
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8541116817473498e-05,
+      "loss": 0.6013,
+      "step": 7683
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8540746059297937e-05,
+      "loss": 0.6377,
+      "step": 7684
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8540375257724344e-05,
+      "loss": 0.6656,
+      "step": 7685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.85400044127546e-05,
+      "loss": 0.604,
+      "step": 7686
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8539633524390598e-05,
+      "loss": 0.6297,
+      "step": 7687
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8539262592634215e-05,
+      "loss": 0.5772,
+      "step": 7688
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8538891617487332e-05,
+      "loss": 0.6523,
+      "step": 7689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8538520598951847e-05,
+      "loss": 0.5978,
+      "step": 7690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8538149537029638e-05,
+      "loss": 0.5836,
+      "step": 7691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8537778431722587e-05,
+      "loss": 0.6514,
+      "step": 7692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8537407283032585e-05,
+      "loss": 0.6159,
+      "step": 7693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8537036090961517e-05,
+      "loss": 0.6329,
+      "step": 7694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.853666485551127e-05,
+      "loss": 0.6026,
+      "step": 7695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8536293576683725e-05,
+      "loss": 0.5984,
+      "step": 7696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8535922254480776e-05,
+      "loss": 0.5948,
+      "step": 7697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8535550888904302e-05,
+      "loss": 0.6059,
+      "step": 7698
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8535179479956197e-05,
+      "loss": 0.6489,
+      "step": 7699
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8534808027638346e-05,
+      "loss": 0.604,
+      "step": 7700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8534436531952637e-05,
+      "loss": 0.579,
+      "step": 7701
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8534064992900953e-05,
+      "loss": 0.5723,
+      "step": 7702
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8533693410485186e-05,
+      "loss": 0.6655,
+      "step": 7703
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8533321784707225e-05,
+      "loss": 0.6169,
+      "step": 7704
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.853295011556896e-05,
+      "loss": 0.6267,
+      "step": 7705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.853257840307227e-05,
+      "loss": 0.6162,
+      "step": 7706
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8532206647219052e-05,
+      "loss": 0.6258,
+      "step": 7707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8531834848011193e-05,
+      "loss": 0.6701,
+      "step": 7708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8531463005450585e-05,
+      "loss": 0.6787,
+      "step": 7709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8531091119539115e-05,
+      "loss": 0.6345,
+      "step": 7710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.853071919027867e-05,
+      "loss": 0.6027,
+      "step": 7711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8530347217671146e-05,
+      "loss": 0.6249,
+      "step": 7712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8529975201718425e-05,
+      "loss": 0.6154,
+      "step": 7713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8529603142422405e-05,
+      "loss": 0.6187,
+      "step": 7714
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8529231039784975e-05,
+      "loss": 0.5785,
+      "step": 7715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852885889380802e-05,
+      "loss": 0.61,
+      "step": 7716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852848670449344e-05,
+      "loss": 0.6191,
+      "step": 7717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8528114471843116e-05,
+      "loss": 0.5855,
+      "step": 7718
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8527742195858948e-05,
+      "loss": 0.6091,
+      "step": 7719
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8527369876542823e-05,
+      "loss": 0.6386,
+      "step": 7720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8526997513896633e-05,
+      "loss": 0.6217,
+      "step": 7721
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8526625107922275e-05,
+      "loss": 0.5909,
+      "step": 7722
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8526252658621637e-05,
+      "loss": 0.626,
+      "step": 7723
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852588016599661e-05,
+      "loss": 0.6108,
+      "step": 7724
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852550763004909e-05,
+      "loss": 0.6479,
+      "step": 7725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852513505078097e-05,
+      "loss": 0.6407,
+      "step": 7726
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852476242819414e-05,
+      "loss": 0.5592,
+      "step": 7727
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8524389762290494e-05,
+      "loss": 0.6887,
+      "step": 7728
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852401705307193e-05,
+      "loss": 0.6479,
+      "step": 7729
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852364430054034e-05,
+      "loss": 0.5799,
+      "step": 7730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852327150469762e-05,
+      "loss": 0.6399,
+      "step": 7731
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8522898665545655e-05,
+      "loss": 0.5979,
+      "step": 7732
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852252578308635e-05,
+      "loss": 0.578,
+      "step": 7733
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852215285732159e-05,
+      "loss": 0.5826,
+      "step": 7734
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8521779888253282e-05,
+      "loss": 0.5619,
+      "step": 7735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.852140687588331e-05,
+      "loss": 0.6189,
+      "step": 7736
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8521033820213575e-05,
+      "loss": 0.6137,
+      "step": 7737
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8520660721245974e-05,
+      "loss": 0.5659,
+      "step": 7738
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.85202875789824e-05,
+      "loss": 0.5533,
+      "step": 7739
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8519914393424748e-05,
+      "loss": 0.6763,
+      "step": 7740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8519541164574914e-05,
+      "loss": 0.5988,
+      "step": 7741
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.85191678924348e-05,
+      "loss": 0.6432,
+      "step": 7742
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8518794577006297e-05,
+      "loss": 0.5999,
+      "step": 7743
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8518421218291303e-05,
+      "loss": 0.6055,
+      "step": 7744
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851804781629172e-05,
+      "loss": 0.6335,
+      "step": 7745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8517674371009436e-05,
+      "loss": 0.5433,
+      "step": 7746
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851730088244636e-05,
+      "loss": 0.6459,
+      "step": 7747
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851692735060438e-05,
+      "loss": 0.6469,
+      "step": 7748
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.85165537754854e-05,
+      "loss": 0.6277,
+      "step": 7749
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8516180157091312e-05,
+      "loss": 0.5973,
+      "step": 7750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8515806495424024e-05,
+      "loss": 0.6339,
+      "step": 7751
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8515432790485427e-05,
+      "loss": 0.6005,
+      "step": 7752
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851505904227742e-05,
+      "loss": 0.6039,
+      "step": 7753
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8514685250801905e-05,
+      "loss": 0.6349,
+      "step": 7754
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851431141606078e-05,
+      "loss": 0.6639,
+      "step": 7755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851393753805595e-05,
+      "loss": 0.615,
+      "step": 7756
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8513563616789304e-05,
+      "loss": 0.6522,
+      "step": 7757
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851318965226275e-05,
+      "loss": 0.5695,
+      "step": 7758
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851281564447819e-05,
+      "loss": 0.6449,
+      "step": 7759
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8512441593437515e-05,
+      "loss": 0.5959,
+      "step": 7760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8512067499142633e-05,
+      "loss": 0.6227,
+      "step": 7761
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8511693361595443e-05,
+      "loss": 0.64,
+      "step": 7762
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.851131918079785e-05,
+      "loss": 0.6077,
+      "step": 7763
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8510944956751747e-05,
+      "loss": 0.6245,
+      "step": 7764
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8510570689459042e-05,
+      "loss": 0.6094,
+      "step": 7765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8510196378921636e-05,
+      "loss": 0.5775,
+      "step": 7766
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8509822025141426e-05,
+      "loss": 0.6556,
+      "step": 7767
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850944762812032e-05,
+      "loss": 0.6357,
+      "step": 7768
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850907318786022e-05,
+      "loss": 0.6097,
+      "step": 7769
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8508698704363026e-05,
+      "loss": 0.6003,
+      "step": 7770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850832417763064e-05,
+      "loss": 0.6178,
+      "step": 7771
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8507949607664974e-05,
+      "loss": 0.5678,
+      "step": 7772
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850757499446792e-05,
+      "loss": 0.5918,
+      "step": 7773
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850720033804138e-05,
+      "loss": 0.617,
+      "step": 7774
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850682563838727e-05,
+      "loss": 0.598,
+      "step": 7775
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8506450895507487e-05,
+      "loss": 0.5755,
+      "step": 7776
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8506076109403938e-05,
+      "loss": 0.6464,
+      "step": 7777
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8505701280078523e-05,
+      "loss": 0.5411,
+      "step": 7778
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850532640753315e-05,
+      "loss": 0.6484,
+      "step": 7779
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8504951491769717e-05,
+      "loss": 0.6437,
+      "step": 7780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850457653279014e-05,
+      "loss": 0.5158,
+      "step": 7781
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8504201530596316e-05,
+      "loss": 0.5704,
+      "step": 7782
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8503826485190157e-05,
+      "loss": 0.5935,
+      "step": 7783
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8503451396573564e-05,
+      "loss": 0.6263,
+      "step": 7784
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8503076264748442e-05,
+      "loss": 0.6649,
+      "step": 7785
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.85027010897167e-05,
+      "loss": 0.6164,
+      "step": 7786
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8502325871480245e-05,
+      "loss": 0.6263,
+      "step": 7787
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8501950610040982e-05,
+      "loss": 0.6747,
+      "step": 7788
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8501575305400817e-05,
+      "loss": 0.6072,
+      "step": 7789
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850119995756166e-05,
+      "loss": 0.5886,
+      "step": 7790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8500824566525413e-05,
+      "loss": 0.6125,
+      "step": 7791
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.850044913229399e-05,
+      "loss": 0.6307,
+      "step": 7792
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8500073654869295e-05,
+      "loss": 0.6672,
+      "step": 7793
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8499698134253234e-05,
+      "loss": 0.6696,
+      "step": 7794
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849932257044772e-05,
+      "loss": 0.5916,
+      "step": 7795
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8498946963454657e-05,
+      "loss": 0.6432,
+      "step": 7796
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8498571313275956e-05,
+      "loss": 0.6501,
+      "step": 7797
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849819561991353e-05,
+      "loss": 0.5474,
+      "step": 7798
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8497819883369278e-05,
+      "loss": 0.6312,
+      "step": 7799
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8497444103645114e-05,
+      "loss": 0.6075,
+      "step": 7800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849706828074295e-05,
+      "loss": 0.7216,
+      "step": 7801
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849669241466469e-05,
+      "loss": 0.616,
+      "step": 7802
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8496316505412252e-05,
+      "loss": 0.6157,
+      "step": 7803
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849594055298754e-05,
+      "loss": 0.6165,
+      "step": 7804
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8495564557392468e-05,
+      "loss": 0.5945,
+      "step": 7805
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849518851862894e-05,
+      "loss": 0.6618,
+      "step": 7806
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8494812436698875e-05,
+      "loss": 0.5417,
+      "step": 7807
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849443631160418e-05,
+      "loss": 0.6077,
+      "step": 7808
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8494060143346764e-05,
+      "loss": 0.5791,
+      "step": 7809
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849368393192854e-05,
+      "loss": 0.5371,
+      "step": 7810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8493307677351425e-05,
+      "loss": 0.6121,
+      "step": 7811
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8492931379617323e-05,
+      "loss": 0.5944,
+      "step": 7812
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849255503872815e-05,
+      "loss": 0.6103,
+      "step": 7813
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8492178654685818e-05,
+      "loss": 0.6159,
+      "step": 7814
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8491802227492237e-05,
+      "loss": 0.6232,
+      "step": 7815
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8491425757149323e-05,
+      "loss": 0.6138,
+      "step": 7816
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8491049243658992e-05,
+      "loss": 0.6585,
+      "step": 7817
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8490672687023147e-05,
+      "loss": 0.6326,
+      "step": 7818
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.849029608724371e-05,
+      "loss": 0.5895,
+      "step": 7819
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8489919444322594e-05,
+      "loss": 0.6515,
+      "step": 7820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848954275826171e-05,
+      "loss": 0.6341,
+      "step": 7821
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848916602906297e-05,
+      "loss": 0.6066,
+      "step": 7822
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8488789256728295e-05,
+      "loss": 0.6209,
+      "step": 7823
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8488412441259595e-05,
+      "loss": 0.6565,
+      "step": 7824
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8488035582658784e-05,
+      "loss": 0.5976,
+      "step": 7825
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848765868092778e-05,
+      "loss": 0.6569,
+      "step": 7826
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8487281736068493e-05,
+      "loss": 0.6057,
+      "step": 7827
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8486904748082845e-05,
+      "loss": 0.594,
+      "step": 7828
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848652771697275e-05,
+      "loss": 0.6428,
+      "step": 7829
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848615064274012e-05,
+      "loss": 0.6465,
+      "step": 7830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8485773525386874e-05,
+      "loss": 0.6424,
+      "step": 7831
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8485396364914926e-05,
+      "loss": 0.623,
+      "step": 7832
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8485019161326195e-05,
+      "loss": 0.5956,
+      "step": 7833
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8484641914622597e-05,
+      "loss": 0.6567,
+      "step": 7834
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8484264624806047e-05,
+      "loss": 0.6126,
+      "step": 7835
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8483887291878465e-05,
+      "loss": 0.5928,
+      "step": 7836
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848350991584177e-05,
+      "loss": 0.5855,
+      "step": 7837
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8483132496697874e-05,
+      "loss": 0.6217,
+      "step": 7838
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8482755034448697e-05,
+      "loss": 0.5666,
+      "step": 7839
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8482377529096157e-05,
+      "loss": 0.6068,
+      "step": 7840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848199998064217e-05,
+      "loss": 0.5945,
+      "step": 7841
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848162238908866e-05,
+      "loss": 0.6511,
+      "step": 7842
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8481244754437547e-05,
+      "loss": 0.6098,
+      "step": 7843
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.848086707669074e-05,
+      "loss": 0.5783,
+      "step": 7844
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8480489355850163e-05,
+      "loss": 0.5849,
+      "step": 7845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8480111591917736e-05,
+      "loss": 0.6091,
+      "step": 7846
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.847973378489538e-05,
+      "loss": 0.5864,
+      "step": 7847
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8479355934785017e-05,
+      "loss": 0.6299,
+      "step": 7848
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.847897804158856e-05,
+      "loss": 0.593,
+      "step": 7849
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.847860010530793e-05,
+      "loss": 0.5479,
+      "step": 7850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8478222125945057e-05,
+      "loss": 0.5959,
+      "step": 7851
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8477844103501844e-05,
+      "loss": 0.5875,
+      "step": 7852
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8477466037980233e-05,
+      "loss": 0.5543,
+      "step": 7853
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8477087929382125e-05,
+      "loss": 0.5755,
+      "step": 7854
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8476709777709457e-05,
+      "loss": 0.6441,
+      "step": 7855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.847633158296414e-05,
+      "loss": 0.5899,
+      "step": 7856
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8475953345148103e-05,
+      "loss": 0.6112,
+      "step": 7857
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8475575064263263e-05,
+      "loss": 0.576,
+      "step": 7858
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8475196740311544e-05,
+      "loss": 0.581,
+      "step": 7859
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.847481837329487e-05,
+      "loss": 0.613,
+      "step": 7860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8474439963215158e-05,
+      "loss": 0.5562,
+      "step": 7861
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8474061510074337e-05,
+      "loss": 0.6426,
+      "step": 7862
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8473683013874328e-05,
+      "loss": 0.6687,
+      "step": 7863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8473304474617056e-05,
+      "loss": 0.6543,
+      "step": 7864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8472925892304438e-05,
+      "loss": 0.6032,
+      "step": 7865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8472547266938407e-05,
+      "loss": 0.5961,
+      "step": 7866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.847216859852088e-05,
+      "loss": 0.6584,
+      "step": 7867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.847178988705378e-05,
+      "loss": 0.5765,
+      "step": 7868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8471411132539037e-05,
+      "loss": 0.5747,
+      "step": 7869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8471032334978573e-05,
+      "loss": 0.5809,
+      "step": 7870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8470653494374314e-05,
+      "loss": 0.5773,
+      "step": 7871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.847027461072818e-05,
+      "loss": 0.5574,
+      "step": 7872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8469895684042104e-05,
+      "loss": 0.5684,
+      "step": 7873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8469516714318006e-05,
+      "loss": 0.5623,
+      "step": 7874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8469137701557814e-05,
+      "loss": 0.6505,
+      "step": 7875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8468758645763452e-05,
+      "loss": 0.5707,
+      "step": 7876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8468379546936848e-05,
+      "loss": 0.6003,
+      "step": 7877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8468000405079927e-05,
+      "loss": 0.6431,
+      "step": 7878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8467621220194617e-05,
+      "loss": 0.5897,
+      "step": 7879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8467241992282842e-05,
+      "loss": 0.6166,
+      "step": 7880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8466862721346533e-05,
+      "loss": 0.5826,
+      "step": 7881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8466483407387615e-05,
+      "loss": 0.6355,
+      "step": 7882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8466104050408012e-05,
+      "loss": 0.6005,
+      "step": 7883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8465724650409655e-05,
+      "loss": 0.5886,
+      "step": 7884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8465345207394473e-05,
+      "loss": 0.6297,
+      "step": 7885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8464965721364393e-05,
+      "loss": 0.6402,
+      "step": 7886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8464586192321346e-05,
+      "loss": 0.6147,
+      "step": 7887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8464206620267253e-05,
+      "loss": 0.6234,
+      "step": 7888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.846382700520405e-05,
+      "loss": 0.5931,
+      "step": 7889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8463447347133664e-05,
+      "loss": 0.6093,
+      "step": 7890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8463067646058022e-05,
+      "loss": 0.6838,
+      "step": 7891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8462687901979053e-05,
+      "loss": 0.6519,
+      "step": 7892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.846230811489869e-05,
+      "loss": 0.5832,
+      "step": 7893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.846192828481886e-05,
+      "loss": 0.6143,
+      "step": 7894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8461548411741493e-05,
+      "loss": 0.6102,
+      "step": 7895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8461168495668523e-05,
+      "loss": 0.6552,
+      "step": 7896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8460788536601882e-05,
+      "loss": 0.6353,
+      "step": 7897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.846040853454349e-05,
+      "loss": 0.5995,
+      "step": 7898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8460028489495288e-05,
+      "loss": 0.5896,
+      "step": 7899
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8459648401459204e-05,
+      "loss": 0.685,
+      "step": 7900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8459268270437166e-05,
+      "loss": 0.633,
+      "step": 7901
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845888809643111e-05,
+      "loss": 0.5871,
+      "step": 7902
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8458507879442968e-05,
+      "loss": 0.5403,
+      "step": 7903
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8458127619474665e-05,
+      "loss": 0.6466,
+      "step": 7904
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845774731652814e-05,
+      "loss": 0.6162,
+      "step": 7905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8457366970605328e-05,
+      "loss": 0.6386,
+      "step": 7906
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8456986581708155e-05,
+      "loss": 0.593,
+      "step": 7907
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8456606149838555e-05,
+      "loss": 0.5741,
+      "step": 7908
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8456225674998468e-05,
+      "loss": 0.6458,
+      "step": 7909
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8455845157189816e-05,
+      "loss": 0.5991,
+      "step": 7910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845546459641454e-05,
+      "loss": 0.6687,
+      "step": 7911
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8455083992674573e-05,
+      "loss": 0.5857,
+      "step": 7912
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845470334597185e-05,
+      "loss": 0.5819,
+      "step": 7913
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8454322656308298e-05,
+      "loss": 0.5681,
+      "step": 7914
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845394192368586e-05,
+      "loss": 0.613,
+      "step": 7915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8453561148106468e-05,
+      "loss": 0.6564,
+      "step": 7916
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8453180329572053e-05,
+      "loss": 0.6456,
+      "step": 7917
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8452799468084553e-05,
+      "loss": 0.6201,
+      "step": 7918
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8452418563645907e-05,
+      "loss": 0.6537,
+      "step": 7919
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8452037616258045e-05,
+      "loss": 0.5774,
+      "step": 7920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8451656625922905e-05,
+      "loss": 0.6262,
+      "step": 7921
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8451275592642423e-05,
+      "loss": 0.6229,
+      "step": 7922
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8450894516418532e-05,
+      "loss": 0.592,
+      "step": 7923
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8450513397253175e-05,
+      "loss": 0.6142,
+      "step": 7924
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.845013223514828e-05,
+      "loss": 0.6164,
+      "step": 7925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844975103010579e-05,
+      "loss": 0.666,
+      "step": 7926
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8449369782127643e-05,
+      "loss": 0.568,
+      "step": 7927
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8448988491215775e-05,
+      "loss": 0.5754,
+      "step": 7928
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8448607157372115e-05,
+      "loss": 0.5713,
+      "step": 7929
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844822578059861e-05,
+      "loss": 0.6553,
+      "step": 7930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8447844360897198e-05,
+      "loss": 0.574,
+      "step": 7931
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8447462898269813e-05,
+      "loss": 0.6107,
+      "step": 7932
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.84470813927184e-05,
+      "loss": 0.642,
+      "step": 7933
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8446699844244885e-05,
+      "loss": 0.6849,
+      "step": 7934
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844631825285122e-05,
+      "loss": 0.6542,
+      "step": 7935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8445936618539336e-05,
+      "loss": 0.6038,
+      "step": 7936
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8445554941311176e-05,
+      "loss": 0.6727,
+      "step": 7937
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844517322116868e-05,
+      "loss": 0.6012,
+      "step": 7938
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844479145811378e-05,
+      "loss": 0.6747,
+      "step": 7939
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844440965214842e-05,
+      "loss": 0.552,
+      "step": 7940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844402780327455e-05,
+      "loss": 0.6219,
+      "step": 7941
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8443645911494093e-05,
+      "loss": 0.6095,
+      "step": 7942
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8443263976809005e-05,
+      "loss": 0.5674,
+      "step": 7943
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8442881999221218e-05,
+      "loss": 0.5891,
+      "step": 7944
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.844249997873267e-05,
+      "loss": 0.6181,
+      "step": 7945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8442117915345315e-05,
+      "loss": 0.6264,
+      "step": 7946
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8441735809061083e-05,
+      "loss": 0.6489,
+      "step": 7947
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8441353659881916e-05,
+      "loss": 0.6507,
+      "step": 7948
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8440971467809764e-05,
+      "loss": 0.6199,
+      "step": 7949
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8440589232846563e-05,
+      "loss": 0.622,
+      "step": 7950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8440206954994252e-05,
+      "loss": 0.5733,
+      "step": 7951
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8439824634254783e-05,
+      "loss": 0.5872,
+      "step": 7952
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843944227063009e-05,
+      "loss": 0.6503,
+      "step": 7953
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843905986412212e-05,
+      "loss": 0.6144,
+      "step": 7954
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8438677414732814e-05,
+      "loss": 0.624,
+      "step": 7955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843829492246412e-05,
+      "loss": 0.6766,
+      "step": 7956
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8437912387317976e-05,
+      "loss": 0.6313,
+      "step": 7957
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8437529809296332e-05,
+      "loss": 0.5836,
+      "step": 7958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8437147188401122e-05,
+      "loss": 0.6559,
+      "step": 7959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.84367645246343e-05,
+      "loss": 0.5647,
+      "step": 7960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8436381817997803e-05,
+      "loss": 0.6607,
+      "step": 7961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8435999068493583e-05,
+      "loss": 0.573,
+      "step": 7962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8435616276123576e-05,
+      "loss": 0.5782,
+      "step": 7963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8435233440889736e-05,
+      "loss": 0.6735,
+      "step": 7964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8434850562794008e-05,
+      "loss": 0.6322,
+      "step": 7965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843446764183833e-05,
+      "loss": 0.5872,
+      "step": 7966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843408467802465e-05,
+      "loss": 0.6478,
+      "step": 7967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843370167135492e-05,
+      "loss": 0.6347,
+      "step": 7968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843331862183108e-05,
+      "loss": 0.63,
+      "step": 7969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8432935529455078e-05,
+      "loss": 0.5903,
+      "step": 7970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843255239422886e-05,
+      "loss": 0.6309,
+      "step": 7971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8432169216154373e-05,
+      "loss": 0.6511,
+      "step": 7972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843178599523357e-05,
+      "loss": 0.5918,
+      "step": 7973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8431402731468386e-05,
+      "loss": 0.6195,
+      "step": 7974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843101942486078e-05,
+      "loss": 0.6242,
+      "step": 7975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8430636075412693e-05,
+      "loss": 0.5738,
+      "step": 7976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.843025268312608e-05,
+      "loss": 0.5832,
+      "step": 7977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8429869248002878e-05,
+      "loss": 0.6033,
+      "step": 7978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842948577004504e-05,
+      "loss": 0.6113,
+      "step": 7979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8429102249254524e-05,
+      "loss": 0.6039,
+      "step": 7980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8428718685633263e-05,
+      "loss": 0.6785,
+      "step": 7981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842833507918322e-05,
+      "loss": 0.6039,
+      "step": 7982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8427951429906334e-05,
+      "loss": 0.6458,
+      "step": 7983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842756773780456e-05,
+      "loss": 0.6257,
+      "step": 7984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842718400287985e-05,
+      "loss": 0.6149,
+      "step": 7985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8426800225134145e-05,
+      "loss": 0.5822,
+      "step": 7986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.84264164045694e-05,
+      "loss": 0.6081,
+      "step": 7987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8426032541187568e-05,
+      "loss": 0.6141,
+      "step": 7988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.84256486349906e-05,
+      "loss": 0.5972,
+      "step": 7989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8425264685980437e-05,
+      "loss": 0.6188,
+      "step": 7990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842488069415904e-05,
+      "loss": 0.6016,
+      "step": 7991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8424496659528356e-05,
+      "loss": 0.5997,
+      "step": 7992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8424112582090337e-05,
+      "loss": 0.5948,
+      "step": 7993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8423728461846938e-05,
+      "loss": 0.5521,
+      "step": 7994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8423344298800105e-05,
+      "loss": 0.6768,
+      "step": 7995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842296009295179e-05,
+      "loss": 0.6217,
+      "step": 7996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8422575844303953e-05,
+      "loss": 0.5933,
+      "step": 7997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.842219155285854e-05,
+      "loss": 0.5582,
+      "step": 7998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8421807218617506e-05,
+      "loss": 0.5937,
+      "step": 7999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8421422841582798e-05,
+      "loss": 0.6219,
+      "step": 8000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8421038421756382e-05,
+      "loss": 0.5839,
+      "step": 8001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8420653959140197e-05,
+      "loss": 0.6523,
+      "step": 8002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8420269453736206e-05,
+      "loss": 0.5525,
+      "step": 8003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.841988490554636e-05,
+      "loss": 0.6148,
+      "step": 8004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8419500314572613e-05,
+      "loss": 0.5858,
+      "step": 8005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8419115680816916e-05,
+      "loss": 0.5965,
+      "step": 8006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8418731004281228e-05,
+      "loss": 0.6684,
+      "step": 8007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8418346284967507e-05,
+      "loss": 0.5803,
+      "step": 8008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8417961522877696e-05,
+      "loss": 0.6114,
+      "step": 8009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8417576718013762e-05,
+      "loss": 0.5537,
+      "step": 8010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8417191870377652e-05,
+      "loss": 0.65,
+      "step": 8011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841680697997133e-05,
+      "loss": 0.5858,
+      "step": 8012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8416422046796746e-05,
+      "loss": 0.5706,
+      "step": 8013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8416037070855853e-05,
+      "loss": 0.5952,
+      "step": 8014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841565205215061e-05,
+      "loss": 0.5835,
+      "step": 8015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841526699068298e-05,
+      "loss": 0.5987,
+      "step": 8016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841488188645491e-05,
+      "loss": 0.6011,
+      "step": 8017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8414496739468362e-05,
+      "loss": 0.7061,
+      "step": 8018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841411154972529e-05,
+      "loss": 0.5822,
+      "step": 8019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8413726317227653e-05,
+      "loss": 0.5954,
+      "step": 8020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8413341041977412e-05,
+      "loss": 0.6042,
+      "step": 8021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8412955723976517e-05,
+      "loss": 0.6409,
+      "step": 8022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841257036322693e-05,
+      "loss": 0.6327,
+      "step": 8023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8412184959730616e-05,
+      "loss": 0.5966,
+      "step": 8024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8411799513489522e-05,
+      "loss": 0.6499,
+      "step": 8025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8411414024505608e-05,
+      "loss": 0.6541,
+      "step": 8026
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8411028492780838e-05,
+      "loss": 0.5822,
+      "step": 8027
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841064291831717e-05,
+      "loss": 0.6054,
+      "step": 8028
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.841025730111656e-05,
+      "loss": 0.599,
+      "step": 8029
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8409871641180975e-05,
+      "loss": 0.6383,
+      "step": 8030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8409485938512366e-05,
+      "loss": 0.6816,
+      "step": 8031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8409100193112697e-05,
+      "loss": 0.6236,
+      "step": 8032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8408714404983925e-05,
+      "loss": 0.6073,
+      "step": 8033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8408328574128016e-05,
+      "loss": 0.548,
+      "step": 8034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8407942700546922e-05,
+      "loss": 0.6432,
+      "step": 8035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8407556784242614e-05,
+      "loss": 0.616,
+      "step": 8036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8407170825217045e-05,
+      "loss": 0.6191,
+      "step": 8037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8406784823472178e-05,
+      "loss": 0.6765,
+      "step": 8038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8406398779009976e-05,
+      "loss": 0.6173,
+      "step": 8039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.84060126918324e-05,
+      "loss": 0.5904,
+      "step": 8040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8405626561941413e-05,
+      "loss": 0.6245,
+      "step": 8041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8405240389338973e-05,
+      "loss": 0.6391,
+      "step": 8042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8404854174027047e-05,
+      "loss": 0.6258,
+      "step": 8043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8404467916007595e-05,
+      "loss": 0.6247,
+      "step": 8044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.840408161528258e-05,
+      "loss": 0.6521,
+      "step": 8045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.840369527185397e-05,
+      "loss": 0.5545,
+      "step": 8046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8403308885723717e-05,
+      "loss": 0.5686,
+      "step": 8047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.840292245689379e-05,
+      "loss": 0.6094,
+      "step": 8048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8402535985366158e-05,
+      "loss": 0.6858,
+      "step": 8049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8402149471142773e-05,
+      "loss": 0.5792,
+      "step": 8050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.840176291422561e-05,
+      "loss": 0.5867,
+      "step": 8051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.840137631461663e-05,
+      "loss": 0.6521,
+      "step": 8052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8400989672317794e-05,
+      "loss": 0.5502,
+      "step": 8053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8400602987331073e-05,
+      "loss": 0.5687,
+      "step": 8054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8400216259658425e-05,
+      "loss": 0.5895,
+      "step": 8055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.839982948930182e-05,
+      "loss": 0.6119,
+      "step": 8056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8399442676263218e-05,
+      "loss": 0.6311,
+      "step": 8057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.839905582054459e-05,
+      "loss": 0.5986,
+      "step": 8058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.83986689221479e-05,
+      "loss": 0.6741,
+      "step": 8059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8398281981075113e-05,
+      "loss": 0.6235,
+      "step": 8060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8397894997328192e-05,
+      "loss": 0.5707,
+      "step": 8061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8397507970909114e-05,
+      "loss": 0.6259,
+      "step": 8062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8397120901819836e-05,
+      "loss": 0.5851,
+      "step": 8063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8396733790062328e-05,
+      "loss": 0.5671,
+      "step": 8064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8396346635638553e-05,
+      "loss": 0.6053,
+      "step": 8065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8395959438550484e-05,
+      "loss": 0.5677,
+      "step": 8066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8395572198800087e-05,
+      "loss": 0.5938,
+      "step": 8067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.839518491638933e-05,
+      "loss": 0.6018,
+      "step": 8068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8394797591320176e-05,
+      "loss": 0.6399,
+      "step": 8069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.83944102235946e-05,
+      "loss": 0.5929,
+      "step": 8070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8394022813214568e-05,
+      "loss": 0.5641,
+      "step": 8071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8393635360182043e-05,
+      "loss": 0.5738,
+      "step": 8072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8393247864499004e-05,
+      "loss": 0.594,
+      "step": 8073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8392860326167413e-05,
+      "loss": 0.5986,
+      "step": 8074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8392472745189237e-05,
+      "loss": 0.5964,
+      "step": 8075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8392085121566455e-05,
+      "loss": 0.5898,
+      "step": 8076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8391697455301026e-05,
+      "loss": 0.6643,
+      "step": 8077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.839130974639493e-05,
+      "loss": 0.5684,
+      "step": 8078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8390921994850125e-05,
+      "loss": 0.5903,
+      "step": 8079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.839053420066859e-05,
+      "loss": 0.5885,
+      "step": 8080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8390146363852294e-05,
+      "loss": 0.6199,
+      "step": 8081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8389758484403207e-05,
+      "loss": 0.581,
+      "step": 8082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.83893705623233e-05,
+      "loss": 0.5674,
+      "step": 8083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8388982597614545e-05,
+      "loss": 0.5716,
+      "step": 8084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8388594590278917e-05,
+      "loss": 0.5945,
+      "step": 8085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8388206540318378e-05,
+      "loss": 0.5901,
+      "step": 8086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8387818447734904e-05,
+      "loss": 0.683,
+      "step": 8087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838743031253047e-05,
+      "loss": 0.6077,
+      "step": 8088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8387042134707048e-05,
+      "loss": 0.6366,
+      "step": 8089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8386653914266604e-05,
+      "loss": 0.5445,
+      "step": 8090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838626565121112e-05,
+      "loss": 0.632,
+      "step": 8091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838587734554256e-05,
+      "loss": 0.572,
+      "step": 8092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8385488997262903e-05,
+      "loss": 0.6136,
+      "step": 8093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838510060637412e-05,
+      "loss": 0.5924,
+      "step": 8094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8384712172878188e-05,
+      "loss": 0.6205,
+      "step": 8095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8384323696777074e-05,
+      "loss": 0.6165,
+      "step": 8096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8383935178072757e-05,
+      "loss": 0.6364,
+      "step": 8097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838354661676721e-05,
+      "loss": 0.5856,
+      "step": 8098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838315801286241e-05,
+      "loss": 0.6318,
+      "step": 8099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8382769366360328e-05,
+      "loss": 0.5599,
+      "step": 8100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8382380677262937e-05,
+      "loss": 0.6601,
+      "step": 8101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838199194557222e-05,
+      "loss": 0.6676,
+      "step": 8102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8381603171290142e-05,
+      "loss": 0.6104,
+      "step": 8103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8381214354418685e-05,
+      "loss": 0.5866,
+      "step": 8104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8380825494959827e-05,
+      "loss": 0.6355,
+      "step": 8105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8380436592915534e-05,
+      "loss": 0.6452,
+      "step": 8106
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.838004764828779e-05,
+      "loss": 0.5596,
+      "step": 8107
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8379658661078574e-05,
+      "loss": 0.6201,
+      "step": 8108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8379269631289855e-05,
+      "loss": 0.5905,
+      "step": 8109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8378880558923614e-05,
+      "loss": 0.6647,
+      "step": 8110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.837849144398183e-05,
+      "loss": 0.5866,
+      "step": 8111
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.837810228646647e-05,
+      "loss": 0.6005,
+      "step": 8112
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8377713086379523e-05,
+      "loss": 0.562,
+      "step": 8113
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.837732384372296e-05,
+      "loss": 0.5944,
+      "step": 8114
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8376934558498763e-05,
+      "loss": 0.5715,
+      "step": 8115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8376545230708907e-05,
+      "loss": 0.6326,
+      "step": 8116
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8376155860355375e-05,
+      "loss": 0.6502,
+      "step": 8117
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8375766447440138e-05,
+      "loss": 0.6067,
+      "step": 8118
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8375376991965182e-05,
+      "loss": 0.5878,
+      "step": 8119
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.837498749393248e-05,
+      "loss": 0.6239,
+      "step": 8120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8374597953344016e-05,
+      "loss": 0.5754,
+      "step": 8121
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8374208370201766e-05,
+      "loss": 0.5705,
+      "step": 8122
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8373818744507706e-05,
+      "loss": 0.6542,
+      "step": 8123
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8373429076263826e-05,
+      "loss": 0.6365,
+      "step": 8124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.83730393654721e-05,
+      "loss": 0.6405,
+      "step": 8125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.837264961213451e-05,
+      "loss": 0.6147,
+      "step": 8126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8372259816253035e-05,
+      "loss": 0.5896,
+      "step": 8127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8371869977829652e-05,
+      "loss": 0.6288,
+      "step": 8128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8371480096866353e-05,
+      "loss": 0.6446,
+      "step": 8129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8371090173365103e-05,
+      "loss": 0.6542,
+      "step": 8130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8370700207327902e-05,
+      "loss": 0.5729,
+      "step": 8131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8370310198756715e-05,
+      "loss": 0.6202,
+      "step": 8132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836992014765353e-05,
+      "loss": 0.5624,
+      "step": 8133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8369530054020332e-05,
+      "loss": 0.5897,
+      "step": 8134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.83691399178591e-05,
+      "loss": 0.6119,
+      "step": 8135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836874973917182e-05,
+      "loss": 0.593,
+      "step": 8136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8368359517960467e-05,
+      "loss": 0.6266,
+      "step": 8137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8367969254227032e-05,
+      "loss": 0.6244,
+      "step": 8138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836757894797349e-05,
+      "loss": 0.5603,
+      "step": 8139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8367188599201833e-05,
+      "loss": 0.6294,
+      "step": 8140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8366798207914042e-05,
+      "loss": 0.6056,
+      "step": 8141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8366407774112094e-05,
+      "loss": 0.5754,
+      "step": 8142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836601729779798e-05,
+      "loss": 0.6972,
+      "step": 8143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836562677897368e-05,
+      "loss": 0.6548,
+      "step": 8144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8365236217641186e-05,
+      "loss": 0.5711,
+      "step": 8145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8364845613802473e-05,
+      "loss": 0.5973,
+      "step": 8146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836445496745953e-05,
+      "loss": 0.6568,
+      "step": 8147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8364064278614338e-05,
+      "loss": 0.6128,
+      "step": 8148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836367354726889e-05,
+      "loss": 0.5562,
+      "step": 8149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836328277342517e-05,
+      "loss": 0.6271,
+      "step": 8150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8362891957085155e-05,
+      "loss": 0.5697,
+      "step": 8151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836250109825084e-05,
+      "loss": 0.5901,
+      "step": 8152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.836211019692421e-05,
+      "loss": 0.6067,
+      "step": 8153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8361719253107244e-05,
+      "loss": 0.5889,
+      "step": 8154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8361328266801937e-05,
+      "loss": 0.594,
+      "step": 8155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.836093723801027e-05,
+      "loss": 0.5746,
+      "step": 8156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8360546166734235e-05,
+      "loss": 0.6587,
+      "step": 8157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8360155052975813e-05,
+      "loss": 0.6667,
+      "step": 8158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8359763896737e-05,
+      "loss": 0.6655,
+      "step": 8159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8359372698019773e-05,
+      "loss": 0.5787,
+      "step": 8160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835898145682613e-05,
+      "loss": 0.6213,
+      "step": 8161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835859017315805e-05,
+      "loss": 0.5635,
+      "step": 8162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835819884701753e-05,
+      "loss": 0.6066,
+      "step": 8163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8357807478406548e-05,
+      "loss": 0.6407,
+      "step": 8164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83574160673271e-05,
+      "loss": 0.578,
+      "step": 8165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8357024613781174e-05,
+      "loss": 0.6302,
+      "step": 8166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835663311777076e-05,
+      "loss": 0.6461,
+      "step": 8167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8356241579297848e-05,
+      "loss": 0.5474,
+      "step": 8168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835584999836442e-05,
+      "loss": 0.6835,
+      "step": 8169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8355458374972476e-05,
+      "loss": 0.6151,
+      "step": 8170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8355066709123996e-05,
+      "loss": 0.5755,
+      "step": 8171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835467500082098e-05,
+      "loss": 0.6708,
+      "step": 8172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835428325006541e-05,
+      "loss": 0.6182,
+      "step": 8173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835389145685928e-05,
+      "loss": 0.6284,
+      "step": 8174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8353499621204582e-05,
+      "loss": 0.5975,
+      "step": 8175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8353107743103307e-05,
+      "loss": 0.5605,
+      "step": 8176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8352715822557445e-05,
+      "loss": 0.6229,
+      "step": 8177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8352323859568984e-05,
+      "loss": 0.5786,
+      "step": 8178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8351931854139924e-05,
+      "loss": 0.6139,
+      "step": 8179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835153980627225e-05,
+      "loss": 0.6179,
+      "step": 8180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8351147715967958e-05,
+      "loss": 0.6351,
+      "step": 8181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.835075558322904e-05,
+      "loss": 0.5386,
+      "step": 8182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8350363408057484e-05,
+      "loss": 0.6175,
+      "step": 8183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8349971190455287e-05,
+      "loss": 0.6218,
+      "step": 8184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834957893042444e-05,
+      "loss": 0.6325,
+      "step": 8185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8349186627966936e-05,
+      "loss": 0.5923,
+      "step": 8186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8348794283084774e-05,
+      "loss": 0.6025,
+      "step": 8187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8348401895779942e-05,
+      "loss": 0.688,
+      "step": 8188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8348009466054433e-05,
+      "loss": 0.6522,
+      "step": 8189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8347616993910243e-05,
+      "loss": 0.5914,
+      "step": 8190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8347224479349367e-05,
+      "loss": 0.5983,
+      "step": 8191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83468319223738e-05,
+      "loss": 0.6276,
+      "step": 8192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8346439322985536e-05,
+      "loss": 0.6308,
+      "step": 8193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8346046681186565e-05,
+      "loss": 0.5689,
+      "step": 8194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834565399697889e-05,
+      "loss": 0.63,
+      "step": 8195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8345261270364506e-05,
+      "loss": 0.5784,
+      "step": 8196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8344868501345402e-05,
+      "loss": 0.6222,
+      "step": 8197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8344475689923577e-05,
+      "loss": 0.6647,
+      "step": 8198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834408283610103e-05,
+      "loss": 0.5924,
+      "step": 8199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8343689939879747e-05,
+      "loss": 0.5728,
+      "step": 8200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834329700126174e-05,
+      "loss": 0.6722,
+      "step": 8201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834290402024899e-05,
+      "loss": 0.586,
+      "step": 8202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8342510996843507e-05,
+      "loss": 0.6526,
+      "step": 8203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8342117931047284e-05,
+      "loss": 0.6287,
+      "step": 8204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8341724822862313e-05,
+      "loss": 0.6678,
+      "step": 8205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8341331672290597e-05,
+      "loss": 0.6658,
+      "step": 8206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.834093847933413e-05,
+      "loss": 0.6583,
+      "step": 8207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8340545243994913e-05,
+      "loss": 0.5769,
+      "step": 8208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8340151966274942e-05,
+      "loss": 0.6503,
+      "step": 8209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8339758646176216e-05,
+      "loss": 0.5692,
+      "step": 8210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8339365283700736e-05,
+      "loss": 0.6294,
+      "step": 8211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83389718788505e-05,
+      "loss": 0.4949,
+      "step": 8212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83385784316275e-05,
+      "loss": 0.6308,
+      "step": 8213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8338184942033743e-05,
+      "loss": 0.6408,
+      "step": 8214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833779141007123e-05,
+      "loss": 0.5514,
+      "step": 8215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8337397835741954e-05,
+      "loss": 0.6927,
+      "step": 8216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833700421904792e-05,
+      "loss": 0.603,
+      "step": 8217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8336610559991124e-05,
+      "loss": 0.6256,
+      "step": 8218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8336216858573567e-05,
+      "loss": 0.5706,
+      "step": 8219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833582311479725e-05,
+      "loss": 0.6199,
+      "step": 8220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833542932866418e-05,
+      "loss": 0.6224,
+      "step": 8221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833503550017635e-05,
+      "loss": 0.6081,
+      "step": 8222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8334641629335764e-05,
+      "loss": 0.5708,
+      "step": 8223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833424771614442e-05,
+      "loss": 0.5851,
+      "step": 8224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8333853760604327e-05,
+      "loss": 0.589,
+      "step": 8225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833345976271748e-05,
+      "loss": 0.6375,
+      "step": 8226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833306572248588e-05,
+      "loss": 0.5744,
+      "step": 8227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8332671639911537e-05,
+      "loss": 0.645,
+      "step": 8228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833227751499645e-05,
+      "loss": 0.6178,
+      "step": 8229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8331883347742617e-05,
+      "loss": 0.6426,
+      "step": 8230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8331489138152045e-05,
+      "loss": 0.6376,
+      "step": 8231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833109488622674e-05,
+      "loss": 0.6552,
+      "step": 8232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83307005919687e-05,
+      "loss": 0.6042,
+      "step": 8233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.833030625537993e-05,
+      "loss": 0.6542,
+      "step": 8234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8329911876462435e-05,
+      "loss": 0.6717,
+      "step": 8235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8329517455218218e-05,
+      "loss": 0.5769,
+      "step": 8236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8329122991649282e-05,
+      "loss": 0.6328,
+      "step": 8237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8328728485757633e-05,
+      "loss": 0.605,
+      "step": 8238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8328333937545278e-05,
+      "loss": 0.604,
+      "step": 8239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8327939347014217e-05,
+      "loss": 0.537,
+      "step": 8240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832754471416646e-05,
+      "loss": 0.5967,
+      "step": 8241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8327150039004003e-05,
+      "loss": 0.556,
+      "step": 8242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8326755321528864e-05,
+      "loss": 0.5376,
+      "step": 8243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832636056174304e-05,
+      "loss": 0.6203,
+      "step": 8244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8325965759648544e-05,
+      "loss": 0.673,
+      "step": 8245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8325570915247374e-05,
+      "loss": 0.5934,
+      "step": 8246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8325176028541537e-05,
+      "loss": 0.5776,
+      "step": 8247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8324781099533045e-05,
+      "loss": 0.6198,
+      "step": 8248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8324386128223907e-05,
+      "loss": 0.5769,
+      "step": 8249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832399111461612e-05,
+      "loss": 0.6323,
+      "step": 8250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8323596058711697e-05,
+      "loss": 0.6215,
+      "step": 8251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8323200960512642e-05,
+      "loss": 0.621,
+      "step": 8252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832280582002097e-05,
+      "loss": 0.6351,
+      "step": 8253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8322410637238685e-05,
+      "loss": 0.6257,
+      "step": 8254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832201541216779e-05,
+      "loss": 0.5804,
+      "step": 8255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8321620144810297e-05,
+      "loss": 0.6129,
+      "step": 8256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832122483516822e-05,
+      "loss": 0.6374,
+      "step": 8257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8320829483243557e-05,
+      "loss": 0.6967,
+      "step": 8258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.832043408903833e-05,
+      "loss": 0.6138,
+      "step": 8259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8320038652554536e-05,
+      "loss": 0.6487,
+      "step": 8260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8319643173794187e-05,
+      "loss": 0.5971,
+      "step": 8261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8319247652759294e-05,
+      "loss": 0.6088,
+      "step": 8262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831885208945187e-05,
+      "loss": 0.577,
+      "step": 8263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8318456483873924e-05,
+      "loss": 0.6879,
+      "step": 8264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831806083602746e-05,
+      "loss": 0.6247,
+      "step": 8265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8317665145914497e-05,
+      "loss": 0.6572,
+      "step": 8266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8317269413537037e-05,
+      "loss": 0.5854,
+      "step": 8267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.83168736388971e-05,
+      "loss": 0.5488,
+      "step": 8268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831647782199669e-05,
+      "loss": 0.6604,
+      "step": 8269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8316081962837822e-05,
+      "loss": 0.5728,
+      "step": 8270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8315686061422503e-05,
+      "loss": 0.5464,
+      "step": 8271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831529011775275e-05,
+      "loss": 0.6306,
+      "step": 8272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8314894131830575e-05,
+      "loss": 0.5888,
+      "step": 8273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831449810365798e-05,
+      "loss": 0.5914,
+      "step": 8274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8314102033236994e-05,
+      "loss": 0.6444,
+      "step": 8275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8313705920569614e-05,
+      "loss": 0.6608,
+      "step": 8276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8313309765657866e-05,
+      "loss": 0.6154,
+      "step": 8277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831291356850375e-05,
+      "loss": 0.5718,
+      "step": 8278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831251732910929e-05,
+      "loss": 0.6516,
+      "step": 8279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8312121047476493e-05,
+      "loss": 0.6643,
+      "step": 8280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8311724723607377e-05,
+      "loss": 0.569,
+      "step": 8281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831132835750395e-05,
+      "loss": 0.6211,
+      "step": 8282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831093194916823e-05,
+      "loss": 0.6065,
+      "step": 8283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8310535498602234e-05,
+      "loss": 0.609,
+      "step": 8284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.831013900580797e-05,
+      "loss": 0.5654,
+      "step": 8285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8309742470787458e-05,
+      "loss": 0.6066,
+      "step": 8286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830934589354271e-05,
+      "loss": 0.6019,
+      "step": 8287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8308949274075742e-05,
+      "loss": 0.6097,
+      "step": 8288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830855261238857e-05,
+      "loss": 0.589,
+      "step": 8289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8308155908483208e-05,
+      "loss": 0.5974,
+      "step": 8290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830775916236167e-05,
+      "loss": 0.624,
+      "step": 8291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830736237402598e-05,
+      "loss": 0.5928,
+      "step": 8292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8306965543478145e-05,
+      "loss": 0.5747,
+      "step": 8293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8306568670720187e-05,
+      "loss": 0.6008,
+      "step": 8294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830617175575412e-05,
+      "loss": 0.6865,
+      "step": 8295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8305774798581964e-05,
+      "loss": 0.6853,
+      "step": 8296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8305377799205733e-05,
+      "loss": 0.623,
+      "step": 8297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830498075762744e-05,
+      "loss": 0.6507,
+      "step": 8298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8304583673849114e-05,
+      "loss": 0.581,
+      "step": 8299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8304186547872764e-05,
+      "loss": 0.6023,
+      "step": 8300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830378937970041e-05,
+      "loss": 0.6356,
+      "step": 8301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.830339216933407e-05,
+      "loss": 0.626,
+      "step": 8302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8302994916775764e-05,
+      "loss": 0.5836,
+      "step": 8303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8302597622027507e-05,
+      "loss": 0.6522,
+      "step": 8304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8302200285091323e-05,
+      "loss": 0.6203,
+      "step": 8305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8301802905969226e-05,
+      "loss": 0.5793,
+      "step": 8306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8301405484663238e-05,
+      "loss": 0.5689,
+      "step": 8307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8301008021175378e-05,
+      "loss": 0.6501,
+      "step": 8308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8300610515507663e-05,
+      "loss": 0.5566,
+      "step": 8309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8300212967662115e-05,
+      "loss": 0.5924,
+      "step": 8310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8299815377640754e-05,
+      "loss": 0.5732,
+      "step": 8311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8299417745445602e-05,
+      "loss": 0.5961,
+      "step": 8312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829902007107868e-05,
+      "loss": 0.659,
+      "step": 8313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8298622354542003e-05,
+      "loss": 0.6225,
+      "step": 8314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8298224595837594e-05,
+      "loss": 0.5756,
+      "step": 8315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8297826794967476e-05,
+      "loss": 0.607,
+      "step": 8316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8297428951933673e-05,
+      "loss": 0.589,
+      "step": 8317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8297031066738203e-05,
+      "loss": 0.6588,
+      "step": 8318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8296633139383087e-05,
+      "loss": 0.5561,
+      "step": 8319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829623516987035e-05,
+      "loss": 0.5653,
+      "step": 8320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829583715820201e-05,
+      "loss": 0.5992,
+      "step": 8321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829543910438009e-05,
+      "loss": 0.615,
+      "step": 8322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829504100840662e-05,
+      "loss": 0.6507,
+      "step": 8323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8294642870283613e-05,
+      "loss": 0.6271,
+      "step": 8324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8294244690013102e-05,
+      "loss": 0.5956,
+      "step": 8325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.82938464675971e-05,
+      "loss": 0.6109,
+      "step": 8326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8293448203037638e-05,
+      "loss": 0.5877,
+      "step": 8327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8293049896336733e-05,
+      "loss": 0.5492,
+      "step": 8328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8292651547496416e-05,
+      "loss": 0.5964,
+      "step": 8329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8292253156518708e-05,
+      "loss": 0.5645,
+      "step": 8330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829185472340563e-05,
+      "loss": 0.6322,
+      "step": 8331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.829145624815921e-05,
+      "loss": 0.594,
+      "step": 8332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8291057730781474e-05,
+      "loss": 0.6107,
+      "step": 8333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8290659171274445e-05,
+      "loss": 0.6522,
+      "step": 8334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8290260569640147e-05,
+      "loss": 0.6,
+      "step": 8335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828986192588061e-05,
+      "loss": 0.5778,
+      "step": 8336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828946323999786e-05,
+      "loss": 0.6712,
+      "step": 8337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8289064511993916e-05,
+      "loss": 0.633,
+      "step": 8338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8288665741870802e-05,
+      "loss": 0.6136,
+      "step": 8339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8288266929630558e-05,
+      "loss": 0.5962,
+      "step": 8340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8287868075275202e-05,
+      "loss": 0.6563,
+      "step": 8341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8287469178806757e-05,
+      "loss": 0.7023,
+      "step": 8342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8287070240227252e-05,
+      "loss": 0.6092,
+      "step": 8343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828667125953872e-05,
+      "loss": 0.6514,
+      "step": 8344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8286272236743186e-05,
+      "loss": 0.593,
+      "step": 8345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8285873171842677e-05,
+      "loss": 0.6494,
+      "step": 8346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828547406483922e-05,
+      "loss": 0.667,
+      "step": 8347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8285074915734835e-05,
+      "loss": 0.6624,
+      "step": 8348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8284675724531566e-05,
+      "loss": 0.5412,
+      "step": 8349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828427649123143e-05,
+      "loss": 0.5567,
+      "step": 8350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8283877215836456e-05,
+      "loss": 0.5766,
+      "step": 8351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828347789834868e-05,
+      "loss": 0.6344,
+      "step": 8352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8283078538770123e-05,
+      "loss": 0.5654,
+      "step": 8353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8282679137102826e-05,
+      "loss": 0.5422,
+      "step": 8354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8282279693348802e-05,
+      "loss": 0.6314,
+      "step": 8355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8281880207510093e-05,
+      "loss": 0.5984,
+      "step": 8356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8281480679588726e-05,
+      "loss": 0.5089,
+      "step": 8357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.828108110958673e-05,
+      "loss": 0.5855,
+      "step": 8358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8280681497506134e-05,
+      "loss": 0.5817,
+      "step": 8359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8280281843348974e-05,
+      "loss": 0.6148,
+      "step": 8360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8279882147117275e-05,
+      "loss": 0.6298,
+      "step": 8361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8279482408813067e-05,
+      "loss": 0.5594,
+      "step": 8362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8279082628438385e-05,
+      "loss": 0.5936,
+      "step": 8363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8278682805995264e-05,
+      "loss": 0.6452,
+      "step": 8364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8278282941485725e-05,
+      "loss": 0.5881,
+      "step": 8365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827788303491181e-05,
+      "loss": 0.5115,
+      "step": 8366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8277483086275546e-05,
+      "loss": 0.6471,
+      "step": 8367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8277083095578966e-05,
+      "loss": 0.5567,
+      "step": 8368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8276683062824103e-05,
+      "loss": 0.629,
+      "step": 8369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827628298801299e-05,
+      "loss": 0.6329,
+      "step": 8370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827588287114766e-05,
+      "loss": 0.5855,
+      "step": 8371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8275482712230143e-05,
+      "loss": 0.63,
+      "step": 8372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8275082511262475e-05,
+      "loss": 0.6069,
+      "step": 8373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8274682268246692e-05,
+      "loss": 0.6259,
+      "step": 8374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827428198318482e-05,
+      "loss": 0.5599,
+      "step": 8375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8273881656078903e-05,
+      "loss": 0.5821,
+      "step": 8376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827348128693097e-05,
+      "loss": 0.6218,
+      "step": 8377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8273080875743053e-05,
+      "loss": 0.5682,
+      "step": 8378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827268042251719e-05,
+      "loss": 0.6341,
+      "step": 8379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8272279927255418e-05,
+      "loss": 0.593,
+      "step": 8380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8271879389959762e-05,
+      "loss": 0.591,
+      "step": 8381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827147881063227e-05,
+      "loss": 0.6614,
+      "step": 8382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.827107818927497e-05,
+      "loss": 0.6117,
+      "step": 8383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.82706775258899e-05,
+      "loss": 0.6706,
+      "step": 8384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8270276820479096e-05,
+      "loss": 0.6427,
+      "step": 8385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826987607304459e-05,
+      "loss": 0.6386,
+      "step": 8386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8269475283588427e-05,
+      "loss": 0.6053,
+      "step": 8387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8269074452112634e-05,
+      "loss": 0.6036,
+      "step": 8388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8268673578619255e-05,
+      "loss": 0.6209,
+      "step": 8389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826827266311032e-05,
+      "loss": 0.5736,
+      "step": 8390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8267871705587873e-05,
+      "loss": 0.5966,
+      "step": 8391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8267470706053948e-05,
+      "loss": 0.6072,
+      "step": 8392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8267069664510583e-05,
+      "loss": 0.6236,
+      "step": 8393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8266668580959815e-05,
+      "loss": 0.5961,
+      "step": 8394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8266267455403684e-05,
+      "loss": 0.6015,
+      "step": 8395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8265866287844226e-05,
+      "loss": 0.6148,
+      "step": 8396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8265465078283484e-05,
+      "loss": 0.6197,
+      "step": 8397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8265063826723494e-05,
+      "loss": 0.6325,
+      "step": 8398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826466253316629e-05,
+      "loss": 0.5842,
+      "step": 8399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8264261197613915e-05,
+      "loss": 0.6083,
+      "step": 8400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826385982006841e-05,
+      "loss": 0.6185,
+      "step": 8401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8263458400531812e-05,
+      "loss": 0.6107,
+      "step": 8402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8263056939006164e-05,
+      "loss": 0.5745,
+      "step": 8403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.82626554354935e-05,
+      "loss": 0.5973,
+      "step": 8404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826225388999587e-05,
+      "loss": 0.6203,
+      "step": 8405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8261852302515304e-05,
+      "loss": 0.5721,
+      "step": 8406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8261450673053848e-05,
+      "loss": 0.6791,
+      "step": 8407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826104900161354e-05,
+      "loss": 0.5139,
+      "step": 8408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8260647288196422e-05,
+      "loss": 0.6135,
+      "step": 8409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.826024553280454e-05,
+      "loss": 0.6648,
+      "step": 8410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825984373543993e-05,
+      "loss": 0.6321,
+      "step": 8411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8259441896104634e-05,
+      "loss": 0.5886,
+      "step": 8412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8259040014800693e-05,
+      "loss": 0.6417,
+      "step": 8413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8258638091530156e-05,
+      "loss": 0.5801,
+      "step": 8414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8258236126295055e-05,
+      "loss": 0.563,
+      "step": 8415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825783411909744e-05,
+      "loss": 0.5696,
+      "step": 8416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825743206993935e-05,
+      "loss": 0.6146,
+      "step": 8417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825702997882283e-05,
+      "loss": 0.6129,
+      "step": 8418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8256627845749927e-05,
+      "loss": 0.5893,
+      "step": 8419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825622567072267e-05,
+      "loss": 0.6284,
+      "step": 8420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825582345374312e-05,
+      "loss": 0.577,
+      "step": 8421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825542119481331e-05,
+      "loss": 0.6144,
+      "step": 8422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8255018893935292e-05,
+      "loss": 0.639,
+      "step": 8423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.82546165511111e-05,
+      "loss": 0.6126,
+      "step": 8424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8254214166342787e-05,
+      "loss": 0.585,
+      "step": 8425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8253811739632393e-05,
+      "loss": 0.5938,
+      "step": 8426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8253409270981966e-05,
+      "loss": 0.6675,
+      "step": 8427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8253006760393548e-05,
+      "loss": 0.616,
+      "step": 8428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8252604207869183e-05,
+      "loss": 0.6354,
+      "step": 8429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8252201613410924e-05,
+      "loss": 0.6245,
+      "step": 8430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825179897702081e-05,
+      "loss": 0.6243,
+      "step": 8431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825139629870089e-05,
+      "loss": 0.4906,
+      "step": 8432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8250993578453205e-05,
+      "loss": 0.5994,
+      "step": 8433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825059081627981e-05,
+      "loss": 0.6668,
+      "step": 8434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.825018801218274e-05,
+      "loss": 0.5942,
+      "step": 8435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8249785166164054e-05,
+      "loss": 0.5797,
+      "step": 8436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8249382278225792e-05,
+      "loss": 0.597,
+      "step": 8437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8248979348370004e-05,
+      "loss": 0.6375,
+      "step": 8438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8248576376598737e-05,
+      "loss": 0.623,
+      "step": 8439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8248173362914033e-05,
+      "loss": 0.6309,
+      "step": 8440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.824777030731795e-05,
+      "loss": 0.5444,
+      "step": 8441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8247367209812528e-05,
+      "loss": 0.5691,
+      "step": 8442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8246964070399817e-05,
+      "loss": 0.6227,
+      "step": 8443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8246560889081868e-05,
+      "loss": 0.5662,
+      "step": 8444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8246157665860725e-05,
+      "loss": 0.6192,
+      "step": 8445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.824575440073844e-05,
+      "loss": 0.5703,
+      "step": 8446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8245351093717066e-05,
+      "loss": 0.6571,
+      "step": 8447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.824494774479865e-05,
+      "loss": 0.5943,
+      "step": 8448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8244544353985235e-05,
+      "loss": 0.6294,
+      "step": 8449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8244140921278878e-05,
+      "loss": 0.5919,
+      "step": 8450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8243737446681623e-05,
+      "loss": 0.5707,
+      "step": 8451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8243333930195525e-05,
+      "loss": 0.5885,
+      "step": 8452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8242930371822635e-05,
+      "loss": 0.6395,
+      "step": 8453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8242526771565e-05,
+      "loss": 0.577,
+      "step": 8454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8242123129424675e-05,
+      "loss": 0.5778,
+      "step": 8455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8241719445403707e-05,
+      "loss": 0.5929,
+      "step": 8456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.824131571950415e-05,
+      "loss": 0.5945,
+      "step": 8457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8240911951728053e-05,
+      "loss": 0.6229,
+      "step": 8458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.824050814207747e-05,
+      "loss": 0.6171,
+      "step": 8459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.824010429055445e-05,
+      "loss": 0.6494,
+      "step": 8460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8239700397161048e-05,
+      "loss": 0.623,
+      "step": 8461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8239296461899313e-05,
+      "loss": 0.5427,
+      "step": 8462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.82388924847713e-05,
+      "loss": 0.5733,
+      "step": 8463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8238488465779066e-05,
+      "loss": 0.5635,
+      "step": 8464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8238084404924656e-05,
+      "loss": 0.5735,
+      "step": 8465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.823768030221013e-05,
+      "loss": 0.5936,
+      "step": 8466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8237276157637534e-05,
+      "loss": 0.5835,
+      "step": 8467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8236871971208925e-05,
+      "loss": 0.6082,
+      "step": 8468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8236467742926357e-05,
+      "loss": 0.5856,
+      "step": 8469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8236063472791885e-05,
+      "loss": 0.6007,
+      "step": 8470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8235659160807565e-05,
+      "loss": 0.6099,
+      "step": 8471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8235254806975447e-05,
+      "loss": 0.5855,
+      "step": 8472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8234850411297587e-05,
+      "loss": 0.5853,
+      "step": 8473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.823444597377604e-05,
+      "loss": 0.6313,
+      "step": 8474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.823404149441286e-05,
+      "loss": 0.6661,
+      "step": 8475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.823363697321011e-05,
+      "loss": 0.6191,
+      "step": 8476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8233232410169834e-05,
+      "loss": 0.6083,
+      "step": 8477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8232827805294094e-05,
+      "loss": 0.6932,
+      "step": 8478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8232423158584945e-05,
+      "loss": 0.6179,
+      "step": 8479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8232018470044444e-05,
+      "loss": 0.617,
+      "step": 8480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8231613739674644e-05,
+      "loss": 0.6285,
+      "step": 8481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8231208967477605e-05,
+      "loss": 0.5949,
+      "step": 8482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8230804153455383e-05,
+      "loss": 0.5586,
+      "step": 8483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8230399297610035e-05,
+      "loss": 0.6415,
+      "step": 8484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8229994399943618e-05,
+      "loss": 0.5998,
+      "step": 8485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8229589460458185e-05,
+      "loss": 0.6039,
+      "step": 8486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8229184479155802e-05,
+      "loss": 0.5936,
+      "step": 8487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8228779456038524e-05,
+      "loss": 0.6112,
+      "step": 8488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8228374391108406e-05,
+      "loss": 0.6242,
+      "step": 8489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8227969284367504e-05,
+      "loss": 0.6264,
+      "step": 8490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8227564135817887e-05,
+      "loss": 0.6109,
+      "step": 8491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8227158945461604e-05,
+      "loss": 0.6459,
+      "step": 8492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8226753713300717e-05,
+      "loss": 0.6239,
+      "step": 8493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8226348439337284e-05,
+      "loss": 0.6216,
+      "step": 8494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8225943123573368e-05,
+      "loss": 0.6263,
+      "step": 8495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8225537766011022e-05,
+      "loss": 0.6219,
+      "step": 8496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.822513236665231e-05,
+      "loss": 0.5558,
+      "step": 8497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8224726925499294e-05,
+      "loss": 0.6225,
+      "step": 8498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8224321442554032e-05,
+      "loss": 0.6172,
+      "step": 8499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.822391591781858e-05,
+      "loss": 0.5996,
+      "step": 8500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.822351035129501e-05,
+      "loss": 0.6161,
+      "step": 8501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8223104742985368e-05,
+      "loss": 0.657,
+      "step": 8502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.822269909289173e-05,
+      "loss": 0.5784,
+      "step": 8503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8222293401016145e-05,
+      "loss": 0.6229,
+      "step": 8504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8221887667360678e-05,
+      "loss": 0.6027,
+      "step": 8505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8221481891927394e-05,
+      "loss": 0.639,
+      "step": 8506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8221076074718357e-05,
+      "loss": 0.6141,
+      "step": 8507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.822067021573562e-05,
+      "loss": 0.6551,
+      "step": 8508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8220264314981247e-05,
+      "loss": 0.5816,
+      "step": 8509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8219858372457307e-05,
+      "loss": 0.6873,
+      "step": 8510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821945238816586e-05,
+      "loss": 0.5838,
+      "step": 8511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821904636210897e-05,
+      "loss": 0.636,
+      "step": 8512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8218640294288694e-05,
+      "loss": 0.6291,
+      "step": 8513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8218234184707102e-05,
+      "loss": 0.6675,
+      "step": 8514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8217828033366258e-05,
+      "loss": 0.6446,
+      "step": 8515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821742184026822e-05,
+      "loss": 0.639,
+      "step": 8516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8217015605415054e-05,
+      "loss": 0.6479,
+      "step": 8517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8216609328808824e-05,
+      "loss": 0.6465,
+      "step": 8518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.82162030104516e-05,
+      "loss": 0.5931,
+      "step": 8519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8215796650345437e-05,
+      "loss": 0.626,
+      "step": 8520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821539024849241e-05,
+      "loss": 0.6196,
+      "step": 8521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821498380489458e-05,
+      "loss": 0.5819,
+      "step": 8522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821457731955401e-05,
+      "loss": 0.6366,
+      "step": 8523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8214170792472764e-05,
+      "loss": 0.5521,
+      "step": 8524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8213764223652914e-05,
+      "loss": 0.6006,
+      "step": 8525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821335761309652e-05,
+      "loss": 0.5994,
+      "step": 8526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.821295096080565e-05,
+      "loss": 0.5965,
+      "step": 8527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8212544266782374e-05,
+      "loss": 0.5759,
+      "step": 8528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8212137531028755e-05,
+      "loss": 0.6198,
+      "step": 8529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8211730753546855e-05,
+      "loss": 0.5975,
+      "step": 8530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8211323934338753e-05,
+      "loss": 0.7057,
+      "step": 8531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8210917073406506e-05,
+      "loss": 0.559,
+      "step": 8532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8210510170752185e-05,
+      "loss": 0.625,
+      "step": 8533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8210103226377858e-05,
+      "loss": 0.5727,
+      "step": 8534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8209696240285593e-05,
+      "loss": 0.604,
+      "step": 8535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8209289212477454e-05,
+      "loss": 0.566,
+      "step": 8536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8208882142955514e-05,
+      "loss": 0.5855,
+      "step": 8537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8208475031721843e-05,
+      "loss": 0.6548,
+      "step": 8538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8208067878778498e-05,
+      "loss": 0.5811,
+      "step": 8539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8207660684127565e-05,
+      "loss": 0.5927,
+      "step": 8540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8207253447771102e-05,
+      "loss": 0.6792,
+      "step": 8541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8206846169711175e-05,
+      "loss": 0.6428,
+      "step": 8542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8206438849949864e-05,
+      "loss": 0.6084,
+      "step": 8543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8206031488489237e-05,
+      "loss": 0.589,
+      "step": 8544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8205624085331356e-05,
+      "loss": 0.6293,
+      "step": 8545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8205216640478295e-05,
+      "loss": 0.6327,
+      "step": 8546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8204809153932128e-05,
+      "loss": 0.5618,
+      "step": 8547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8204401625694924e-05,
+      "loss": 0.6853,
+      "step": 8548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8203994055768748e-05,
+      "loss": 0.5829,
+      "step": 8549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.820358644415568e-05,
+      "loss": 0.6008,
+      "step": 8550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8203178790857782e-05,
+      "loss": 0.6104,
+      "step": 8551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8202771095877135e-05,
+      "loss": 0.5854,
+      "step": 8552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8202363359215803e-05,
+      "loss": 0.6265,
+      "step": 8553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8201955580875862e-05,
+      "loss": 0.6682,
+      "step": 8554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.820154776085938e-05,
+      "loss": 0.5782,
+      "step": 8555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8201139899168433e-05,
+      "loss": 0.652,
+      "step": 8556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8200731995805094e-05,
+      "loss": 0.5734,
+      "step": 8557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8200324050771434e-05,
+      "loss": 0.6131,
+      "step": 8558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8199916064069523e-05,
+      "loss": 0.5886,
+      "step": 8559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8199508035701442e-05,
+      "loss": 0.6286,
+      "step": 8560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8199099965669257e-05,
+      "loss": 0.5613,
+      "step": 8561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8198691853975044e-05,
+      "loss": 0.5784,
+      "step": 8562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8198283700620875e-05,
+      "loss": 0.6091,
+      "step": 8563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8197875505608826e-05,
+      "loss": 0.5977,
+      "step": 8564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.819746726894097e-05,
+      "loss": 0.6046,
+      "step": 8565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.819705899061938e-05,
+      "loss": 0.5968,
+      "step": 8566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.819665067064614e-05,
+      "loss": 0.5978,
+      "step": 8567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8196242309023313e-05,
+      "loss": 0.6505,
+      "step": 8568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8195833905752978e-05,
+      "loss": 0.6445,
+      "step": 8569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8195425460837213e-05,
+      "loss": 0.581,
+      "step": 8570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8195016974278087e-05,
+      "loss": 0.5558,
+      "step": 8571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8194608446077686e-05,
+      "loss": 0.5689,
+      "step": 8572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8194199876238074e-05,
+      "loss": 0.6502,
+      "step": 8573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8193791264761334e-05,
+      "loss": 0.6721,
+      "step": 8574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8193382611649544e-05,
+      "loss": 0.6522,
+      "step": 8575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8192973916904773e-05,
+      "loss": 0.6606,
+      "step": 8576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8192565180529105e-05,
+      "loss": 0.6406,
+      "step": 8577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8192156402524613e-05,
+      "loss": 0.5952,
+      "step": 8578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8191747582893376e-05,
+      "loss": 0.6155,
+      "step": 8579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.819133872163747e-05,
+      "loss": 0.6707,
+      "step": 8580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8190929818758972e-05,
+      "loss": 0.5811,
+      "step": 8581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.819052087425996e-05,
+      "loss": 0.5875,
+      "step": 8582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8190111888142513e-05,
+      "loss": 0.6489,
+      "step": 8583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8189702860408714e-05,
+      "loss": 0.5932,
+      "step": 8584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.818929379106063e-05,
+      "loss": 0.6177,
+      "step": 8585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.818888468010035e-05,
+      "loss": 0.6341,
+      "step": 8586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8188475527529946e-05,
+      "loss": 0.598,
+      "step": 8587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.81880663333515e-05,
+      "loss": 0.63,
+      "step": 8588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.818765709756709e-05,
+      "loss": 0.6596,
+      "step": 8589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.81872478201788e-05,
+      "loss": 0.647,
+      "step": 8590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.81868385011887e-05,
+      "loss": 0.6523,
+      "step": 8591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8186429140598886e-05,
+      "loss": 0.6006,
+      "step": 8592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.818601973841142e-05,
+      "loss": 0.6069,
+      "step": 8593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8185610294628395e-05,
+      "loss": 0.5882,
+      "step": 8594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8185200809251883e-05,
+      "loss": 0.6965,
+      "step": 8595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.818479128228397e-05,
+      "loss": 0.648,
+      "step": 8596
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8184381713726732e-05,
+      "loss": 0.6093,
+      "step": 8597
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8183972103582257e-05,
+      "loss": 0.6695,
+      "step": 8598
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8183562451852625e-05,
+      "loss": 0.6111,
+      "step": 8599
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.818315275853991e-05,
+      "loss": 0.6465,
+      "step": 8600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8182743023646204e-05,
+      "loss": 0.5891,
+      "step": 8601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8182333247173583e-05,
+      "loss": 0.638,
+      "step": 8602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8181923429124132e-05,
+      "loss": 0.5937,
+      "step": 8603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.818151356949993e-05,
+      "loss": 0.6279,
+      "step": 8604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8181103668303062e-05,
+      "loss": 0.5729,
+      "step": 8605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.818069372553561e-05,
+      "loss": 0.6237,
+      "step": 8606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.818028374119966e-05,
+      "loss": 0.5915,
+      "step": 8607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8179873715297286e-05,
+      "loss": 0.5811,
+      "step": 8608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817946364783058e-05,
+      "loss": 0.7168,
+      "step": 8609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8179053538801627e-05,
+      "loss": 0.6059,
+      "step": 8610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817864338821251e-05,
+      "loss": 0.6336,
+      "step": 8611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8178233196065303e-05,
+      "loss": 0.6163,
+      "step": 8612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8177822962362105e-05,
+      "loss": 0.6236,
+      "step": 8613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817741268710499e-05,
+      "loss": 0.5562,
+      "step": 8614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8177002370296046e-05,
+      "loss": 0.663,
+      "step": 8615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817659201193736e-05,
+      "loss": 0.6027,
+      "step": 8616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8176181612031014e-05,
+      "loss": 0.6487,
+      "step": 8617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8175771170579097e-05,
+      "loss": 0.6391,
+      "step": 8618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817536068758369e-05,
+      "loss": 0.5714,
+      "step": 8619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817495016304688e-05,
+      "loss": 0.6972,
+      "step": 8620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8174539596970756e-05,
+      "loss": 0.66,
+      "step": 8621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8174128989357402e-05,
+      "loss": 0.6391,
+      "step": 8622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8173718340208907e-05,
+      "loss": 0.5916,
+      "step": 8623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817330764952735e-05,
+      "loss": 0.5787,
+      "step": 8624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8172896917314828e-05,
+      "loss": 0.6079,
+      "step": 8625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8172486143573424e-05,
+      "loss": 0.5834,
+      "step": 8626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817207532830522e-05,
+      "loss": 0.6114,
+      "step": 8627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8171664471512307e-05,
+      "loss": 0.604,
+      "step": 8628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817125357319678e-05,
+      "loss": 0.6215,
+      "step": 8629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8170842633360715e-05,
+      "loss": 0.5773,
+      "step": 8630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.817043165200621e-05,
+      "loss": 0.5704,
+      "step": 8631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8170020629135346e-05,
+      "loss": 0.6587,
+      "step": 8632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8169609564750216e-05,
+      "loss": 0.5796,
+      "step": 8633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8169198458852903e-05,
+      "loss": 0.5755,
+      "step": 8634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8168787311445503e-05,
+      "loss": 0.6201,
+      "step": 8635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8168376122530104e-05,
+      "loss": 0.6211,
+      "step": 8636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8167964892108793e-05,
+      "loss": 0.644,
+      "step": 8637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8167553620183656e-05,
+      "loss": 0.5933,
+      "step": 8638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8167142306756793e-05,
+      "loss": 0.5948,
+      "step": 8639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8166730951830288e-05,
+      "loss": 0.5645,
+      "step": 8640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816631955540623e-05,
+      "loss": 0.6104,
+      "step": 8641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816590811748671e-05,
+      "loss": 0.6062,
+      "step": 8642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816549663807382e-05,
+      "loss": 0.5883,
+      "step": 8643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8165085117169646e-05,
+      "loss": 0.6445,
+      "step": 8644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816467355477629e-05,
+      "loss": 0.5443,
+      "step": 8645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816426195089583e-05,
+      "loss": 0.6184,
+      "step": 8646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8163850305530368e-05,
+      "loss": 0.6073,
+      "step": 8647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8163438618681993e-05,
+      "loss": 0.6089,
+      "step": 8648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8163026890352794e-05,
+      "loss": 0.5713,
+      "step": 8649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8162615120544863e-05,
+      "loss": 0.643,
+      "step": 8650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8162203309260296e-05,
+      "loss": 0.6446,
+      "step": 8651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816179145650118e-05,
+      "loss": 0.6091,
+      "step": 8652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8161379562269616e-05,
+      "loss": 0.5834,
+      "step": 8653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.816096762656769e-05,
+      "loss": 0.6376,
+      "step": 8654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8160555649397498e-05,
+      "loss": 0.5715,
+      "step": 8655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8160143630761133e-05,
+      "loss": 0.6083,
+      "step": 8656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815973157066069e-05,
+      "loss": 0.5972,
+      "step": 8657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8159319469098257e-05,
+      "loss": 0.5574,
+      "step": 8658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8158907326075937e-05,
+      "loss": 0.5912,
+      "step": 8659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8158495141595816e-05,
+      "loss": 0.5853,
+      "step": 8660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8158082915659993e-05,
+      "loss": 0.5679,
+      "step": 8661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815767064827056e-05,
+      "loss": 0.5797,
+      "step": 8662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8157258339429614e-05,
+      "loss": 0.5933,
+      "step": 8663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8156845989139253e-05,
+      "loss": 0.5446,
+      "step": 8664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8156433597401565e-05,
+      "loss": 0.672,
+      "step": 8665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815602116421865e-05,
+      "loss": 0.6378,
+      "step": 8666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8155608689592604e-05,
+      "loss": 0.6109,
+      "step": 8667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815519617352552e-05,
+      "loss": 0.6025,
+      "step": 8668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8154783616019495e-05,
+      "loss": 0.5587,
+      "step": 8669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815437101707663e-05,
+      "loss": 0.6381,
+      "step": 8670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8153958376699016e-05,
+      "loss": 0.5545,
+      "step": 8671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8153545694888752e-05,
+      "loss": 0.6147,
+      "step": 8672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815313297164793e-05,
+      "loss": 0.5794,
+      "step": 8673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8152720206978655e-05,
+      "loss": 0.6116,
+      "step": 8674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8152307400883023e-05,
+      "loss": 0.5748,
+      "step": 8675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8151894553363126e-05,
+      "loss": 0.6355,
+      "step": 8676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8151481664421065e-05,
+      "loss": 0.5808,
+      "step": 8677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8151068734058944e-05,
+      "loss": 0.6487,
+      "step": 8678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8150655762278853e-05,
+      "loss": 0.6048,
+      "step": 8679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.815024274908289e-05,
+      "loss": 0.593,
+      "step": 8680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8149829694473162e-05,
+      "loss": 0.6461,
+      "step": 8681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8149416598451755e-05,
+      "loss": 0.5983,
+      "step": 8682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8149003461020782e-05,
+      "loss": 0.6648,
+      "step": 8683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8148590282182335e-05,
+      "loss": 0.601,
+      "step": 8684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8148177061938512e-05,
+      "loss": 0.6298,
+      "step": 8685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8147763800291414e-05,
+      "loss": 0.6074,
+      "step": 8686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8147350497243145e-05,
+      "loss": 0.5866,
+      "step": 8687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8146937152795804e-05,
+      "loss": 0.6309,
+      "step": 8688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8146523766951484e-05,
+      "loss": 0.5646,
+      "step": 8689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8146110339712294e-05,
+      "loss": 0.5993,
+      "step": 8690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8145696871080332e-05,
+      "loss": 0.6482,
+      "step": 8691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8145283361057698e-05,
+      "loss": 0.6162,
+      "step": 8692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8144869809646493e-05,
+      "loss": 0.6566,
+      "step": 8693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8144456216848822e-05,
+      "loss": 0.6507,
+      "step": 8694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8144042582666777e-05,
+      "loss": 0.5734,
+      "step": 8695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814362890710247e-05,
+      "loss": 0.5896,
+      "step": 8696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8143215190158003e-05,
+      "loss": 0.6323,
+      "step": 8697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814280143183547e-05,
+      "loss": 0.5698,
+      "step": 8698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814238763213698e-05,
+      "loss": 0.6903,
+      "step": 8699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814197379106464e-05,
+      "loss": 0.6148,
+      "step": 8700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8141559908620537e-05,
+      "loss": 0.5747,
+      "step": 8701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814114598480679e-05,
+      "loss": 0.6477,
+      "step": 8702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.814073201962549e-05,
+      "loss": 0.638,
+      "step": 8703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8140318013078754e-05,
+      "loss": 0.5961,
+      "step": 8704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8139903965168674e-05,
+      "loss": 0.6301,
+      "step": 8705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8139489875897358e-05,
+      "loss": 0.596,
+      "step": 8706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813907574526691e-05,
+      "loss": 0.5962,
+      "step": 8707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8138661573279434e-05,
+      "loss": 0.6059,
+      "step": 8708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813824735993704e-05,
+      "loss": 0.5509,
+      "step": 8709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813783310524182e-05,
+      "loss": 0.586,
+      "step": 8710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8137418809195893e-05,
+      "loss": 0.5348,
+      "step": 8711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8137004471801357e-05,
+      "loss": 0.6067,
+      "step": 8712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8136590093060316e-05,
+      "loss": 0.5718,
+      "step": 8713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8136175672974876e-05,
+      "loss": 0.6593,
+      "step": 8714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8135761211547146e-05,
+      "loss": 0.6249,
+      "step": 8715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8135346708779235e-05,
+      "loss": 0.5597,
+      "step": 8716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813493216467324e-05,
+      "loss": 0.6487,
+      "step": 8717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8134517579231274e-05,
+      "loss": 0.6214,
+      "step": 8718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8134102952455442e-05,
+      "loss": 0.6606,
+      "step": 8719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813368828434785e-05,
+      "loss": 0.6026,
+      "step": 8720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8133273574910607e-05,
+      "loss": 0.6618,
+      "step": 8721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813285882414582e-05,
+      "loss": 0.547,
+      "step": 8722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8132444032055592e-05,
+      "loss": 0.6216,
+      "step": 8723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8132029198642034e-05,
+      "loss": 0.546,
+      "step": 8724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.813161432390726e-05,
+      "loss": 0.5972,
+      "step": 8725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8131199407853365e-05,
+      "loss": 0.6264,
+      "step": 8726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8130784450482468e-05,
+      "loss": 0.6366,
+      "step": 8727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8130369451796675e-05,
+      "loss": 0.6236,
+      "step": 8728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8129954411798093e-05,
+      "loss": 0.6003,
+      "step": 8729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.812953933048883e-05,
+      "loss": 0.5847,
+      "step": 8730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8129124207870997e-05,
+      "loss": 0.624,
+      "step": 8731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8128709043946705e-05,
+      "loss": 0.5916,
+      "step": 8732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8128293838718063e-05,
+      "loss": 0.6117,
+      "step": 8733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8127878592187177e-05,
+      "loss": 0.6359,
+      "step": 8734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8127463304356157e-05,
+      "loss": 0.598,
+      "step": 8735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.812704797522712e-05,
+      "loss": 0.5707,
+      "step": 8736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8126632604802173e-05,
+      "loss": 0.6428,
+      "step": 8737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.812621719308342e-05,
+      "loss": 0.5137,
+      "step": 8738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8125801740072982e-05,
+      "loss": 0.5937,
+      "step": 8739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8125386245772965e-05,
+      "loss": 0.6292,
+      "step": 8740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8124970710185482e-05,
+      "loss": 0.6436,
+      "step": 8741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.812455513331264e-05,
+      "loss": 0.5912,
+      "step": 8742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8124139515156556e-05,
+      "loss": 0.6171,
+      "step": 8743
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8123723855719338e-05,
+      "loss": 0.633,
+      "step": 8744
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8123308155003104e-05,
+      "loss": 0.5775,
+      "step": 8745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.812289241300996e-05,
+      "loss": 0.6133,
+      "step": 8746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.812247662974202e-05,
+      "loss": 0.6062,
+      "step": 8747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.81220608052014e-05,
+      "loss": 0.6525,
+      "step": 8748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8121644939390208e-05,
+      "loss": 0.6218,
+      "step": 8749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8121229032310558e-05,
+      "loss": 0.6142,
+      "step": 8750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8120813083964567e-05,
+      "loss": 0.535,
+      "step": 8751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8120397094354346e-05,
+      "loss": 0.6205,
+      "step": 8752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811998106348201e-05,
+      "loss": 0.6198,
+      "step": 8753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8119564991349672e-05,
+      "loss": 0.5721,
+      "step": 8754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8119148877959444e-05,
+      "loss": 0.6206,
+      "step": 8755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8118732723313445e-05,
+      "loss": 0.5955,
+      "step": 8756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8118316527413785e-05,
+      "loss": 0.621,
+      "step": 8757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8117900290262587e-05,
+      "loss": 0.5673,
+      "step": 8758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8117484011861955e-05,
+      "loss": 0.6428,
+      "step": 8759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811706769221401e-05,
+      "loss": 0.6849,
+      "step": 8760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811665133132087e-05,
+      "loss": 0.6597,
+      "step": 8761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8116234929184644e-05,
+      "loss": 0.6242,
+      "step": 8762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8115818485807452e-05,
+      "loss": 0.5794,
+      "step": 8763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811540200119141e-05,
+      "loss": 0.6309,
+      "step": 8764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8114985475338635e-05,
+      "loss": 0.6107,
+      "step": 8765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811456890825124e-05,
+      "loss": 0.5813,
+      "step": 8766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8114152299931343e-05,
+      "loss": 0.5909,
+      "step": 8767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8113735650381067e-05,
+      "loss": 0.6306,
+      "step": 8768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811331895960252e-05,
+      "loss": 0.6148,
+      "step": 8769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8112902227597824e-05,
+      "loss": 0.573,
+      "step": 8770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8112485454369095e-05,
+      "loss": 0.571,
+      "step": 8771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811206863991845e-05,
+      "loss": 0.5768,
+      "step": 8772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8111651784248013e-05,
+      "loss": 0.6011,
+      "step": 8773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8111234887359895e-05,
+      "loss": 0.5988,
+      "step": 8774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.811081794925622e-05,
+      "loss": 0.6052,
+      "step": 8775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.81104009699391e-05,
+      "loss": 0.6101,
+      "step": 8776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810998394941066e-05,
+      "loss": 0.6309,
+      "step": 8777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8109566887673013e-05,
+      "loss": 0.6479,
+      "step": 8778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8109149784728283e-05,
+      "loss": 0.5618,
+      "step": 8779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810873264057859e-05,
+      "loss": 0.5893,
+      "step": 8780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8108315455226048e-05,
+      "loss": 0.571,
+      "step": 8781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8107898228672788e-05,
+      "loss": 0.597,
+      "step": 8782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8107480960920916e-05,
+      "loss": 0.6074,
+      "step": 8783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810706365197256e-05,
+      "loss": 0.5722,
+      "step": 8784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8106646301829836e-05,
+      "loss": 0.6374,
+      "step": 8785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8106228910494874e-05,
+      "loss": 0.599,
+      "step": 8786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8105811477969788e-05,
+      "loss": 0.5784,
+      "step": 8787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8105394004256698e-05,
+      "loss": 0.6059,
+      "step": 8788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8104976489357725e-05,
+      "loss": 0.6805,
+      "step": 8789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8104558933274997e-05,
+      "loss": 0.609,
+      "step": 8790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8104141336010632e-05,
+      "loss": 0.6515,
+      "step": 8791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8103723697566747e-05,
+      "loss": 0.5424,
+      "step": 8792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8103306017945474e-05,
+      "loss": 0.6235,
+      "step": 8793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810288829714893e-05,
+      "loss": 0.637,
+      "step": 8794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810247053517923e-05,
+      "loss": 0.6434,
+      "step": 8795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810205273203851e-05,
+      "loss": 0.6128,
+      "step": 8796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.810163488772889e-05,
+      "loss": 0.6066,
+      "step": 8797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8101217002252483e-05,
+      "loss": 0.6325,
+      "step": 8798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8100799075611425e-05,
+      "loss": 0.6655,
+      "step": 8799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8100381107807835e-05,
+      "loss": 0.6602,
+      "step": 8800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809996309884383e-05,
+      "loss": 0.6081,
+      "step": 8801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8099545048721547e-05,
+      "loss": 0.6114,
+      "step": 8802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.80991269574431e-05,
+      "loss": 0.6041,
+      "step": 8803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8098708825010618e-05,
+      "loss": 0.6915,
+      "step": 8804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8098290651426228e-05,
+      "loss": 0.5693,
+      "step": 8805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8097872436692046e-05,
+      "loss": 0.6433,
+      "step": 8806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8097454180810206e-05,
+      "loss": 0.5904,
+      "step": 8807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809703588378283e-05,
+      "loss": 0.658,
+      "step": 8808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809661754561204e-05,
+      "loss": 0.6045,
+      "step": 8809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809619916629997e-05,
+      "loss": 0.6238,
+      "step": 8810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8095780745848735e-05,
+      "loss": 0.5949,
+      "step": 8811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809536228426047e-05,
+      "loss": 0.5707,
+      "step": 8812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8094943781537303e-05,
+      "loss": 0.6837,
+      "step": 8813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809452523768135e-05,
+      "loss": 0.55,
+      "step": 8814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8094106652694746e-05,
+      "loss": 0.6702,
+      "step": 8815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8093688026579612e-05,
+      "loss": 0.6248,
+      "step": 8816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8093269359338085e-05,
+      "loss": 0.657,
+      "step": 8817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809285065097228e-05,
+      "loss": 0.6247,
+      "step": 8818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8092431901484336e-05,
+      "loss": 0.5154,
+      "step": 8819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8092013110876373e-05,
+      "loss": 0.5987,
+      "step": 8820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809159427915052e-05,
+      "loss": 0.6206,
+      "step": 8821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809117540630891e-05,
+      "loss": 0.588,
+      "step": 8822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8090756492353665e-05,
+      "loss": 0.6085,
+      "step": 8823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.809033753728692e-05,
+      "loss": 0.5862,
+      "step": 8824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8089918541110797e-05,
+      "loss": 0.6241,
+      "step": 8825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808949950382743e-05,
+      "loss": 0.5949,
+      "step": 8826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808908042543895e-05,
+      "loss": 0.5898,
+      "step": 8827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8088661305947482e-05,
+      "loss": 0.6489,
+      "step": 8828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8088242145355152e-05,
+      "loss": 0.5664,
+      "step": 8829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.80878229436641e-05,
+      "loss": 0.5969,
+      "step": 8830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808740370087645e-05,
+      "loss": 0.6204,
+      "step": 8831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8086984416994335e-05,
+      "loss": 0.5645,
+      "step": 8832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808656509201988e-05,
+      "loss": 0.578,
+      "step": 8833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808614572595522e-05,
+      "loss": 0.6062,
+      "step": 8834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808572631880249e-05,
+      "loss": 0.6006,
+      "step": 8835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808530687056381e-05,
+      "loss": 0.5494,
+      "step": 8836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808488738124132e-05,
+      "loss": 0.629,
+      "step": 8837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8084467850837154e-05,
+      "loss": 0.6423,
+      "step": 8838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8084048279353434e-05,
+      "loss": 0.6575,
+      "step": 8839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8083628666792302e-05,
+      "loss": 0.6021,
+      "step": 8840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808320901315588e-05,
+      "loss": 0.5871,
+      "step": 8841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808278931844631e-05,
+      "loss": 0.6356,
+      "step": 8842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8082369582665718e-05,
+      "loss": 0.6513,
+      "step": 8843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808194980581624e-05,
+      "loss": 0.6338,
+      "step": 8844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8081529987900007e-05,
+      "loss": 0.6326,
+      "step": 8845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8081110128919156e-05,
+      "loss": 0.5653,
+      "step": 8846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8080690228875814e-05,
+      "loss": 0.5941,
+      "step": 8847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.808027028777212e-05,
+      "loss": 0.5536,
+      "step": 8848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8079850305610205e-05,
+      "loss": 0.5924,
+      "step": 8849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8079430282392206e-05,
+      "loss": 0.6084,
+      "step": 8850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8079010218120255e-05,
+      "loss": 0.6233,
+      "step": 8851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807859011279649e-05,
+      "loss": 0.5902,
+      "step": 8852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807816996642304e-05,
+      "loss": 0.5571,
+      "step": 8853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807774977900204e-05,
+      "loss": 0.6416,
+      "step": 8854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807732955053563e-05,
+      "loss": 0.6402,
+      "step": 8855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8076909281025946e-05,
+      "loss": 0.6103,
+      "step": 8856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807648897047512e-05,
+      "loss": 0.6038,
+      "step": 8857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8076068618885284e-05,
+      "loss": 0.6419,
+      "step": 8858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807564822625858e-05,
+      "loss": 0.6027,
+      "step": 8859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807522779259714e-05,
+      "loss": 0.6557,
+      "step": 8860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8074807317903108e-05,
+      "loss": 0.6199,
+      "step": 8861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807438680217861e-05,
+      "loss": 0.6195,
+      "step": 8862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8073966245425794e-05,
+      "loss": 0.667,
+      "step": 8863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8073545647646784e-05,
+      "loss": 0.5772,
+      "step": 8864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807312500884373e-05,
+      "loss": 0.6285,
+      "step": 8865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807270432901876e-05,
+      "loss": 0.5888,
+      "step": 8866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8072283608174017e-05,
+      "loss": 0.5963,
+      "step": 8867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8071862846311633e-05,
+      "loss": 0.6026,
+      "step": 8868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8071442043433755e-05,
+      "loss": 0.588,
+      "step": 8869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8071021199542515e-05,
+      "loss": 0.646,
+      "step": 8870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.807060031464005e-05,
+      "loss": 0.5422,
+      "step": 8871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8070179388728503e-05,
+      "loss": 0.5809,
+      "step": 8872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806975842181001e-05,
+      "loss": 0.5854,
+      "step": 8873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8069337413886713e-05,
+      "loss": 0.5908,
+      "step": 8874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806891636496075e-05,
+      "loss": 0.5773,
+      "step": 8875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8068495275034255e-05,
+      "loss": 0.637,
+      "step": 8876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8068074144109377e-05,
+      "loss": 0.6689,
+      "step": 8877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806765297218825e-05,
+      "loss": 0.5742,
+      "step": 8878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8067231759273016e-05,
+      "loss": 0.6435,
+      "step": 8879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8066810505365815e-05,
+      "loss": 0.5884,
+      "step": 8880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806638921046879e-05,
+      "loss": 0.6601,
+      "step": 8881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8065967874584075e-05,
+      "loss": 0.6311,
+      "step": 8882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8065546497713816e-05,
+      "loss": 0.6142,
+      "step": 8883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8065125079860153e-05,
+      "loss": 0.5747,
+      "step": 8884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8064703621025226e-05,
+      "loss": 0.574,
+      "step": 8885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806428212121118e-05,
+      "loss": 0.649,
+      "step": 8886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8063860580420158e-05,
+      "loss": 0.5762,
+      "step": 8887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8063438998654292e-05,
+      "loss": 0.602,
+      "step": 8888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8063017375915736e-05,
+      "loss": 0.6022,
+      "step": 8889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8062595712206622e-05,
+      "loss": 0.5816,
+      "step": 8890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8062174007529103e-05,
+      "loss": 0.5987,
+      "step": 8891
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8061752261885314e-05,
+      "loss": 0.6717,
+      "step": 8892
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8061330475277405e-05,
+      "loss": 0.5754,
+      "step": 8893
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8060908647707508e-05,
+      "loss": 0.5996,
+      "step": 8894
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.806048677917778e-05,
+      "loss": 0.6431,
+      "step": 8895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8060064869690355e-05,
+      "loss": 0.6007,
+      "step": 8896
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8059642919247382e-05,
+      "loss": 0.6242,
+      "step": 8897
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8059220927851e-05,
+      "loss": 0.6418,
+      "step": 8898
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805879889550336e-05,
+      "loss": 0.6076,
+      "step": 8899
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80583768222066e-05,
+      "loss": 0.5647,
+      "step": 8900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8057954707962866e-05,
+      "loss": 0.6425,
+      "step": 8901
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8057532552774307e-05,
+      "loss": 0.6327,
+      "step": 8902
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805711035664306e-05,
+      "loss": 0.6349,
+      "step": 8903
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805668811957128e-05,
+      "loss": 0.6059,
+      "step": 8904
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805626584156111e-05,
+      "loss": 0.5835,
+      "step": 8905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8055843522614692e-05,
+      "loss": 0.574,
+      "step": 8906
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8055421162734173e-05,
+      "loss": 0.5946,
+      "step": 8907
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80549987619217e-05,
+      "loss": 0.6668,
+      "step": 8908
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8054576320179422e-05,
+      "loss": 0.6394,
+      "step": 8909
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805415383750948e-05,
+      "loss": 0.6085,
+      "step": 8910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8053731313914023e-05,
+      "loss": 0.5825,
+      "step": 8911
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8053308749395203e-05,
+      "loss": 0.6371,
+      "step": 8912
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8052886143955157e-05,
+      "loss": 0.5658,
+      "step": 8913
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805246349759604e-05,
+      "loss": 0.6074,
+      "step": 8914
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805204081032e-05,
+      "loss": 0.6938,
+      "step": 8915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.805161808212918e-05,
+      "loss": 0.6409,
+      "step": 8916
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8051195313025732e-05,
+      "loss": 0.5652,
+      "step": 8917
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8050772503011803e-05,
+      "loss": 0.6611,
+      "step": 8918
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8050349652089537e-05,
+      "loss": 0.6144,
+      "step": 8919
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8049926760261092e-05,
+      "loss": 0.6397,
+      "step": 8920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804950382752861e-05,
+      "loss": 0.5608,
+      "step": 8921
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804908085389424e-05,
+      "loss": 0.5769,
+      "step": 8922
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8048657839360134e-05,
+      "loss": 0.634,
+      "step": 8923
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804823478392844e-05,
+      "loss": 0.5912,
+      "step": 8924
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804781168760131e-05,
+      "loss": 0.6594,
+      "step": 8925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8047388550380887e-05,
+      "loss": 0.6215,
+      "step": 8926
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8046965372269333e-05,
+      "loss": 0.6114,
+      "step": 8927
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8046542153268788e-05,
+      "loss": 0.6127,
+      "step": 8928
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8046118893381407e-05,
+      "loss": 0.6482,
+      "step": 8929
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8045695592609334e-05,
+      "loss": 0.6223,
+      "step": 8930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8045272250954732e-05,
+      "loss": 0.6681,
+      "step": 8931
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804484886841974e-05,
+      "loss": 0.6166,
+      "step": 8932
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804442544500652e-05,
+      "loss": 0.608,
+      "step": 8933
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8044001980717218e-05,
+      "loss": 0.5977,
+      "step": 8934
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804357847555398e-05,
+      "loss": 0.6438,
+      "step": 8935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8043154929518972e-05,
+      "loss": 0.6594,
+      "step": 8936
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804273134261433e-05,
+      "loss": 0.6288,
+      "step": 8937
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804230771484222e-05,
+      "loss": 0.659,
+      "step": 8938
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804188404620479e-05,
+      "loss": 0.6337,
+      "step": 8939
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.804146033670419e-05,
+      "loss": 0.5646,
+      "step": 8940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8041036586342577e-05,
+      "loss": 0.5909,
+      "step": 8941
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80406127951221e-05,
+      "loss": 0.5743,
+      "step": 8942
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8040188963044912e-05,
+      "loss": 0.6074,
+      "step": 8943
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8039765090113176e-05,
+      "loss": 0.6896,
+      "step": 8944
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8039341176329035e-05,
+      "loss": 0.6281,
+      "step": 8945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8038917221694646e-05,
+      "loss": 0.5995,
+      "step": 8946
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8038493226212167e-05,
+      "loss": 0.6172,
+      "step": 8947
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8038069189883748e-05,
+      "loss": 0.5935,
+      "step": 8948
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8037645112711545e-05,
+      "loss": 0.6277,
+      "step": 8949
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8037220994697714e-05,
+      "loss": 0.6133,
+      "step": 8950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.803679683584441e-05,
+      "loss": 0.6602,
+      "step": 8951
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8036372636153787e-05,
+      "loss": 0.5903,
+      "step": 8952
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8035948395628e-05,
+      "loss": 0.6477,
+      "step": 8953
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.803552411426921e-05,
+      "loss": 0.6626,
+      "step": 8954
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8035099792079564e-05,
+      "loss": 0.6106,
+      "step": 8955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8034675429061223e-05,
+      "loss": 0.63,
+      "step": 8956
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8034251025216344e-05,
+      "loss": 0.566,
+      "step": 8957
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8033826580547086e-05,
+      "loss": 0.6479,
+      "step": 8958
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8033402095055597e-05,
+      "loss": 0.5928,
+      "step": 8959
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8032977568744042e-05,
+      "loss": 0.5673,
+      "step": 8960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8032553001614576e-05,
+      "loss": 0.6186,
+      "step": 8961
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8032128393669356e-05,
+      "loss": 0.5892,
+      "step": 8962
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8031703744910536e-05,
+      "loss": 0.6571,
+      "step": 8963
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8031279055340278e-05,
+      "loss": 0.5246,
+      "step": 8964
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.803085432496074e-05,
+      "loss": 0.6232,
+      "step": 8965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.803042955377408e-05,
+      "loss": 0.5971,
+      "step": 8966
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8030004741782455e-05,
+      "loss": 0.5911,
+      "step": 8967
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8029579888988024e-05,
+      "loss": 0.5854,
+      "step": 8968
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8029154995392944e-05,
+      "loss": 0.606,
+      "step": 8969
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802873006099938e-05,
+      "loss": 0.6263,
+      "step": 8970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8028305085809488e-05,
+      "loss": 0.595,
+      "step": 8971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8027880069825425e-05,
+      "loss": 0.6109,
+      "step": 8972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802745501304935e-05,
+      "loss": 0.5647,
+      "step": 8973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8027029915483426e-05,
+      "loss": 0.6302,
+      "step": 8974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8026604777129814e-05,
+      "loss": 0.5949,
+      "step": 8975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802617959799067e-05,
+      "loss": 0.6871,
+      "step": 8976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802575437806816e-05,
+      "loss": 0.5697,
+      "step": 8977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8025329117364436e-05,
+      "loss": 0.6574,
+      "step": 8978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802490381588167e-05,
+      "loss": 0.6324,
+      "step": 8979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8024478473622013e-05,
+      "loss": 0.6628,
+      "step": 8980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8024053090587633e-05,
+      "loss": 0.5861,
+      "step": 8981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8023627666780692e-05,
+      "loss": 0.5673,
+      "step": 8982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8023202202203345e-05,
+      "loss": 0.6308,
+      "step": 8983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802277669685776e-05,
+      "loss": 0.6313,
+      "step": 8984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80223511507461e-05,
+      "loss": 0.5748,
+      "step": 8985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802192556387052e-05,
+      "loss": 0.6214,
+      "step": 8986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8021499936233185e-05,
+      "loss": 0.5556,
+      "step": 8987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802107426783626e-05,
+      "loss": 0.6536,
+      "step": 8988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.802064855868191e-05,
+      "loss": 0.5928,
+      "step": 8989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80202228087723e-05,
+      "loss": 0.6723,
+      "step": 8990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8019797018109583e-05,
+      "loss": 0.642,
+      "step": 8991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801937118669593e-05,
+      "loss": 0.557,
+      "step": 8992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8018945314533506e-05,
+      "loss": 0.6374,
+      "step": 8993
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801851940162447e-05,
+      "loss": 0.5864,
+      "step": 8994
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801809344797099e-05,
+      "loss": 0.6384,
+      "step": 8995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801766745357523e-05,
+      "loss": 0.5638,
+      "step": 8996
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8017241418439348e-05,
+      "loss": 0.6404,
+      "step": 8997
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8016815342565516e-05,
+      "loss": 0.5531,
+      "step": 8998
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80163892259559e-05,
+      "loss": 0.6076,
+      "step": 8999
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8015963068612664e-05,
+      "loss": 0.5851,
+      "step": 9000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8015536870537968e-05,
+      "loss": 0.6173,
+      "step": 9001
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8015110631733984e-05,
+      "loss": 0.6119,
+      "step": 9002
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8014684352202874e-05,
+      "loss": 0.6097,
+      "step": 9003
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8014258031946806e-05,
+      "loss": 0.5823,
+      "step": 9004
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801383167096795e-05,
+      "loss": 0.6444,
+      "step": 9005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8013405269268463e-05,
+      "loss": 0.5899,
+      "step": 9006
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801297882685052e-05,
+      "loss": 0.6113,
+      "step": 9007
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8012552343716283e-05,
+      "loss": 0.5821,
+      "step": 9008
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801212581986792e-05,
+      "loss": 0.6229,
+      "step": 9009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8011699255307602e-05,
+      "loss": 0.5339,
+      "step": 9010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8011272650037493e-05,
+      "loss": 0.5652,
+      "step": 9011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.801084600405976e-05,
+      "loss": 0.6729,
+      "step": 9012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8010419317376576e-05,
+      "loss": 0.6336,
+      "step": 9013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8009992589990103e-05,
+      "loss": 0.624,
+      "step": 9014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8009565821902512e-05,
+      "loss": 0.5806,
+      "step": 9015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800913901311597e-05,
+      "loss": 0.642,
+      "step": 9016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8008712163632646e-05,
+      "loss": 0.6214,
+      "step": 9017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8008285273454715e-05,
+      "loss": 0.6111,
+      "step": 9018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8007858342584338e-05,
+      "loss": 0.6385,
+      "step": 9019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8007431371023688e-05,
+      "loss": 0.636,
+      "step": 9020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8007004358774933e-05,
+      "loss": 0.5925,
+      "step": 9021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8006577305840248e-05,
+      "loss": 0.6566,
+      "step": 9022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8006150212221795e-05,
+      "loss": 0.6561,
+      "step": 9023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8005723077921748e-05,
+      "loss": 0.5873,
+      "step": 9024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800529590294228e-05,
+      "loss": 0.6245,
+      "step": 9025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8004868687285556e-05,
+      "loss": 0.6199,
+      "step": 9026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800444143095375e-05,
+      "loss": 0.6473,
+      "step": 9027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8004014133949035e-05,
+      "loss": 0.623,
+      "step": 9028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800358679627358e-05,
+      "loss": 0.7063,
+      "step": 9029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8003159417929558e-05,
+      "loss": 0.5729,
+      "step": 9030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8002731998919136e-05,
+      "loss": 0.6174,
+      "step": 9031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8002304539244488e-05,
+      "loss": 0.607,
+      "step": 9032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800187703890779e-05,
+      "loss": 0.6285,
+      "step": 9033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800144949791121e-05,
+      "loss": 0.6423,
+      "step": 9034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.800102191625692e-05,
+      "loss": 0.5908,
+      "step": 9035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.80005942939471e-05,
+      "loss": 0.6053,
+      "step": 9036
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8000166630983914e-05,
+      "loss": 0.566,
+      "step": 9037
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799973892736954e-05,
+      "loss": 0.6045,
+      "step": 9038
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7999311183106147e-05,
+      "loss": 0.5757,
+      "step": 9039
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799888339819591e-05,
+      "loss": 0.6125,
+      "step": 9040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7998455572641006e-05,
+      "loss": 0.6148,
+      "step": 9041
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7998027706443605e-05,
+      "loss": 0.5967,
+      "step": 9042
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7997599799605888e-05,
+      "loss": 0.5664,
+      "step": 9043
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799717185213002e-05,
+      "loss": 0.6094,
+      "step": 9044
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799674386401818e-05,
+      "loss": 0.6308,
+      "step": 9045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7996315835272545e-05,
+      "loss": 0.5893,
+      "step": 9046
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7995887765895285e-05,
+      "loss": 0.6406,
+      "step": 9047
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7995459655888577e-05,
+      "loss": 0.6371,
+      "step": 9048
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79950315052546e-05,
+      "loss": 0.6056,
+      "step": 9049
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7994603313995524e-05,
+      "loss": 0.646,
+      "step": 9050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7994175082113528e-05,
+      "loss": 0.6867,
+      "step": 9051
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7993746809610787e-05,
+      "loss": 0.5653,
+      "step": 9052
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799331849648948e-05,
+      "loss": 0.6345,
+      "step": 9053
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7992890142751775e-05,
+      "loss": 0.5859,
+      "step": 9054
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7992461748399858e-05,
+      "loss": 0.6557,
+      "step": 9055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79920333134359e-05,
+      "loss": 0.5644,
+      "step": 9056
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799160483786208e-05,
+      "loss": 0.6265,
+      "step": 9057
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7991176321680578e-05,
+      "loss": 0.5841,
+      "step": 9058
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799074776489357e-05,
+      "loss": 0.6156,
+      "step": 9059
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.799031916750323e-05,
+      "loss": 0.6061,
+      "step": 9060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798989052951174e-05,
+      "loss": 0.5967,
+      "step": 9061
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7989461850921273e-05,
+      "loss": 0.6129,
+      "step": 9062
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7989033131734013e-05,
+      "loss": 0.6543,
+      "step": 9063
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7988604371952134e-05,
+      "loss": 0.6123,
+      "step": 9064
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798817557157782e-05,
+      "loss": 0.6115,
+      "step": 9065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7987746730613246e-05,
+      "loss": 0.6194,
+      "step": 9066
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7987317849060584e-05,
+      "loss": 0.5655,
+      "step": 9067
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7986888926922028e-05,
+      "loss": 0.586,
+      "step": 9068
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798645996419975e-05,
+      "loss": 0.6384,
+      "step": 9069
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798603096089593e-05,
+      "loss": 0.5895,
+      "step": 9070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7985601917012748e-05,
+      "loss": 0.6032,
+      "step": 9071
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7985172832552382e-05,
+      "loss": 0.6088,
+      "step": 9072
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7984743707517017e-05,
+      "loss": 0.6129,
+      "step": 9073
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7984314541908828e-05,
+      "loss": 0.5805,
+      "step": 9074
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7983885335729997e-05,
+      "loss": 0.6856,
+      "step": 9075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798345608898271e-05,
+      "loss": 0.6508,
+      "step": 9076
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7983026801669144e-05,
+      "loss": 0.6134,
+      "step": 9077
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798259747379148e-05,
+      "loss": 0.6325,
+      "step": 9078
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79821681053519e-05,
+      "loss": 0.6084,
+      "step": 9079
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7981738696352588e-05,
+      "loss": 0.6215,
+      "step": 9080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7981309246795722e-05,
+      "loss": 0.5407,
+      "step": 9081
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798087975668349e-05,
+      "loss": 0.5551,
+      "step": 9082
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7980450226018065e-05,
+      "loss": 0.5974,
+      "step": 9083
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.798002065480164e-05,
+      "loss": 0.669,
+      "step": 9084
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7979591043036387e-05,
+      "loss": 0.6423,
+      "step": 9085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7979161390724503e-05,
+      "loss": 0.6367,
+      "step": 9086
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7978731697868158e-05,
+      "loss": 0.6175,
+      "step": 9087
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7978301964469542e-05,
+      "loss": 0.5977,
+      "step": 9088
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7977872190530835e-05,
+      "loss": 0.5352,
+      "step": 9089
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797744237605423e-05,
+      "loss": 0.5766,
+      "step": 9090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7977012521041896e-05,
+      "loss": 0.6202,
+      "step": 9091
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7976582625496028e-05,
+      "loss": 0.5789,
+      "step": 9092
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797615268941881e-05,
+      "loss": 0.5592,
+      "step": 9093
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7975722712812416e-05,
+      "loss": 0.6285,
+      "step": 9094
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797529269567905e-05,
+      "loss": 0.6467,
+      "step": 9095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7974862638020882e-05,
+      "loss": 0.6393,
+      "step": 9096
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7974432539840097e-05,
+      "loss": 0.6269,
+      "step": 9097
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797400240113889e-05,
+      "loss": 0.6283,
+      "step": 9098
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797357222191944e-05,
+      "loss": 0.5552,
+      "step": 9099
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7973142002183933e-05,
+      "loss": 0.6283,
+      "step": 9100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797271174193456e-05,
+      "loss": 0.618,
+      "step": 9101
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79722814411735e-05,
+      "loss": 0.6449,
+      "step": 9102
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7971851099902948e-05,
+      "loss": 0.6095,
+      "step": 9103
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797142071812508e-05,
+      "loss": 0.5944,
+      "step": 9104
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7970990295842093e-05,
+      "loss": 0.6313,
+      "step": 9105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7970559833056167e-05,
+      "loss": 0.6263,
+      "step": 9106
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.797012932976949e-05,
+      "loss": 0.5507,
+      "step": 9107
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796969878598426e-05,
+      "loss": 0.569,
+      "step": 9108
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796926820170265e-05,
+      "loss": 0.6545,
+      "step": 9109
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7968837576926857e-05,
+      "loss": 0.6824,
+      "step": 9110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7968406911659063e-05,
+      "loss": 0.6616,
+      "step": 9111
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7967976205901464e-05,
+      "loss": 0.6344,
+      "step": 9112
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7967545459656242e-05,
+      "loss": 0.6362,
+      "step": 9113
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796711467292559e-05,
+      "loss": 0.5913,
+      "step": 9114
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7966683845711693e-05,
+      "loss": 0.5473,
+      "step": 9115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7966252978016746e-05,
+      "loss": 0.5458,
+      "step": 9116
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796582206984293e-05,
+      "loss": 0.6148,
+      "step": 9117
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796539112119244e-05,
+      "loss": 0.6011,
+      "step": 9118
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796496013206747e-05,
+      "loss": 0.6484,
+      "step": 9119
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7964529102470197e-05,
+      "loss": 0.6232,
+      "step": 9120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7964098032402823e-05,
+      "loss": 0.6282,
+      "step": 9121
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7963666921867535e-05,
+      "loss": 0.6067,
+      "step": 9122
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796323577086652e-05,
+      "loss": 0.5906,
+      "step": 9123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796280457940198e-05,
+      "loss": 0.6079,
+      "step": 9124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796237334747609e-05,
+      "loss": 0.6054,
+      "step": 9125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796194207509105e-05,
+      "loss": 0.6243,
+      "step": 9126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7961510762249054e-05,
+      "loss": 0.657,
+      "step": 9127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7961079408952287e-05,
+      "loss": 0.657,
+      "step": 9128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7960648015202944e-05,
+      "loss": 0.6098,
+      "step": 9129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.796021658100322e-05,
+      "loss": 0.6106,
+      "step": 9130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79597851063553e-05,
+      "loss": 0.6351,
+      "step": 9131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7959353591261385e-05,
+      "loss": 0.5548,
+      "step": 9132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.795892203572366e-05,
+      "loss": 0.5905,
+      "step": 9133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7958490439744324e-05,
+      "loss": 0.603,
+      "step": 9134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7958058803325568e-05,
+      "loss": 0.6023,
+      "step": 9135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7957627126469584e-05,
+      "loss": 0.5723,
+      "step": 9136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7957195409178562e-05,
+      "loss": 0.6237,
+      "step": 9137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7956763651454703e-05,
+      "loss": 0.582,
+      "step": 9138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7956331853300196e-05,
+      "loss": 0.6277,
+      "step": 9139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.795590001471724e-05,
+      "loss": 0.6288,
+      "step": 9140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7955468135708024e-05,
+      "loss": 0.5967,
+      "step": 9141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7955036216274747e-05,
+      "loss": 0.6109,
+      "step": 9142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7954604256419597e-05,
+      "loss": 0.6188,
+      "step": 9143
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7954172256144774e-05,
+      "loss": 0.6385,
+      "step": 9144
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7953740215452473e-05,
+      "loss": 0.6215,
+      "step": 9145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.795330813434489e-05,
+      "loss": 0.5743,
+      "step": 9146
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7952876012824218e-05,
+      "loss": 0.6074,
+      "step": 9147
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7952443850892655e-05,
+      "loss": 0.6449,
+      "step": 9148
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7952011648552393e-05,
+      "loss": 0.5923,
+      "step": 9149
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7951579405805633e-05,
+      "loss": 0.6461,
+      "step": 9150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.795114712265457e-05,
+      "loss": 0.6064,
+      "step": 9151
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7950714799101398e-05,
+      "loss": 0.5721,
+      "step": 9152
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7950282435148317e-05,
+      "loss": 0.586,
+      "step": 9153
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794985003079752e-05,
+      "loss": 0.6135,
+      "step": 9154
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794941758605121e-05,
+      "loss": 0.657,
+      "step": 9155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794898510091158e-05,
+      "loss": 0.6008,
+      "step": 9156
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794855257538083e-05,
+      "loss": 0.5602,
+      "step": 9157
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7948120009461155e-05,
+      "loss": 0.6351,
+      "step": 9158
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7947687403154756e-05,
+      "loss": 0.6179,
+      "step": 9159
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7947254756463825e-05,
+      "loss": 0.6395,
+      "step": 9160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794682206939057e-05,
+      "loss": 0.5059,
+      "step": 9161
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7946389341937184e-05,
+      "loss": 0.5653,
+      "step": 9162
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7945956574105864e-05,
+      "loss": 0.6118,
+      "step": 9163
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7945523765898815e-05,
+      "loss": 0.6459,
+      "step": 9164
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794509091731823e-05,
+      "loss": 0.6014,
+      "step": 9165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7944658028366313e-05,
+      "loss": 0.6333,
+      "step": 9166
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7944225099045262e-05,
+      "loss": 0.6281,
+      "step": 9167
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7943792129357277e-05,
+      "loss": 0.6613,
+      "step": 9168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7943359119304556e-05,
+      "loss": 0.5732,
+      "step": 9169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79429260688893e-05,
+      "loss": 0.6474,
+      "step": 9170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7942492978113716e-05,
+      "loss": 0.5981,
+      "step": 9171
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7942059846979994e-05,
+      "loss": 0.5983,
+      "step": 9172
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.794162667549034e-05,
+      "loss": 0.6501,
+      "step": 9173
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7941193463646954e-05,
+      "loss": 0.6204,
+      "step": 9174
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7940760211452045e-05,
+      "loss": 0.6387,
+      "step": 9175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.79403269189078e-05,
+      "loss": 0.6148,
+      "step": 9176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7939893586016436e-05,
+      "loss": 0.6143,
+      "step": 9177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7939460212780142e-05,
+      "loss": 0.6182,
+      "step": 9178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7939026799201125e-05,
+      "loss": 0.6284,
+      "step": 9179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.793859334528159e-05,
+      "loss": 0.5725,
+      "step": 9180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7938159851023738e-05,
+      "loss": 0.5308,
+      "step": 9181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7937726316429768e-05,
+      "loss": 0.5865,
+      "step": 9182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7937292741501888e-05,
+      "loss": 0.6178,
+      "step": 9183
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.79368591262423e-05,
+      "loss": 0.5998,
+      "step": 9184
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7936425470653205e-05,
+      "loss": 0.5702,
+      "step": 9185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7935991774736806e-05,
+      "loss": 0.5797,
+      "step": 9186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7935558038495313e-05,
+      "loss": 0.6303,
+      "step": 9187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.793512426193092e-05,
+      "loss": 0.6339,
+      "step": 9188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7934690445045844e-05,
+      "loss": 0.7195,
+      "step": 9189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7934256587842277e-05,
+      "loss": 0.5889,
+      "step": 9190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.793382269032243e-05,
+      "loss": 0.586,
+      "step": 9191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7933388752488506e-05,
+      "loss": 0.6506,
+      "step": 9192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.793295477434271e-05,
+      "loss": 0.5837,
+      "step": 9193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7932520755887248e-05,
+      "loss": 0.6275,
+      "step": 9194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7932086697124327e-05,
+      "loss": 0.6591,
+      "step": 9195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7931652598056146e-05,
+      "loss": 0.6496,
+      "step": 9196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7931218458684917e-05,
+      "loss": 0.6052,
+      "step": 9197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7930784279012847e-05,
+      "loss": 0.6142,
+      "step": 9198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7930350059042137e-05,
+      "loss": 0.6546,
+      "step": 9199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7929915798774994e-05,
+      "loss": 0.5635,
+      "step": 9200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792948149821363e-05,
+      "loss": 0.624,
+      "step": 9201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792904715736025e-05,
+      "loss": 0.6221,
+      "step": 9202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7928612776217055e-05,
+      "loss": 0.5404,
+      "step": 9203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7928178354786257e-05,
+      "loss": 0.6229,
+      "step": 9204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792774389307006e-05,
+      "loss": 0.6256,
+      "step": 9205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7927309391070678e-05,
+      "loss": 0.5776,
+      "step": 9206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7926874848790315e-05,
+      "loss": 0.6598,
+      "step": 9207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792644026623118e-05,
+      "loss": 0.6296,
+      "step": 9208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7926005643395484e-05,
+      "loss": 0.6441,
+      "step": 9209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7925570980285427e-05,
+      "loss": 0.6683,
+      "step": 9210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7925136276903226e-05,
+      "loss": 0.6068,
+      "step": 9211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7924701533251083e-05,
+      "loss": 0.6119,
+      "step": 9212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7924266749331214e-05,
+      "loss": 0.5887,
+      "step": 9213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7923831925145822e-05,
+      "loss": 0.5421,
+      "step": 9214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792339706069712e-05,
+      "loss": 0.6461,
+      "step": 9215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7922962155987318e-05,
+      "loss": 0.5224,
+      "step": 9216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7922527211018625e-05,
+      "loss": 0.5492,
+      "step": 9217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.792209222579325e-05,
+      "loss": 0.629,
+      "step": 9218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7921657200313406e-05,
+      "loss": 0.6845,
+      "step": 9219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.79212221345813e-05,
+      "loss": 0.6344,
+      "step": 9220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7920787028599147e-05,
+      "loss": 0.561,
+      "step": 9221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7920351882369154e-05,
+      "loss": 0.6127,
+      "step": 9222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7919916695893533e-05,
+      "loss": 0.5739,
+      "step": 9223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7919481469174495e-05,
+      "loss": 0.6297,
+      "step": 9224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7919046202214254e-05,
+      "loss": 0.6034,
+      "step": 9225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7918610895015018e-05,
+      "loss": 0.6227,
+      "step": 9226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7918175547579003e-05,
+      "loss": 0.633,
+      "step": 9227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.791774015990842e-05,
+      "loss": 0.6468,
+      "step": 9228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.791730473200548e-05,
+      "loss": 0.5989,
+      "step": 9229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7916869263872393e-05,
+      "loss": 0.6163,
+      "step": 9230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7916433755511376e-05,
+      "loss": 0.6458,
+      "step": 9231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7915998206924644e-05,
+      "loss": 0.602,
+      "step": 9232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.79155626181144e-05,
+      "loss": 0.6023,
+      "step": 9233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.791512698908287e-05,
+      "loss": 0.6234,
+      "step": 9234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.791469131983226e-05,
+      "loss": 0.5726,
+      "step": 9235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7914255610364785e-05,
+      "loss": 0.6225,
+      "step": 9236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7913819860682663e-05,
+      "loss": 0.5911,
+      "step": 9237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7913384070788102e-05,
+      "loss": 0.5651,
+      "step": 9238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7912948240683318e-05,
+      "loss": 0.6033,
+      "step": 9239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7912512370370528e-05,
+      "loss": 0.6084,
+      "step": 9240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7912076459851943e-05,
+      "loss": 0.633,
+      "step": 9241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7911640509129782e-05,
+      "loss": 0.5432,
+      "step": 9242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7911204518206256e-05,
+      "loss": 0.6162,
+      "step": 9243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7910768487083586e-05,
+      "loss": 0.5935,
+      "step": 9244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7910332415763985e-05,
+      "loss": 0.6107,
+      "step": 9245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790989630424967e-05,
+      "loss": 0.6184,
+      "step": 9246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790946015254285e-05,
+      "loss": 0.6447,
+      "step": 9247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7909023960645748e-05,
+      "loss": 0.6537,
+      "step": 9248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790858772856058e-05,
+      "loss": 0.6211,
+      "step": 9249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7908151456289562e-05,
+      "loss": 0.5825,
+      "step": 9250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790771514383491e-05,
+      "loss": 0.6277,
+      "step": 9251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790727879119884e-05,
+      "loss": 0.6257,
+      "step": 9252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7906842398383574e-05,
+      "loss": 0.6481,
+      "step": 9253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7906405965391326e-05,
+      "loss": 0.5863,
+      "step": 9254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790596949222431e-05,
+      "loss": 0.6401,
+      "step": 9255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790553297888475e-05,
+      "loss": 0.6912,
+      "step": 9256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7905096425374862e-05,
+      "loss": 0.634,
+      "step": 9257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7904659831696863e-05,
+      "loss": 0.5768,
+      "step": 9258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7904223197852975e-05,
+      "loss": 0.6134,
+      "step": 9259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7903786523845413e-05,
+      "loss": 0.5823,
+      "step": 9260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7903349809676395e-05,
+      "loss": 0.6223,
+      "step": 9261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7902913055348145e-05,
+      "loss": 0.5988,
+      "step": 9262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7902476260862872e-05,
+      "loss": 0.6264,
+      "step": 9263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790203942622281e-05,
+      "loss": 0.6096,
+      "step": 9264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790160255143017e-05,
+      "loss": 0.5619,
+      "step": 9265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7901165636487175e-05,
+      "loss": 0.6307,
+      "step": 9266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7900728681396037e-05,
+      "loss": 0.6444,
+      "step": 9267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.790029168615899e-05,
+      "loss": 0.6791,
+      "step": 9268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7899854650778245e-05,
+      "loss": 0.5719,
+      "step": 9269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7899417575256022e-05,
+      "loss": 0.5897,
+      "step": 9270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7898980459594546e-05,
+      "loss": 0.6097,
+      "step": 9271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789854330379604e-05,
+      "loss": 0.6012,
+      "step": 9272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789810610786272e-05,
+      "loss": 0.6057,
+      "step": 9273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789766887179681e-05,
+      "loss": 0.548,
+      "step": 9274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7897231595600534e-05,
+      "loss": 0.6254,
+      "step": 9275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7896794279276105e-05,
+      "loss": 0.6331,
+      "step": 9276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7896356922825757e-05,
+      "loss": 0.5682,
+      "step": 9277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7895919526251705e-05,
+      "loss": 0.6153,
+      "step": 9278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7895482089556176e-05,
+      "loss": 0.5502,
+      "step": 9279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7895044612741385e-05,
+      "loss": 0.6268,
+      "step": 9280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7894607095809562e-05,
+      "loss": 0.6101,
+      "step": 9281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7894169538762934e-05,
+      "loss": 0.6202,
+      "step": 9282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7893731941603714e-05,
+      "loss": 0.5935,
+      "step": 9283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789329430433413e-05,
+      "loss": 0.5807,
+      "step": 9284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789285662695641e-05,
+      "loss": 0.6587,
+      "step": 9285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7892418909472768e-05,
+      "loss": 0.5956,
+      "step": 9286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7891981151885437e-05,
+      "loss": 0.6705,
+      "step": 9287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7891543354196636e-05,
+      "loss": 0.6346,
+      "step": 9288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7891105516408594e-05,
+      "loss": 0.6349,
+      "step": 9289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7890667638523537e-05,
+      "loss": 0.6421,
+      "step": 9290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.789022972054368e-05,
+      "loss": 0.619,
+      "step": 9291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.788979176247126e-05,
+      "loss": 0.6148,
+      "step": 9292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7889353764308498e-05,
+      "loss": 0.6264,
+      "step": 9293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7888915726057616e-05,
+      "loss": 0.5694,
+      "step": 9294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7888477647720847e-05,
+      "loss": 0.6239,
+      "step": 9295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7888039529300408e-05,
+      "loss": 0.5638,
+      "step": 9296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7887601370798533e-05,
+      "loss": 0.6924,
+      "step": 9297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7887163172217444e-05,
+      "loss": 0.5809,
+      "step": 9298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7886724933559367e-05,
+      "loss": 0.6377,
+      "step": 9299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7886286654826535e-05,
+      "loss": 0.6239,
+      "step": 9300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.788584833602117e-05,
+      "loss": 0.5627,
+      "step": 9301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7885409977145497e-05,
+      "loss": 0.5771,
+      "step": 9302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7884971578201752e-05,
+      "loss": 0.6164,
+      "step": 9303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7884533139192153e-05,
+      "loss": 0.6371,
+      "step": 9304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7884094660118933e-05,
+      "loss": 0.6939,
+      "step": 9305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.788365614098432e-05,
+      "loss": 0.6008,
+      "step": 9306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.788321758179054e-05,
+      "loss": 0.6079,
+      "step": 9307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7882778982539822e-05,
+      "loss": 0.6298,
+      "step": 9308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.78823403432344e-05,
+      "loss": 0.6655,
+      "step": 9309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7881901663876494e-05,
+      "loss": 0.6078,
+      "step": 9310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7881462944468337e-05,
+      "loss": 0.6182,
+      "step": 9311
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7881024185012158e-05,
+      "loss": 0.6398,
+      "step": 9312
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.788058538551019e-05,
+      "loss": 0.574,
+      "step": 9313
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7880146545964662e-05,
+      "loss": 0.5974,
+      "step": 9314
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7879707666377798e-05,
+      "loss": 0.5664,
+      "step": 9315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787926874675183e-05,
+      "loss": 0.6215,
+      "step": 9316
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7878829787088993e-05,
+      "loss": 0.5853,
+      "step": 9317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787839078739151e-05,
+      "loss": 0.5756,
+      "step": 9318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787795174766162e-05,
+      "loss": 0.5872,
+      "step": 9319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787751266790155e-05,
+      "loss": 0.6149,
+      "step": 9320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787707354811353e-05,
+      "loss": 0.6281,
+      "step": 9321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7876634388299795e-05,
+      "loss": 0.5621,
+      "step": 9322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.787619518846257e-05,
+      "loss": 0.5597,
+      "step": 9323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7875755948604093e-05,
+      "loss": 0.6032,
+      "step": 9324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7875316668726595e-05,
+      "loss": 0.6276,
+      "step": 9325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7874877348832302e-05,
+      "loss": 0.6084,
+      "step": 9326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7874437988923455e-05,
+      "loss": 0.5719,
+      "step": 9327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7873998589002283e-05,
+      "loss": 0.6283,
+      "step": 9328
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7873559149071015e-05,
+      "loss": 0.5943,
+      "step": 9329
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.7873119669131887e-05,
+      "loss": 0.6168,
+      "step": 9330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7872680149187135e-05,
+      "loss": 0.6082,
+      "step": 9331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7872240589238986e-05,
+      "loss": 0.6835,
+      "step": 9332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7871800989289678e-05,
+      "loss": 0.6413,
+      "step": 9333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7871361349341446e-05,
+      "loss": 0.6603,
+      "step": 9334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.787092166939652e-05,
+      "loss": 0.5854,
+      "step": 9335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7870481949457136e-05,
+      "loss": 0.6195,
+      "step": 9336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.787004218952553e-05,
+      "loss": 0.6312,
+      "step": 9337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7869602389603932e-05,
+      "loss": 0.5941,
+      "step": 9338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7869162549694582e-05,
+      "loss": 0.6798,
+      "step": 9339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786872266979971e-05,
+      "loss": 0.6196,
+      "step": 9340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786828274992156e-05,
+      "loss": 0.6315,
+      "step": 9341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786784279006235e-05,
+      "loss": 0.5968,
+      "step": 9342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7867402790224335e-05,
+      "loss": 0.5951,
+      "step": 9343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786696275040974e-05,
+      "loss": 0.6419,
+      "step": 9344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7866522670620802e-05,
+      "loss": 0.5972,
+      "step": 9345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786608255085976e-05,
+      "loss": 0.5825,
+      "step": 9346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786564239112885e-05,
+      "loss": 0.5866,
+      "step": 9347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7865202191430305e-05,
+      "loss": 0.6254,
+      "step": 9348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7864761951766368e-05,
+      "loss": 0.5888,
+      "step": 9349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786432167213927e-05,
+      "loss": 0.6145,
+      "step": 9350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786388135255125e-05,
+      "loss": 0.6915,
+      "step": 9351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7863440993004543e-05,
+      "loss": 0.6031,
+      "step": 9352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7863000593501395e-05,
+      "loss": 0.6519,
+      "step": 9353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7862560154044036e-05,
+      "loss": 0.601,
+      "step": 9354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7862119674634703e-05,
+      "loss": 0.5982,
+      "step": 9355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786167915527564e-05,
+      "loss": 0.6168,
+      "step": 9356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7861238595969083e-05,
+      "loss": 0.6253,
+      "step": 9357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786079799671727e-05,
+      "loss": 0.6066,
+      "step": 9358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.786035735752244e-05,
+      "loss": 0.6219,
+      "step": 9359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7859916678386836e-05,
+      "loss": 0.607,
+      "step": 9360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785947595931269e-05,
+      "loss": 0.627,
+      "step": 9361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7859035200302244e-05,
+      "loss": 0.6357,
+      "step": 9362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7858594401357742e-05,
+      "loss": 0.614,
+      "step": 9363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7858153562481417e-05,
+      "loss": 0.5784,
+      "step": 9364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785771268367551e-05,
+      "loss": 0.6692,
+      "step": 9365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785727176494227e-05,
+      "loss": 0.6496,
+      "step": 9366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7856830806283927e-05,
+      "loss": 0.6249,
+      "step": 9367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7856389807702728e-05,
+      "loss": 0.581,
+      "step": 9368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785594876920091e-05,
+      "loss": 0.6261,
+      "step": 9369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7855507690780716e-05,
+      "loss": 0.5762,
+      "step": 9370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7855066572444386e-05,
+      "loss": 0.5931,
+      "step": 9371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7854625414194163e-05,
+      "loss": 0.6468,
+      "step": 9372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785418421603229e-05,
+      "loss": 0.6267,
+      "step": 9373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7853742977961005e-05,
+      "loss": 0.6253,
+      "step": 9374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785330169998255e-05,
+      "loss": 0.5951,
+      "step": 9375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785286038209917e-05,
+      "loss": 0.6828,
+      "step": 9376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785241902431311e-05,
+      "loss": 0.6096,
+      "step": 9377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7851977626626604e-05,
+      "loss": 0.566,
+      "step": 9378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7851536189041903e-05,
+      "loss": 0.5817,
+      "step": 9379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785109471156125e-05,
+      "loss": 0.608,
+      "step": 9380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.785065319418688e-05,
+      "loss": 0.5849,
+      "step": 9381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7850211636921045e-05,
+      "loss": 0.6293,
+      "step": 9382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7849770039765982e-05,
+      "loss": 0.654,
+      "step": 9383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784932840272394e-05,
+      "loss": 0.5889,
+      "step": 9384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7848886725797165e-05,
+      "loss": 0.5798,
+      "step": 9385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7848445008987897e-05,
+      "loss": 0.5858,
+      "step": 9386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784800325229838e-05,
+      "loss": 0.6098,
+      "step": 9387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784756145573086e-05,
+      "loss": 0.5934,
+      "step": 9388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7847119619287583e-05,
+      "loss": 0.6259,
+      "step": 9389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7846677742970793e-05,
+      "loss": 0.6381,
+      "step": 9390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7846235826782735e-05,
+      "loss": 0.5994,
+      "step": 9391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7845793870725653e-05,
+      "loss": 0.5703,
+      "step": 9392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7845351874801794e-05,
+      "loss": 0.5903,
+      "step": 9393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784490983901341e-05,
+      "loss": 0.603,
+      "step": 9394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784446776336274e-05,
+      "loss": 0.5908,
+      "step": 9395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7844025647852027e-05,
+      "loss": 0.6368,
+      "step": 9396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7843583492483526e-05,
+      "loss": 0.6769,
+      "step": 9397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784314129725948e-05,
+      "loss": 0.6175,
+      "step": 9398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7842699062182137e-05,
+      "loss": 0.5843,
+      "step": 9399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7842256787253746e-05,
+      "loss": 0.6426,
+      "step": 9400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7841814472476544e-05,
+      "loss": 0.6186,
+      "step": 9401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784137211785279e-05,
+      "loss": 0.5755,
+      "step": 9402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7840929723384728e-05,
+      "loss": 0.6383,
+      "step": 9403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7840487289074605e-05,
+      "loss": 0.6613,
+      "step": 9404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.784004481492467e-05,
+      "loss": 0.5808,
+      "step": 9405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7839602300937172e-05,
+      "loss": 0.613,
+      "step": 9406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783915974711436e-05,
+      "loss": 0.5908,
+      "step": 9407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783871715345848e-05,
+      "loss": 0.5952,
+      "step": 9408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7838274519971782e-05,
+      "loss": 0.5826,
+      "step": 9409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7837831846656514e-05,
+      "loss": 0.5699,
+      "step": 9410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7837389133514933e-05,
+      "loss": 0.5923,
+      "step": 9411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783694638054928e-05,
+      "loss": 0.6071,
+      "step": 9412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7836503587761803e-05,
+      "loss": 0.6436,
+      "step": 9413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783606075515476e-05,
+      "loss": 0.6795,
+      "step": 9414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7835617882730398e-05,
+      "loss": 0.652,
+      "step": 9415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783517497049096e-05,
+      "loss": 0.5799,
+      "step": 9416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783473201843871e-05,
+      "loss": 0.5768,
+      "step": 9417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7834289026575892e-05,
+      "loss": 0.603,
+      "step": 9418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7833845994904756e-05,
+      "loss": 0.643,
+      "step": 9419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7833402923427553e-05,
+      "loss": 0.637,
+      "step": 9420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7832959812146538e-05,
+      "loss": 0.6497,
+      "step": 9421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783251666106396e-05,
+      "loss": 0.6575,
+      "step": 9422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7832073470182068e-05,
+      "loss": 0.5781,
+      "step": 9423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783163023950312e-05,
+      "loss": 0.5707,
+      "step": 9424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7831186969029363e-05,
+      "loss": 0.5692,
+      "step": 9425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7830743658763052e-05,
+      "loss": 0.5803,
+      "step": 9426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.783030030870644e-05,
+      "loss": 0.6652,
+      "step": 9427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7829856918861778e-05,
+      "loss": 0.6264,
+      "step": 9428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782941348923132e-05,
+      "loss": 0.6182,
+      "step": 9429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782897001981732e-05,
+      "loss": 0.6197,
+      "step": 9430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782852651062203e-05,
+      "loss": 0.6739,
+      "step": 9431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7828082961647703e-05,
+      "loss": 0.5615,
+      "step": 9432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7827639372896597e-05,
+      "loss": 0.5742,
+      "step": 9433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782719574437096e-05,
+      "loss": 0.6581,
+      "step": 9434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782675207607305e-05,
+      "loss": 0.6228,
+      "step": 9435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7826308368005123e-05,
+      "loss": 0.6022,
+      "step": 9436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782586462016943e-05,
+      "loss": 0.6099,
+      "step": 9437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782542083256822e-05,
+      "loss": 0.5673,
+      "step": 9438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7824977005203764e-05,
+      "loss": 0.676,
+      "step": 9439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7824533138078302e-05,
+      "loss": 0.5877,
+      "step": 9440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7824089231194103e-05,
+      "loss": 0.623,
+      "step": 9441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7823645284553407e-05,
+      "loss": 0.579,
+      "step": 9442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7823201298158483e-05,
+      "loss": 0.5935,
+      "step": 9443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.782275727201158e-05,
+      "loss": 0.6254,
+      "step": 9444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7822313206114957e-05,
+      "loss": 0.6397,
+      "step": 9445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7821869100470872e-05,
+      "loss": 0.6146,
+      "step": 9446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7821424955081575e-05,
+      "loss": 0.6205,
+      "step": 9447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7820980769949334e-05,
+      "loss": 0.614,
+      "step": 9448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7820536545076393e-05,
+      "loss": 0.606,
+      "step": 9449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7820092280465017e-05,
+      "loss": 0.6579,
+      "step": 9450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781964797611746e-05,
+      "loss": 0.633,
+      "step": 9451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7819203632035984e-05,
+      "loss": 0.6053,
+      "step": 9452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7818759248222847e-05,
+      "loss": 0.6168,
+      "step": 9453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.78183148246803e-05,
+      "loss": 0.5989,
+      "step": 9454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7817870361410607e-05,
+      "loss": 0.6198,
+      "step": 9455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7817425858416025e-05,
+      "loss": 0.5843,
+      "step": 9456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7816981315698812e-05,
+      "loss": 0.5605,
+      "step": 9457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781653673326123e-05,
+      "loss": 0.5844,
+      "step": 9458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7816092111105534e-05,
+      "loss": 0.6711,
+      "step": 9459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7815647449233988e-05,
+      "loss": 0.594,
+      "step": 9460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7815202747648844e-05,
+      "loss": 0.6374,
+      "step": 9461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7814758006352367e-05,
+      "loss": 0.6009,
+      "step": 9462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7814313225346816e-05,
+      "loss": 0.597,
+      "step": 9463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781386840463445e-05,
+      "loss": 0.5801,
+      "step": 9464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7813423544217534e-05,
+      "loss": 0.5582,
+      "step": 9465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781297864409832e-05,
+      "loss": 0.6362,
+      "step": 9466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781253370427908e-05,
+      "loss": 0.6695,
+      "step": 9467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7812088724762063e-05,
+      "loss": 0.6277,
+      "step": 9468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7811643705549536e-05,
+      "loss": 0.6008,
+      "step": 9469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.781119864664376e-05,
+      "loss": 0.6484,
+      "step": 9470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7810753548046994e-05,
+      "loss": 0.6225,
+      "step": 9471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7810308409761505e-05,
+      "loss": 0.6105,
+      "step": 9472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.780986323178955e-05,
+      "loss": 0.5713,
+      "step": 9473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7809418014133394e-05,
+      "loss": 0.6149,
+      "step": 9474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7808972756795297e-05,
+      "loss": 0.6381,
+      "step": 9475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.780852745977752e-05,
+      "loss": 0.5368,
+      "step": 9476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.7808082123082334e-05,
+      "loss": 0.577,
+      "step": 9477
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7807636746711992e-05,
+      "loss": 0.614,
+      "step": 9478
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7807191330668764e-05,
+      "loss": 0.6549,
+      "step": 9479
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7806745874954904e-05,
+      "loss": 0.62,
+      "step": 9480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7806300379572688e-05,
+      "loss": 0.5783,
+      "step": 9481
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7805854844524374e-05,
+      "loss": 0.5769,
+      "step": 9482
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780540926981222e-05,
+      "loss": 0.5963,
+      "step": 9483
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7804963655438498e-05,
+      "loss": 0.5694,
+      "step": 9484
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7804518001405473e-05,
+      "loss": 0.6753,
+      "step": 9485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7804072307715406e-05,
+      "loss": 0.6617,
+      "step": 9486
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780362657437056e-05,
+      "loss": 0.5894,
+      "step": 9487
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.78031808013732e-05,
+      "loss": 0.579,
+      "step": 9488
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7802734988725596e-05,
+      "loss": 0.6554,
+      "step": 9489
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780228913643001e-05,
+      "loss": 0.6574,
+      "step": 9490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7801843244488705e-05,
+      "loss": 0.6007,
+      "step": 9491
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7801397312903953e-05,
+      "loss": 0.5882,
+      "step": 9492
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7800951341678018e-05,
+      "loss": 0.6448,
+      "step": 9493
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.780050533081316e-05,
+      "loss": 0.6258,
+      "step": 9494
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7800059280311653e-05,
+      "loss": 0.6405,
+      "step": 9495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779961319017576e-05,
+      "loss": 0.6277,
+      "step": 9496
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7799167060407744e-05,
+      "loss": 0.6685,
+      "step": 9497
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7798720891009882e-05,
+      "loss": 0.5708,
+      "step": 9498
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7798274681984435e-05,
+      "loss": 0.6674,
+      "step": 9499
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7797828433333668e-05,
+      "loss": 0.6537,
+      "step": 9500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779738214505985e-05,
+      "loss": 0.5645,
+      "step": 9501
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7796935817165252e-05,
+      "loss": 0.6537,
+      "step": 9502
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7796489449652137e-05,
+      "loss": 0.5892,
+      "step": 9503
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779604304252278e-05,
+      "loss": 0.6007,
+      "step": 9504
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7795596595779443e-05,
+      "loss": 0.5979,
+      "step": 9505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7795150109424394e-05,
+      "loss": 0.6704,
+      "step": 9506
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779470358345991e-05,
+      "loss": 0.5984,
+      "step": 9507
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.779425701788825e-05,
+      "loss": 0.56,
+      "step": 9508
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7793810412711687e-05,
+      "loss": 0.6521,
+      "step": 9509
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7793363767932494e-05,
+      "loss": 0.602,
+      "step": 9510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7792917083552936e-05,
+      "loss": 0.6072,
+      "step": 9511
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7792470359575285e-05,
+      "loss": 0.623,
+      "step": 9512
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7792023596001806e-05,
+      "loss": 0.5464,
+      "step": 9513
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7791576792834777e-05,
+      "loss": 0.642,
+      "step": 9514
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7791129950076466e-05,
+      "loss": 0.6174,
+      "step": 9515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7790683067729136e-05,
+      "loss": 0.6063,
+      "step": 9516
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7790236145795068e-05,
+      "loss": 0.5678,
+      "step": 9517
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7789789184276528e-05,
+      "loss": 0.6273,
+      "step": 9518
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778934218317579e-05,
+      "loss": 0.6,
+      "step": 9519
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7788895142495117e-05,
+      "loss": 0.5952,
+      "step": 9520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7788448062236788e-05,
+      "loss": 0.6212,
+      "step": 9521
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778800094240308e-05,
+      "loss": 0.6184,
+      "step": 9522
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7787553782996254e-05,
+      "loss": 0.6253,
+      "step": 9523
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7787106584018586e-05,
+      "loss": 0.6076,
+      "step": 9524
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778665934547235e-05,
+      "loss": 0.544,
+      "step": 9525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7786212067359816e-05,
+      "loss": 0.665,
+      "step": 9526
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7785764749683257e-05,
+      "loss": 0.5744,
+      "step": 9527
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778531739244495e-05,
+      "loss": 0.5557,
+      "step": 9528
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7784869995647166e-05,
+      "loss": 0.5892,
+      "step": 9529
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7784422559292174e-05,
+      "loss": 0.5785,
+      "step": 9530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7783975083382255e-05,
+      "loss": 0.594,
+      "step": 9531
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7783527567919678e-05,
+      "loss": 0.6049,
+      "step": 9532
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7783080012906717e-05,
+      "loss": 0.5535,
+      "step": 9533
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7782632418345648e-05,
+      "loss": 0.6278,
+      "step": 9534
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7782184784238743e-05,
+      "loss": 0.5947,
+      "step": 9535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7781737110588284e-05,
+      "loss": 0.6306,
+      "step": 9536
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778128939739653e-05,
+      "loss": 0.6735,
+      "step": 9537
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.778084164466577e-05,
+      "loss": 0.6424,
+      "step": 9538
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7780393852398275e-05,
+      "loss": 0.6355,
+      "step": 9539
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7779946020596323e-05,
+      "loss": 0.6076,
+      "step": 9540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7779498149262184e-05,
+      "loss": 0.5893,
+      "step": 9541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7779050238398135e-05,
+      "loss": 0.5617,
+      "step": 9542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7778602288006457e-05,
+      "loss": 0.6526,
+      "step": 9543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777815429808942e-05,
+      "loss": 0.6431,
+      "step": 9544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7777706268649307e-05,
+      "loss": 0.5521,
+      "step": 9545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7777258199688385e-05,
+      "loss": 0.6185,
+      "step": 9546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7776810091208938e-05,
+      "loss": 0.5902,
+      "step": 9547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777636194321324e-05,
+      "loss": 0.6017,
+      "step": 9548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7775913755703577e-05,
+      "loss": 0.6134,
+      "step": 9549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777546552868221e-05,
+      "loss": 0.6329,
+      "step": 9550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7775017262151434e-05,
+      "loss": 0.6371,
+      "step": 9551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7774568956113515e-05,
+      "loss": 0.6513,
+      "step": 9552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777412061057073e-05,
+      "loss": 0.6256,
+      "step": 9553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7773672225525364e-05,
+      "loss": 0.574,
+      "step": 9554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7773223800979696e-05,
+      "loss": 0.6343,
+      "step": 9555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7772775336936e-05,
+      "loss": 0.6237,
+      "step": 9556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7772326833396554e-05,
+      "loss": 0.5931,
+      "step": 9557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7771878290363638e-05,
+      "loss": 0.6079,
+      "step": 9558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7771429707839535e-05,
+      "loss": 0.6094,
+      "step": 9559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777098108582652e-05,
+      "loss": 0.6161,
+      "step": 9560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7770532424326874e-05,
+      "loss": 0.6354,
+      "step": 9561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.777008372334288e-05,
+      "loss": 0.5938,
+      "step": 9562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7769634982876815e-05,
+      "loss": 0.5813,
+      "step": 9563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7769186202930955e-05,
+      "loss": 0.6197,
+      "step": 9564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7768737383507587e-05,
+      "loss": 0.606,
+      "step": 9565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776828852460899e-05,
+      "loss": 0.5766,
+      "step": 9566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7767839626237444e-05,
+      "loss": 0.6127,
+      "step": 9567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7767390688395228e-05,
+      "loss": 0.6308,
+      "step": 9568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7766941711084627e-05,
+      "loss": 0.5867,
+      "step": 9569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776649269430792e-05,
+      "loss": 0.6989,
+      "step": 9570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7766043638067388e-05,
+      "loss": 0.5809,
+      "step": 9571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7765594542365315e-05,
+      "loss": 0.6766,
+      "step": 9572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776514540720398e-05,
+      "loss": 0.6515,
+      "step": 9573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776469623258567e-05,
+      "loss": 0.6378,
+      "step": 9574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7764247018512664e-05,
+      "loss": 0.6213,
+      "step": 9575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7763797764987245e-05,
+      "loss": 0.6026,
+      "step": 9576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7763348472011693e-05,
+      "loss": 0.6094,
+      "step": 9577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7762899139588295e-05,
+      "loss": 0.5562,
+      "step": 9578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7762449767719335e-05,
+      "loss": 0.5499,
+      "step": 9579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7762000356407094e-05,
+      "loss": 0.6167,
+      "step": 9580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7761550905653854e-05,
+      "loss": 0.6184,
+      "step": 9581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7761101415461903e-05,
+      "loss": 0.563,
+      "step": 9582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.776065188583352e-05,
+      "loss": 0.5843,
+      "step": 9583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7760202316770995e-05,
+      "loss": 0.6599,
+      "step": 9584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775975270827661e-05,
+      "loss": 0.6401,
+      "step": 9585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7759303060352645e-05,
+      "loss": 0.6051,
+      "step": 9586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7758853373001392e-05,
+      "loss": 0.6087,
+      "step": 9587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775840364622513e-05,
+      "loss": 0.5695,
+      "step": 9588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7757953880026148e-05,
+      "loss": 0.6286,
+      "step": 9589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775750407440673e-05,
+      "loss": 0.6233,
+      "step": 9590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7757054229369164e-05,
+      "loss": 0.579,
+      "step": 9591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7756604344915732e-05,
+      "loss": 0.579,
+      "step": 9592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775615442104872e-05,
+      "loss": 0.6544,
+      "step": 9593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7755704457770423e-05,
+      "loss": 0.6327,
+      "step": 9594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775525445508311e-05,
+      "loss": 0.6068,
+      "step": 9595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7754804412989086e-05,
+      "loss": 0.6133,
+      "step": 9596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7754354331490625e-05,
+      "loss": 0.5929,
+      "step": 9597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7753904210590023e-05,
+      "loss": 0.5856,
+      "step": 9598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775345405028956e-05,
+      "loss": 0.612,
+      "step": 9599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7753003850591525e-05,
+      "loss": 0.6686,
+      "step": 9600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775255361149821e-05,
+      "loss": 0.6651,
+      "step": 9601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7752103333011895e-05,
+      "loss": 0.5508,
+      "step": 9602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775165301513488e-05,
+      "loss": 0.625,
+      "step": 9603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775120265786944e-05,
+      "loss": 0.6549,
+      "step": 9604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775075226121787e-05,
+      "loss": 0.6946,
+      "step": 9605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.775030182518246e-05,
+      "loss": 0.5583,
+      "step": 9606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774985134976549e-05,
+      "loss": 0.6327,
+      "step": 9607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7749400834969263e-05,
+      "loss": 0.6378,
+      "step": 9608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774895028079606e-05,
+      "loss": 0.6295,
+      "step": 9609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774849968724817e-05,
+      "loss": 0.6439,
+      "step": 9610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7748049054327884e-05,
+      "loss": 0.6297,
+      "step": 9611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774759838203749e-05,
+      "loss": 0.6115,
+      "step": 9612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7747147670379283e-05,
+      "loss": 0.5767,
+      "step": 9613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7746696919355547e-05,
+      "loss": 0.6089,
+      "step": 9614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7746246128968575e-05,
+      "loss": 0.5846,
+      "step": 9615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7745795299220662e-05,
+      "loss": 0.6433,
+      "step": 9616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7745344430114093e-05,
+      "loss": 0.6053,
+      "step": 9617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7744893521651157e-05,
+      "loss": 0.5423,
+      "step": 9618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774444257383415e-05,
+      "loss": 0.6799,
+      "step": 9619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7743991586665363e-05,
+      "loss": 0.6056,
+      "step": 9620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.774354056014709e-05,
+      "loss": 0.6348,
+      "step": 9621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7743089494281617e-05,
+      "loss": 0.5851,
+      "step": 9622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7742638389071238e-05,
+      "loss": 0.6365,
+      "step": 9623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.7742187244518244e-05,
+      "loss": 0.6277,
+      "step": 9624
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.774173606062493e-05,
+      "loss": 0.6227,
+      "step": 9625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7741284837393592e-05,
+      "loss": 0.6016,
+      "step": 9626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7740833574826514e-05,
+      "loss": 0.5864,
+      "step": 9627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7740382272925997e-05,
+      "loss": 0.6626,
+      "step": 9628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7739930931694325e-05,
+      "loss": 0.6083,
+      "step": 9629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7739479551133803e-05,
+      "loss": 0.6094,
+      "step": 9630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773902813124672e-05,
+      "loss": 0.5886,
+      "step": 9631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7738576672035365e-05,
+      "loss": 0.633,
+      "step": 9632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7738125173502035e-05,
+      "loss": 0.6067,
+      "step": 9633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7737673635649025e-05,
+      "loss": 0.6315,
+      "step": 9634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773722205847863e-05,
+      "loss": 0.6598,
+      "step": 9635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7736770441993146e-05,
+      "loss": 0.5886,
+      "step": 9636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773631878619486e-05,
+      "loss": 0.6137,
+      "step": 9637
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7735867091086078e-05,
+      "loss": 0.5803,
+      "step": 9638
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773541535666909e-05,
+      "loss": 0.6325,
+      "step": 9639
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7734963582946188e-05,
+      "loss": 0.6352,
+      "step": 9640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7734511769919672e-05,
+      "loss": 0.6327,
+      "step": 9641
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7734059917591837e-05,
+      "loss": 0.6569,
+      "step": 9642
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7733608025964976e-05,
+      "loss": 0.6379,
+      "step": 9643
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773315609504139e-05,
+      "loss": 0.615,
+      "step": 9644
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7732704124823374e-05,
+      "loss": 0.6095,
+      "step": 9645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773225211531322e-05,
+      "loss": 0.6099,
+      "step": 9646
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773180006651323e-05,
+      "loss": 0.6495,
+      "step": 9647
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.77313479784257e-05,
+      "loss": 0.602,
+      "step": 9648
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7730895851052928e-05,
+      "loss": 0.6198,
+      "step": 9649
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.773044368439721e-05,
+      "loss": 0.6156,
+      "step": 9650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7729991478460845e-05,
+      "loss": 0.6661,
+      "step": 9651
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7729539233246128e-05,
+      "loss": 0.6407,
+      "step": 9652
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7729086948755358e-05,
+      "loss": 0.5829,
+      "step": 9653
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7728634624990834e-05,
+      "loss": 0.6068,
+      "step": 9654
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772818226195485e-05,
+      "loss": 0.5908,
+      "step": 9655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7727729859649715e-05,
+      "loss": 0.6223,
+      "step": 9656
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772727741807772e-05,
+      "loss": 0.648,
+      "step": 9657
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772682493724116e-05,
+      "loss": 0.6498,
+      "step": 9658
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7726372417142344e-05,
+      "loss": 0.5799,
+      "step": 9659
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7725919857783565e-05,
+      "loss": 0.6248,
+      "step": 9660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772546725916713e-05,
+      "loss": 0.5944,
+      "step": 9661
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772501462129533e-05,
+      "loss": 0.6498,
+      "step": 9662
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7724561944170465e-05,
+      "loss": 0.5965,
+      "step": 9663
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772410922779484e-05,
+      "loss": 0.6583,
+      "step": 9664
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772365647217076e-05,
+      "loss": 0.5938,
+      "step": 9665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7723203677300513e-05,
+      "loss": 0.6085,
+      "step": 9666
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7722750843186407e-05,
+      "loss": 0.61,
+      "step": 9667
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7722297969830742e-05,
+      "loss": 0.5856,
+      "step": 9668
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772184505723582e-05,
+      "loss": 0.6283,
+      "step": 9669
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7721392105403944e-05,
+      "loss": 0.5836,
+      "step": 9670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7720939114337408e-05,
+      "loss": 0.5827,
+      "step": 9671
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7720486084038526e-05,
+      "loss": 0.6223,
+      "step": 9672
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.772003301450959e-05,
+      "loss": 0.5771,
+      "step": 9673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7719579905752905e-05,
+      "loss": 0.58,
+      "step": 9674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7719126757770774e-05,
+      "loss": 0.6326,
+      "step": 9675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.77186735705655e-05,
+      "loss": 0.5344,
+      "step": 9676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7718220344139384e-05,
+      "loss": 0.6255,
+      "step": 9677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771776707849473e-05,
+      "loss": 0.6173,
+      "step": 9678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7717313773633842e-05,
+      "loss": 0.6184,
+      "step": 9679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7716860429559024e-05,
+      "loss": 0.6153,
+      "step": 9680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7716407046272578e-05,
+      "loss": 0.5467,
+      "step": 9681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7715953623776803e-05,
+      "loss": 0.6045,
+      "step": 9682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771550016207401e-05,
+      "loss": 0.5441,
+      "step": 9683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7715046661166507e-05,
+      "loss": 0.5713,
+      "step": 9684
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7714593121056586e-05,
+      "loss": 0.608,
+      "step": 9685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771413954174656e-05,
+      "loss": 0.5827,
+      "step": 9686
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771368592323873e-05,
+      "loss": 0.588,
+      "step": 9687
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7713232265535406e-05,
+      "loss": 0.6023,
+      "step": 9688
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771277856863889e-05,
+      "loss": 0.6427,
+      "step": 9689
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7712324832551485e-05,
+      "loss": 0.5868,
+      "step": 9690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.77118710572755e-05,
+      "loss": 0.6218,
+      "step": 9691
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771141724281324e-05,
+      "loss": 0.6452,
+      "step": 9692
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.771096338916701e-05,
+      "loss": 0.5849,
+      "step": 9693
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7710509496339115e-05,
+      "loss": 0.6195,
+      "step": 9694
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7710055564331864e-05,
+      "loss": 0.6012,
+      "step": 9695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7709601593147564e-05,
+      "loss": 0.5863,
+      "step": 9696
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7709147582788522e-05,
+      "loss": 0.6564,
+      "step": 9697
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.770869353325704e-05,
+      "loss": 0.6043,
+      "step": 9698
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.770823944455543e-05,
+      "loss": 0.6184,
+      "step": 9699
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7707785316686e-05,
+      "loss": 0.616,
+      "step": 9700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.770733114965105e-05,
+      "loss": 0.6323,
+      "step": 9701
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.77068769434529e-05,
+      "loss": 0.6447,
+      "step": 9702
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7706422698093846e-05,
+      "loss": 0.6279,
+      "step": 9703
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7705968413576202e-05,
+      "loss": 0.5919,
+      "step": 9704
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7705514089902278e-05,
+      "loss": 0.5548,
+      "step": 9705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7705059727074378e-05,
+      "loss": 0.6217,
+      "step": 9706
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7704605325094813e-05,
+      "loss": 0.6137,
+      "step": 9707
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7704150883965893e-05,
+      "loss": 0.5953,
+      "step": 9708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7703696403689926e-05,
+      "loss": 0.6045,
+      "step": 9709
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7703241884269223e-05,
+      "loss": 0.6153,
+      "step": 9710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.770278732570609e-05,
+      "loss": 0.5723,
+      "step": 9711
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.770233272800284e-05,
+      "loss": 0.5661,
+      "step": 9712
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7701878091161783e-05,
+      "loss": 0.5975,
+      "step": 9713
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7701423415185223e-05,
+      "loss": 0.5888,
+      "step": 9714
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7700968700075477e-05,
+      "loss": 0.5822,
+      "step": 9715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7700513945834855e-05,
+      "loss": 0.553,
+      "step": 9716
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7700059152465664e-05,
+      "loss": 0.5631,
+      "step": 9717
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769960431997022e-05,
+      "loss": 0.6745,
+      "step": 9718
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769914944835083e-05,
+      "loss": 0.5633,
+      "step": 9719
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769869453760981e-05,
+      "loss": 0.6907,
+      "step": 9720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7698239587749467e-05,
+      "loss": 0.6041,
+      "step": 9721
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769778459877211e-05,
+      "loss": 0.5459,
+      "step": 9722
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7697329570680063e-05,
+      "loss": 0.5788,
+      "step": 9723
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7696874503475624e-05,
+      "loss": 0.5532,
+      "step": 9724
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7696419397161112e-05,
+      "loss": 0.6401,
+      "step": 9725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769596425173884e-05,
+      "loss": 0.6343,
+      "step": 9726
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7695509067211117e-05,
+      "loss": 0.6183,
+      "step": 9727
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7695053843580266e-05,
+      "loss": 0.6282,
+      "step": 9728
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7694598580848588e-05,
+      "loss": 0.6637,
+      "step": 9729
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7694143279018397e-05,
+      "loss": 0.5737,
+      "step": 9730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7693687938092015e-05,
+      "loss": 0.5987,
+      "step": 9731
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7693232558071756e-05,
+      "loss": 0.6254,
+      "step": 9732
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7692777138959924e-05,
+      "loss": 0.6467,
+      "step": 9733
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769232168075884e-05,
+      "loss": 0.6308,
+      "step": 9734
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7691866183470815e-05,
+      "loss": 0.6854,
+      "step": 9735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7691410647098168e-05,
+      "loss": 0.5983,
+      "step": 9736
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7690955071643206e-05,
+      "loss": 0.6133,
+      "step": 9737
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7690499457108252e-05,
+      "loss": 0.7268,
+      "step": 9738
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.769004380349562e-05,
+      "loss": 0.6272,
+      "step": 9739
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.768958811080762e-05,
+      "loss": 0.6018,
+      "step": 9740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7689132379046574e-05,
+      "loss": 0.592,
+      "step": 9741
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7688676608214793e-05,
+      "loss": 0.6022,
+      "step": 9742
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7688220798314593e-05,
+      "loss": 0.5597,
+      "step": 9743
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.768776494934829e-05,
+      "loss": 0.6804,
+      "step": 9744
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7687309061318206e-05,
+      "loss": 0.5586,
+      "step": 9745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.768685313422665e-05,
+      "loss": 0.5375,
+      "step": 9746
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7686397168075947e-05,
+      "loss": 0.581,
+      "step": 9747
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7685941162868405e-05,
+      "loss": 0.6554,
+      "step": 9748
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7685485118606345e-05,
+      "loss": 0.5951,
+      "step": 9749
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7685029035292085e-05,
+      "loss": 0.583,
+      "step": 9750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7684572912927943e-05,
+      "loss": 0.6819,
+      "step": 9751
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7684116751516233e-05,
+      "loss": 0.5848,
+      "step": 9752
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7683660551059278e-05,
+      "loss": 0.5612,
+      "step": 9753
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7683204311559394e-05,
+      "loss": 0.6027,
+      "step": 9754
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7682748033018897e-05,
+      "loss": 0.6123,
+      "step": 9755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7682291715440105e-05,
+      "loss": 0.6501,
+      "step": 9756
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7681835358825342e-05,
+      "loss": 0.6259,
+      "step": 9757
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7681378963176922e-05,
+      "loss": 0.4998,
+      "step": 9758
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7680922528497167e-05,
+      "loss": 0.6297,
+      "step": 9759
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7680466054788397e-05,
+      "loss": 0.6163,
+      "step": 9760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7680009542052925e-05,
+      "loss": 0.5668,
+      "step": 9761
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7679552990293078e-05,
+      "loss": 0.6702,
+      "step": 9762
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7679096399511174e-05,
+      "loss": 0.642,
+      "step": 9763
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7678639769709533e-05,
+      "loss": 0.6118,
+      "step": 9764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.767818310089047e-05,
+      "loss": 0.6362,
+      "step": 9765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7677726393056315e-05,
+      "loss": 0.6044,
+      "step": 9766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.767726964620938e-05,
+      "loss": 0.6149,
+      "step": 9767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7676812860351994e-05,
+      "loss": 0.5982,
+      "step": 9768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.767635603548647e-05,
+      "loss": 0.6145,
+      "step": 9769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.767589917161513e-05,
+      "loss": 0.5771,
+      "step": 9770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.7675442268740303e-05,
+      "loss": 0.6332,
+      "step": 9771
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7674985326864307e-05,
+      "loss": 0.6009,
+      "step": 9772
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.767452834598946e-05,
+      "loss": 0.6603,
+      "step": 9773
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7674071326118088e-05,
+      "loss": 0.6621,
+      "step": 9774
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7673614267252513e-05,
+      "loss": 0.5383,
+      "step": 9775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7673157169395054e-05,
+      "loss": 0.6162,
+      "step": 9776
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7672700032548035e-05,
+      "loss": 0.6068,
+      "step": 9777
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7672242856713783e-05,
+      "loss": 0.527,
+      "step": 9778
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.767178564189462e-05,
+      "loss": 0.6106,
+      "step": 9779
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.767132838809287e-05,
+      "loss": 0.5648,
+      "step": 9780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7670871095310845e-05,
+      "loss": 0.6325,
+      "step": 9781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.767041376355088e-05,
+      "loss": 0.6107,
+      "step": 9782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76699563928153e-05,
+      "loss": 0.6324,
+      "step": 9783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.766949898310642e-05,
+      "loss": 0.6545,
+      "step": 9784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7669041534426573e-05,
+      "loss": 0.597,
+      "step": 9785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.766858404677808e-05,
+      "loss": 0.5928,
+      "step": 9786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7668126520163267e-05,
+      "loss": 0.5631,
+      "step": 9787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7667668954584454e-05,
+      "loss": 0.5943,
+      "step": 9788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.766721135004397e-05,
+      "loss": 0.6369,
+      "step": 9789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.766675370654414e-05,
+      "loss": 0.657,
+      "step": 9790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7666296024087292e-05,
+      "loss": 0.5673,
+      "step": 9791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7665838302675745e-05,
+      "loss": 0.6155,
+      "step": 9792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7665380542311832e-05,
+      "loss": 0.5879,
+      "step": 9793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7664922742997875e-05,
+      "loss": 0.6646,
+      "step": 9794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7664464904736198e-05,
+      "loss": 0.5926,
+      "step": 9795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7664007027529135e-05,
+      "loss": 0.5792,
+      "step": 9796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7663549111379e-05,
+      "loss": 0.5402,
+      "step": 9797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7663091156288135e-05,
+      "loss": 0.5946,
+      "step": 9798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7662633162258854e-05,
+      "loss": 0.6091,
+      "step": 9799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7662175129293494e-05,
+      "loss": 0.51,
+      "step": 9800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7661717057394377e-05,
+      "loss": 0.5919,
+      "step": 9801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.766125894656383e-05,
+      "loss": 0.6572,
+      "step": 9802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7660800796804184e-05,
+      "loss": 0.6157,
+      "step": 9803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7660342608117766e-05,
+      "loss": 0.6316,
+      "step": 9804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76598843805069e-05,
+      "loss": 0.6296,
+      "step": 9805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7659426113973924e-05,
+      "loss": 0.6072,
+      "step": 9806
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765896780852116e-05,
+      "loss": 0.5871,
+      "step": 9807
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7658509464150936e-05,
+      "loss": 0.6266,
+      "step": 9808
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765805108086558e-05,
+      "loss": 0.612,
+      "step": 9809
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7657592658667426e-05,
+      "loss": 0.5994,
+      "step": 9810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7657134197558804e-05,
+      "loss": 0.5527,
+      "step": 9811
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7656675697542035e-05,
+      "loss": 0.6016,
+      "step": 9812
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7656217158619457e-05,
+      "loss": 0.5892,
+      "step": 9813
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76557585807934e-05,
+      "loss": 0.5741,
+      "step": 9814
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7655299964066186e-05,
+      "loss": 0.6336,
+      "step": 9815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7654841308440158e-05,
+      "loss": 0.6438,
+      "step": 9816
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7654382613917634e-05,
+      "loss": 0.6249,
+      "step": 9817
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765392388050095e-05,
+      "loss": 0.5814,
+      "step": 9818
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765346510819244e-05,
+      "loss": 0.571,
+      "step": 9819
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7653006296994432e-05,
+      "loss": 0.6174,
+      "step": 9820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7652547446909256e-05,
+      "loss": 0.6354,
+      "step": 9821
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7652088557939248e-05,
+      "loss": 0.5905,
+      "step": 9822
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7651629630086733e-05,
+      "loss": 0.6612,
+      "step": 9823
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765117066335405e-05,
+      "loss": 0.6297,
+      "step": 9824
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.765071165774353e-05,
+      "loss": 0.6071,
+      "step": 9825
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7650252613257496e-05,
+      "loss": 0.6135,
+      "step": 9826
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7649793529898295e-05,
+      "loss": 0.5942,
+      "step": 9827
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764933440766825e-05,
+      "loss": 0.5745,
+      "step": 9828
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76488752465697e-05,
+      "loss": 0.622,
+      "step": 9829
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764841604660497e-05,
+      "loss": 0.6008,
+      "step": 9830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76479568077764e-05,
+      "loss": 0.6257,
+      "step": 9831
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7647497530086322e-05,
+      "loss": 0.5572,
+      "step": 9832
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764703821353707e-05,
+      "loss": 0.5362,
+      "step": 9833
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7646578858130974e-05,
+      "loss": 0.6278,
+      "step": 9834
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7646119463870376e-05,
+      "loss": 0.6185,
+      "step": 9835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7645660030757603e-05,
+      "loss": 0.6172,
+      "step": 9836
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7645200558794993e-05,
+      "loss": 0.6746,
+      "step": 9837
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764474104798488e-05,
+      "loss": 0.6107,
+      "step": 9838
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76442814983296e-05,
+      "loss": 0.5734,
+      "step": 9839
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7643821909831485e-05,
+      "loss": 0.5623,
+      "step": 9840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7643362282492875e-05,
+      "loss": 0.6248,
+      "step": 9841
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7642902616316096e-05,
+      "loss": 0.6582,
+      "step": 9842
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7642442911303497e-05,
+      "loss": 0.5934,
+      "step": 9843
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7641983167457402e-05,
+      "loss": 0.6182,
+      "step": 9844
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764152338478016e-05,
+      "loss": 0.6105,
+      "step": 9845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7641063563274096e-05,
+      "loss": 0.578,
+      "step": 9846
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764060370294155e-05,
+      "loss": 0.6153,
+      "step": 9847
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.764014380378486e-05,
+      "loss": 0.5999,
+      "step": 9848
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.763968386580636e-05,
+      "loss": 0.6309,
+      "step": 9849
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.763922388900839e-05,
+      "loss": 0.6169,
+      "step": 9850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7638763873393287e-05,
+      "loss": 0.6602,
+      "step": 9851
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7638303818963385e-05,
+      "loss": 0.544,
+      "step": 9852
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7637843725721024e-05,
+      "loss": 0.6704,
+      "step": 9853
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7637383593668545e-05,
+      "loss": 0.659,
+      "step": 9854
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.763692342280828e-05,
+      "loss": 0.6261,
+      "step": 9855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7636463213142575e-05,
+      "loss": 0.6391,
+      "step": 9856
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7636002964673763e-05,
+      "loss": 0.5808,
+      "step": 9857
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.763554267740418e-05,
+      "loss": 0.6169,
+      "step": 9858
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7635082351336172e-05,
+      "loss": 0.6582,
+      "step": 9859
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7634621986472073e-05,
+      "loss": 0.684,
+      "step": 9860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7634161582814224e-05,
+      "loss": 0.631,
+      "step": 9861
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7633701140364963e-05,
+      "loss": 0.6263,
+      "step": 9862
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7633240659126635e-05,
+      "loss": 0.6047,
+      "step": 9863
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7632780139101574e-05,
+      "loss": 0.6251,
+      "step": 9864
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7632319580292117e-05,
+      "loss": 0.6145,
+      "step": 9865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7631858982700613e-05,
+      "loss": 0.6105,
+      "step": 9866
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76313983463294e-05,
+      "loss": 0.6642,
+      "step": 9867
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7630937671180815e-05,
+      "loss": 0.6235,
+      "step": 9868
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76304769572572e-05,
+      "loss": 0.6559,
+      "step": 9869
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76300162045609e-05,
+      "loss": 0.6086,
+      "step": 9870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.762955541309425e-05,
+      "loss": 0.612,
+      "step": 9871
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7629094582859597e-05,
+      "loss": 0.6177,
+      "step": 9872
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.762863371385928e-05,
+      "loss": 0.6586,
+      "step": 9873
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7628172806095637e-05,
+      "loss": 0.5877,
+      "step": 9874
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7627711859571018e-05,
+      "loss": 0.5913,
+      "step": 9875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7627250874287757e-05,
+      "loss": 0.5944,
+      "step": 9876
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7626789850248205e-05,
+      "loss": 0.6358,
+      "step": 9877
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7626328787454697e-05,
+      "loss": 0.6118,
+      "step": 9878
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7625867685909584e-05,
+      "loss": 0.6277,
+      "step": 9879
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7625406545615198e-05,
+      "loss": 0.599,
+      "step": 9880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7624945366573892e-05,
+      "loss": 0.5765,
+      "step": 9881
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7624484148788005e-05,
+      "loss": 0.5562,
+      "step": 9882
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7624022892259878e-05,
+      "loss": 0.6569,
+      "step": 9883
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7623561596991862e-05,
+      "loss": 0.594,
+      "step": 9884
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7623100262986292e-05,
+      "loss": 0.6062,
+      "step": 9885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.762263889024552e-05,
+      "loss": 0.594,
+      "step": 9886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7622177478771888e-05,
+      "loss": 0.6186,
+      "step": 9887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.762171602856774e-05,
+      "loss": 0.6137,
+      "step": 9888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.762125453963542e-05,
+      "loss": 0.6482,
+      "step": 9889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7620793011977273e-05,
+      "loss": 0.6516,
+      "step": 9890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7620331445595647e-05,
+      "loss": 0.6485,
+      "step": 9891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7619869840492883e-05,
+      "loss": 0.6516,
+      "step": 9892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761940819667133e-05,
+      "loss": 0.6247,
+      "step": 9893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761894651413333e-05,
+      "loss": 0.6403,
+      "step": 9894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7618484792881237e-05,
+      "loss": 0.5505,
+      "step": 9895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7618023032917387e-05,
+      "loss": 0.6642,
+      "step": 9896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761756123424413e-05,
+      "loss": 0.6468,
+      "step": 9897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7617099396863818e-05,
+      "loss": 0.5989,
+      "step": 9898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761663752077879e-05,
+      "loss": 0.549,
+      "step": 9899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7616175605991393e-05,
+      "loss": 0.602,
+      "step": 9900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7615713652503984e-05,
+      "loss": 0.6209,
+      "step": 9901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7615251660318903e-05,
+      "loss": 0.5968,
+      "step": 9902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7614789629438494e-05,
+      "loss": 0.6369,
+      "step": 9903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761432755986511e-05,
+      "loss": 0.5525,
+      "step": 9904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76138654516011e-05,
+      "loss": 0.6506,
+      "step": 9905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7613403304648808e-05,
+      "loss": 0.5942,
+      "step": 9906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7612941119010583e-05,
+      "loss": 0.5775,
+      "step": 9907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7612478894688777e-05,
+      "loss": 0.5794,
+      "step": 9908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7612016631685734e-05,
+      "loss": 0.6027,
+      "step": 9909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7611554330003806e-05,
+      "loss": 0.6049,
+      "step": 9910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761109198964534e-05,
+      "loss": 0.5971,
+      "step": 9911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.761062961061269e-05,
+      "loss": 0.6596,
+      "step": 9912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7610167192908202e-05,
+      "loss": 0.6216,
+      "step": 9913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7609704736534222e-05,
+      "loss": 0.5893,
+      "step": 9914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.7609242241493104e-05,
+      "loss": 0.5968,
+      "step": 9915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.76087797077872e-05,
+      "loss": 0.6867,
+      "step": 9916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.760831713541886e-05,
+      "loss": 0.569,
+      "step": 9917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.760785452439043e-05,
+      "loss": 0.6372,
+      "step": 9918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760739187470426e-05,
+      "loss": 0.6538,
+      "step": 9919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760692918636271e-05,
+      "loss": 0.5708,
+      "step": 9920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760646645936812e-05,
+      "loss": 0.6132,
+      "step": 9921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7606003693722845e-05,
+      "loss": 0.598,
+      "step": 9922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7605540889429242e-05,
+      "loss": 0.5993,
+      "step": 9923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760507804648966e-05,
+      "loss": 0.6203,
+      "step": 9924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7604615164906444e-05,
+      "loss": 0.5556,
+      "step": 9925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7604152244681955e-05,
+      "loss": 0.6091,
+      "step": 9926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760368928581854e-05,
+      "loss": 0.5832,
+      "step": 9927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7603226288318552e-05,
+      "loss": 0.6412,
+      "step": 9928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7602763252184347e-05,
+      "loss": 0.5824,
+      "step": 9929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7602300177418273e-05,
+      "loss": 0.5843,
+      "step": 9930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7601837064022686e-05,
+      "loss": 0.5964,
+      "step": 9931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.760137391199994e-05,
+      "loss": 0.614,
+      "step": 9932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7600910721352384e-05,
+      "loss": 0.5646,
+      "step": 9933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7600447492082378e-05,
+      "loss": 0.6508,
+      "step": 9934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759998422419227e-05,
+      "loss": 0.6251,
+      "step": 9935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7599520917684417e-05,
+      "loss": 0.5634,
+      "step": 9936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7599057572561175e-05,
+      "loss": 0.604,
+      "step": 9937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7598594188824892e-05,
+      "loss": 0.6384,
+      "step": 9938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759813076647793e-05,
+      "loss": 0.6183,
+      "step": 9939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7597667305522635e-05,
+      "loss": 0.5876,
+      "step": 9940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7597203805961373e-05,
+      "loss": 0.5663,
+      "step": 9941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7596740267796493e-05,
+      "loss": 0.5856,
+      "step": 9942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7596276691030346e-05,
+      "loss": 0.6118,
+      "step": 9943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7595813075665298e-05,
+      "loss": 0.594,
+      "step": 9944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75953494217037e-05,
+      "loss": 0.6595,
+      "step": 9945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75948857291479e-05,
+      "loss": 0.5834,
+      "step": 9946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759442199800027e-05,
+      "loss": 0.5184,
+      "step": 9947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759395822826315e-05,
+      "loss": 0.6407,
+      "step": 9948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759349441993891e-05,
+      "loss": 0.6162,
+      "step": 9949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7593030573029896e-05,
+      "loss": 0.669,
+      "step": 9950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759256668753847e-05,
+      "loss": 0.6013,
+      "step": 9951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7592102763466993e-05,
+      "loss": 0.6351,
+      "step": 9952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759163880081782e-05,
+      "loss": 0.6262,
+      "step": 9953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7591174799593303e-05,
+      "loss": 0.6206,
+      "step": 9954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7590710759795806e-05,
+      "loss": 0.5624,
+      "step": 9955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.759024668142768e-05,
+      "loss": 0.5702,
+      "step": 9956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758978256449129e-05,
+      "loss": 0.6195,
+      "step": 9957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7589318408988995e-05,
+      "loss": 0.5873,
+      "step": 9958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758885421492315e-05,
+      "loss": 0.6573,
+      "step": 9959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758838998229611e-05,
+      "loss": 0.6602,
+      "step": 9960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758792571111024e-05,
+      "loss": 0.6471,
+      "step": 9961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75874614013679e-05,
+      "loss": 0.6191,
+      "step": 9962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7586997053071444e-05,
+      "loss": 0.5339,
+      "step": 9963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7586532666223234e-05,
+      "loss": 0.6342,
+      "step": 9964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7586068240825633e-05,
+      "loss": 0.6273,
+      "step": 9965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7585603776880992e-05,
+      "loss": 0.5888,
+      "step": 9966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758513927439168e-05,
+      "loss": 0.6249,
+      "step": 9967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7584674733360055e-05,
+      "loss": 0.6199,
+      "step": 9968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7584210153788475e-05,
+      "loss": 0.6601,
+      "step": 9969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75837455356793e-05,
+      "loss": 0.6065,
+      "step": 9970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7583280879034896e-05,
+      "loss": 0.6355,
+      "step": 9971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758281618385762e-05,
+      "loss": 0.5687,
+      "step": 9972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758235145014983e-05,
+      "loss": 0.6542,
+      "step": 9973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75818866779139e-05,
+      "loss": 0.627,
+      "step": 9974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.758142186715218e-05,
+      "loss": 0.619,
+      "step": 9975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7580957017867034e-05,
+      "loss": 0.6411,
+      "step": 9976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7580492130060825e-05,
+      "loss": 0.5898,
+      "step": 9977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7580027203735915e-05,
+      "loss": 0.6203,
+      "step": 9978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7579562238894668e-05,
+      "loss": 0.5797,
+      "step": 9979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7579097235539447e-05,
+      "loss": 0.6421,
+      "step": 9980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.757863219367261e-05,
+      "loss": 0.617,
+      "step": 9981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7578167113296525e-05,
+      "loss": 0.6462,
+      "step": 9982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7577701994413555e-05,
+      "loss": 0.6702,
+      "step": 9983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.757723683702606e-05,
+      "loss": 0.633,
+      "step": 9984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7576771641136407e-05,
+      "loss": 0.6247,
+      "step": 9985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7576306406746955e-05,
+      "loss": 0.6054,
+      "step": 9986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7575841133860073e-05,
+      "loss": 0.6333,
+      "step": 9987
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7575375822478126e-05,
+      "loss": 0.6539,
+      "step": 9988
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7574910472603474e-05,
+      "loss": 0.6105,
+      "step": 9989
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7574445084238482e-05,
+      "loss": 0.6443,
+      "step": 9990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7573979657385517e-05,
+      "loss": 0.613,
+      "step": 9991
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7573514192046945e-05,
+      "loss": 0.5918,
+      "step": 9992
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7573048688225126e-05,
+      "loss": 0.6282,
+      "step": 9993
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7572583145922434e-05,
+      "loss": 0.6331,
+      "step": 9994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7572117565141223e-05,
+      "loss": 0.6247,
+      "step": 9995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.757165194588387e-05,
+      "loss": 0.5267,
+      "step": 9996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7571186288152735e-05,
+      "loss": 0.6287,
+      "step": 9997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7570720591950187e-05,
+      "loss": 0.6013,
+      "step": 9998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7570254857278584e-05,
+      "loss": 0.6639,
+      "step": 9999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7569789084140305e-05,
+      "loss": 0.5918,
+      "step": 10000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7569323272537707e-05,
+      "loss": 0.5948,
+      "step": 10001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7568857422473166e-05,
+      "loss": 0.6397,
+      "step": 10002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756839153394904e-05,
+      "loss": 0.5744,
+      "step": 10003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75679256069677e-05,
+      "loss": 0.6312,
+      "step": 10004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756745964153151e-05,
+      "loss": 0.6242,
+      "step": 10005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756699363764285e-05,
+      "loss": 0.6281,
+      "step": 10006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7566527595304072e-05,
+      "loss": 0.587,
+      "step": 10007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7566061514517554e-05,
+      "loss": 0.6139,
+      "step": 10008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756559539528566e-05,
+      "loss": 0.6251,
+      "step": 10009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7565129237610765e-05,
+      "loss": 0.5792,
+      "step": 10010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7564663041495226e-05,
+      "loss": 0.6225,
+      "step": 10011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7564196806941424e-05,
+      "loss": 0.7179,
+      "step": 10012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756373053395172e-05,
+      "loss": 0.6004,
+      "step": 10013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7563264222528488e-05,
+      "loss": 0.6379,
+      "step": 10014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7562797872674094e-05,
+      "loss": 0.6367,
+      "step": 10015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.756233148439091e-05,
+      "loss": 0.6159,
+      "step": 10016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7561865057681303e-05,
+      "loss": 0.5705,
+      "step": 10017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7561398592547647e-05,
+      "loss": 0.649,
+      "step": 10018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7560932088992308e-05,
+      "loss": 0.6125,
+      "step": 10019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7560465547017662e-05,
+      "loss": 0.5998,
+      "step": 10020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7559998966626074e-05,
+      "loss": 0.6436,
+      "step": 10021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755953234781992e-05,
+      "loss": 0.6702,
+      "step": 10022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7559065690601568e-05,
+      "loss": 0.5862,
+      "step": 10023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7558598994973387e-05,
+      "loss": 0.614,
+      "step": 10024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755813226093775e-05,
+      "loss": 0.6248,
+      "step": 10025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755766548849703e-05,
+      "loss": 0.6626,
+      "step": 10026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75571986776536e-05,
+      "loss": 0.6153,
+      "step": 10027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7556731828409832e-05,
+      "loss": 0.5847,
+      "step": 10028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755626494076809e-05,
+      "loss": 0.6042,
+      "step": 10029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755579801473076e-05,
+      "loss": 0.6278,
+      "step": 10030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7555331050300204e-05,
+      "loss": 0.6434,
+      "step": 10031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.75548640474788e-05,
+      "loss": 0.5701,
+      "step": 10032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755439700626892e-05,
+      "loss": 0.5797,
+      "step": 10033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755392992667293e-05,
+      "loss": 0.6224,
+      "step": 10034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7553462808693217e-05,
+      "loss": 0.6119,
+      "step": 10035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7552995652332144e-05,
+      "loss": 0.6205,
+      "step": 10036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7552528457592086e-05,
+      "loss": 0.5596,
+      "step": 10037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7552061224475424e-05,
+      "loss": 0.6084,
+      "step": 10038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7551593952984525e-05,
+      "loss": 0.5871,
+      "step": 10039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7551126643121766e-05,
+      "loss": 0.6701,
+      "step": 10040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.755065929488952e-05,
+      "loss": 0.6393,
+      "step": 10041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7550191908290162e-05,
+      "loss": 0.5718,
+      "step": 10042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754972448332607e-05,
+      "loss": 0.556,
+      "step": 10043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7549257019999614e-05,
+      "loss": 0.6065,
+      "step": 10044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7548789518313176e-05,
+      "loss": 0.5391,
+      "step": 10045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7548321978269125e-05,
+      "loss": 0.5636,
+      "step": 10046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754785439986984e-05,
+      "loss": 0.6485,
+      "step": 10047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7547386783117697e-05,
+      "loss": 0.5265,
+      "step": 10048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7546919128015073e-05,
+      "loss": 0.5807,
+      "step": 10049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754645143456434e-05,
+      "loss": 0.67,
+      "step": 10050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7545983702767876e-05,
+      "loss": 0.585,
+      "step": 10051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754551593262806e-05,
+      "loss": 0.6335,
+      "step": 10052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754504812414727e-05,
+      "loss": 0.6415,
+      "step": 10053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7544580277327877e-05,
+      "loss": 0.5771,
+      "step": 10054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7544112392172267e-05,
+      "loss": 0.5595,
+      "step": 10055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754364446868281e-05,
+      "loss": 0.5614,
+      "step": 10056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7543176506861886e-05,
+      "loss": 0.649,
+      "step": 10057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754270850671187e-05,
+      "loss": 0.6484,
+      "step": 10058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7542240468235147e-05,
+      "loss": 0.6157,
+      "step": 10059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754177239143409e-05,
+      "loss": 0.6037,
+      "step": 10060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754130427631108e-05,
+      "loss": 0.6004,
+      "step": 10061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.7540836122868494e-05,
+      "loss": 0.6107,
+      "step": 10062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.754036793110871e-05,
+      "loss": 0.5739,
+      "step": 10063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.753989970103411e-05,
+      "loss": 0.6571,
+      "step": 10064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.753943143264707e-05,
+      "loss": 0.6162,
+      "step": 10065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7538963125949972e-05,
+      "loss": 0.5906,
+      "step": 10066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7538494780945196e-05,
+      "loss": 0.5565,
+      "step": 10067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7538026397635116e-05,
+      "loss": 0.6951,
+      "step": 10068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7537557976022118e-05,
+      "loss": 0.5763,
+      "step": 10069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7537089516108578e-05,
+      "loss": 0.6384,
+      "step": 10070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7536621017896883e-05,
+      "loss": 0.6533,
+      "step": 10071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.753615248138941e-05,
+      "loss": 0.5534,
+      "step": 10072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7535683906588537e-05,
+      "loss": 0.597,
+      "step": 10073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7535215293496647e-05,
+      "loss": 0.5987,
+      "step": 10074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7534746642116124e-05,
+      "loss": 0.6077,
+      "step": 10075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.753427795244934e-05,
+      "loss": 0.5945,
+      "step": 10076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.753380922449869e-05,
+      "loss": 0.6653,
+      "step": 10077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7533340458266546e-05,
+      "loss": 0.6074,
+      "step": 10078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7532871653755292e-05,
+      "loss": 0.6047,
+      "step": 10079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.753240281096731e-05,
+      "loss": 0.619,
+      "step": 10080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7531933929904988e-05,
+      "loss": 0.6245,
+      "step": 10081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7531465010570704e-05,
+      "loss": 0.6375,
+      "step": 10082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7530996052966838e-05,
+      "loss": 0.607,
+      "step": 10083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7530527057095775e-05,
+      "loss": 0.6094,
+      "step": 10084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.75300580229599e-05,
+      "loss": 0.6701,
+      "step": 10085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7529588950561593e-05,
+      "loss": 0.5513,
+      "step": 10086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752911983990324e-05,
+      "loss": 0.6387,
+      "step": 10087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7528650690987225e-05,
+      "loss": 0.6445,
+      "step": 10088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752818150381593e-05,
+      "loss": 0.6649,
+      "step": 10089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752771227839174e-05,
+      "loss": 0.5611,
+      "step": 10090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7527243014717043e-05,
+      "loss": 0.5398,
+      "step": 10091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7526773712794216e-05,
+      "loss": 0.6205,
+      "step": 10092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752630437262565e-05,
+      "loss": 0.6592,
+      "step": 10093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7525834994213725e-05,
+      "loss": 0.5474,
+      "step": 10094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752536557756083e-05,
+      "loss": 0.6046,
+      "step": 10095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7524896122669346e-05,
+      "loss": 0.6304,
+      "step": 10096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7524426629541662e-05,
+      "loss": 0.6199,
+      "step": 10097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7523957098180164e-05,
+      "loss": 0.6201,
+      "step": 10098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7523487528587235e-05,
+      "loss": 0.6491,
+      "step": 10099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7523017920765262e-05,
+      "loss": 0.6485,
+      "step": 10100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7522548274716636e-05,
+      "loss": 0.5656,
+      "step": 10101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7522078590443733e-05,
+      "loss": 0.6421,
+      "step": 10102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752160886794895e-05,
+      "loss": 0.5615,
+      "step": 10103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752113910723467e-05,
+      "loss": 0.6054,
+      "step": 10104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7520669308303277e-05,
+      "loss": 0.6153,
+      "step": 10105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.752019947115716e-05,
+      "loss": 0.5732,
+      "step": 10106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751972959579871e-05,
+      "loss": 0.582,
+      "step": 10107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7519259682230312e-05,
+      "loss": 0.6027,
+      "step": 10108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751878973045435e-05,
+      "loss": 0.6663,
+      "step": 10109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7518319740473217e-05,
+      "loss": 0.6224,
+      "step": 10110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7517849712289302e-05,
+      "loss": 0.5591,
+      "step": 10111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7517379645904986e-05,
+      "loss": 0.6171,
+      "step": 10112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7516909541322666e-05,
+      "loss": 0.5776,
+      "step": 10113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7516439398544727e-05,
+      "loss": 0.6699,
+      "step": 10114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7515969217573557e-05,
+      "loss": 0.6091,
+      "step": 10115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7515498998411548e-05,
+      "loss": 0.6393,
+      "step": 10116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7515028741061083e-05,
+      "loss": 0.6369,
+      "step": 10117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751455844552456e-05,
+      "loss": 0.5738,
+      "step": 10118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7514088111804367e-05,
+      "loss": 0.5817,
+      "step": 10119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7513617739902888e-05,
+      "loss": 0.614,
+      "step": 10120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751314732982252e-05,
+      "loss": 0.6048,
+      "step": 10121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7512676881565647e-05,
+      "loss": 0.6128,
+      "step": 10122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7512206395134664e-05,
+      "loss": 0.6095,
+      "step": 10123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7511735870531955e-05,
+      "loss": 0.5931,
+      "step": 10124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7511265307759925e-05,
+      "loss": 0.6556,
+      "step": 10125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751079470682095e-05,
+      "loss": 0.5885,
+      "step": 10126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.751032406771743e-05,
+      "loss": 0.5985,
+      "step": 10127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750985339045175e-05,
+      "loss": 0.5621,
+      "step": 10128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7509382675026312e-05,
+      "loss": 0.5999,
+      "step": 10129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7508911921443495e-05,
+      "loss": 0.587,
+      "step": 10130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7508441129705703e-05,
+      "loss": 0.6776,
+      "step": 10131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750797029981532e-05,
+      "loss": 0.545,
+      "step": 10132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7507499431774742e-05,
+      "loss": 0.6486,
+      "step": 10133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750702852558636e-05,
+      "loss": 0.5888,
+      "step": 10134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7506557581252567e-05,
+      "loss": 0.6211,
+      "step": 10135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750608659877576e-05,
+      "loss": 0.5806,
+      "step": 10136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7505615578158325e-05,
+      "loss": 0.5467,
+      "step": 10137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750514451940266e-05,
+      "loss": 0.6234,
+      "step": 10138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7504673422511155e-05,
+      "loss": 0.5886,
+      "step": 10139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.750420228748621e-05,
+      "loss": 0.6254,
+      "step": 10140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7503731114330215e-05,
+      "loss": 0.6187,
+      "step": 10141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7503259903045566e-05,
+      "loss": 0.5846,
+      "step": 10142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7502788653634656e-05,
+      "loss": 0.6152,
+      "step": 10143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7502317366099878e-05,
+      "loss": 0.6104,
+      "step": 10144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7501846040443626e-05,
+      "loss": 0.5857,
+      "step": 10145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.75013746766683e-05,
+      "loss": 0.5912,
+      "step": 10146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7500903274776294e-05,
+      "loss": 0.5865,
+      "step": 10147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7500431834769998e-05,
+      "loss": 0.621,
+      "step": 10148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749996035665181e-05,
+      "loss": 0.6528,
+      "step": 10149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749948884042413e-05,
+      "loss": 0.5997,
+      "step": 10150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7499017286089352e-05,
+      "loss": 0.5825,
+      "step": 10151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749854569364987e-05,
+      "loss": 0.6592,
+      "step": 10152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749807406310808e-05,
+      "loss": 0.6843,
+      "step": 10153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7497602394466383e-05,
+      "loss": 0.6653,
+      "step": 10154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749713068772717e-05,
+      "loss": 0.5565,
+      "step": 10155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7496658942892838e-05,
+      "loss": 0.5603,
+      "step": 10156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749618715996579e-05,
+      "loss": 0.602,
+      "step": 10157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7495715338948416e-05,
+      "loss": 0.5998,
+      "step": 10158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7495243479843124e-05,
+      "loss": 0.6402,
+      "step": 10159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7494771582652302e-05,
+      "loss": 0.6127,
+      "step": 10160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7494299647378347e-05,
+      "loss": 0.5827,
+      "step": 10161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7493827674023665e-05,
+      "loss": 0.6248,
+      "step": 10162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7493355662590647e-05,
+      "loss": 0.5807,
+      "step": 10163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749288361308169e-05,
+      "loss": 0.6146,
+      "step": 10164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7492411525499205e-05,
+      "loss": 0.6116,
+      "step": 10165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.749193939984558e-05,
+      "loss": 0.6127,
+      "step": 10166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7491467236123217e-05,
+      "loss": 0.6206,
+      "step": 10167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7490995034334513e-05,
+      "loss": 0.6546,
+      "step": 10168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7490522794481873e-05,
+      "loss": 0.5636,
+      "step": 10169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7490050516567692e-05,
+      "loss": 0.6065,
+      "step": 10170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748957820059437e-05,
+      "loss": 0.6559,
+      "step": 10171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7489105846564306e-05,
+      "loss": 0.5646,
+      "step": 10172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7488633454479903e-05,
+      "loss": 0.5901,
+      "step": 10173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7488161024343564e-05,
+      "loss": 0.5781,
+      "step": 10174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7487688556157682e-05,
+      "loss": 0.6257,
+      "step": 10175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7487216049924662e-05,
+      "loss": 0.5712,
+      "step": 10176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7486743505646903e-05,
+      "loss": 0.6563,
+      "step": 10177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748627092332681e-05,
+      "loss": 0.5877,
+      "step": 10178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748579830296678e-05,
+      "loss": 0.6134,
+      "step": 10179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748532564456922e-05,
+      "loss": 0.5717,
+      "step": 10180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7484852948136526e-05,
+      "loss": 0.5585,
+      "step": 10181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.74843802136711e-05,
+      "loss": 0.6001,
+      "step": 10182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748390744117535e-05,
+      "loss": 0.6035,
+      "step": 10183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7483434630651675e-05,
+      "loss": 0.6076,
+      "step": 10184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7482961782102473e-05,
+      "loss": 0.6396,
+      "step": 10185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7482488895530152e-05,
+      "loss": 0.6141,
+      "step": 10186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7482015970937114e-05,
+      "loss": 0.6524,
+      "step": 10187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7481543008325764e-05,
+      "loss": 0.5794,
+      "step": 10188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.74810700076985e-05,
+      "loss": 0.6231,
+      "step": 10189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.748059696905773e-05,
+      "loss": 0.5666,
+      "step": 10190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7480123892405855e-05,
+      "loss": 0.5488,
+      "step": 10191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7479650777745277e-05,
+      "loss": 0.5686,
+      "step": 10192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7479177625078406e-05,
+      "loss": 0.5896,
+      "step": 10193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7478704434407644e-05,
+      "loss": 0.6346,
+      "step": 10194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747823120573539e-05,
+      "loss": 0.601,
+      "step": 10195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747775793906406e-05,
+      "loss": 0.5989,
+      "step": 10196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7477284634396052e-05,
+      "loss": 0.6429,
+      "step": 10197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7476811291733765e-05,
+      "loss": 0.5796,
+      "step": 10198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747633791107961e-05,
+      "loss": 0.5897,
+      "step": 10199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7475864492436e-05,
+      "loss": 0.5849,
+      "step": 10200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7475391035805324e-05,
+      "loss": 0.5971,
+      "step": 10201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7474917541190005e-05,
+      "loss": 0.6243,
+      "step": 10202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7474444008592435e-05,
+      "loss": 0.5991,
+      "step": 10203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747397043801503e-05,
+      "loss": 0.6423,
+      "step": 10204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747349682946019e-05,
+      "loss": 0.6276,
+      "step": 10205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747302318293032e-05,
+      "loss": 0.5841,
+      "step": 10206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7472549498427837e-05,
+      "loss": 0.619,
+      "step": 10207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747207577595514e-05,
+      "loss": 0.5771,
+      "step": 10208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7471602015514638e-05,
+      "loss": 0.5979,
+      "step": 10209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7471128217108737e-05,
+      "loss": 0.6694,
+      "step": 10210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.7470654380739846e-05,
+      "loss": 0.6312,
+      "step": 10211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.747018050641037e-05,
+      "loss": 0.5524,
+      "step": 10212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7469706594122722e-05,
+      "loss": 0.5789,
+      "step": 10213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7469232643879304e-05,
+      "loss": 0.6013,
+      "step": 10214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7468758655682532e-05,
+      "loss": 0.6284,
+      "step": 10215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7468284629534808e-05,
+      "loss": 0.6331,
+      "step": 10216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746781056543854e-05,
+      "loss": 0.5672,
+      "step": 10217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7467336463396143e-05,
+      "loss": 0.5868,
+      "step": 10218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746686232341002e-05,
+      "loss": 0.5618,
+      "step": 10219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7466388145482588e-05,
+      "loss": 0.5861,
+      "step": 10220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7465913929616244e-05,
+      "loss": 0.6325,
+      "step": 10221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746543967581341e-05,
+      "loss": 0.5399,
+      "step": 10222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7464965384076486e-05,
+      "loss": 0.5766,
+      "step": 10223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746449105440789e-05,
+      "loss": 0.5527,
+      "step": 10224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746401668681003e-05,
+      "loss": 0.6167,
+      "step": 10225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746354228128531e-05,
+      "loss": 0.6196,
+      "step": 10226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746306783783615e-05,
+      "loss": 0.5672,
+      "step": 10227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7462593356464958e-05,
+      "loss": 0.6318,
+      "step": 10228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746211883717414e-05,
+      "loss": 0.6103,
+      "step": 10229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7461644279966113e-05,
+      "loss": 0.5914,
+      "step": 10230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7461169684843285e-05,
+      "loss": 0.5995,
+      "step": 10231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.746069505180807e-05,
+      "loss": 0.5766,
+      "step": 10232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7460220380862877e-05,
+      "loss": 0.6814,
+      "step": 10233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745974567201012e-05,
+      "loss": 0.5815,
+      "step": 10234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745927092525221e-05,
+      "loss": 0.6173,
+      "step": 10235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745879614059156e-05,
+      "loss": 0.6376,
+      "step": 10236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7458321318030584e-05,
+      "loss": 0.6192,
+      "step": 10237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745784645757169e-05,
+      "loss": 0.6088,
+      "step": 10238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.74573715592173e-05,
+      "loss": 0.5718,
+      "step": 10239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7456896622969817e-05,
+      "loss": 0.5998,
+      "step": 10240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7456421648831658e-05,
+      "loss": 0.6632,
+      "step": 10241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7455946636805237e-05,
+      "loss": 0.595,
+      "step": 10242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7455471586892968e-05,
+      "loss": 0.5824,
+      "step": 10243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7454996499097265e-05,
+      "loss": 0.6518,
+      "step": 10244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745452137342054e-05,
+      "loss": 0.6081,
+      "step": 10245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745404620986521e-05,
+      "loss": 0.6005,
+      "step": 10246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7453571008433686e-05,
+      "loss": 0.5798,
+      "step": 10247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745309576912839e-05,
+      "loss": 0.6042,
+      "step": 10248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745262049195173e-05,
+      "loss": 0.5679,
+      "step": 10249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.745214517690612e-05,
+      "loss": 0.6723,
+      "step": 10250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7451669823993976e-05,
+      "loss": 0.6029,
+      "step": 10251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7451194433217718e-05,
+      "loss": 0.605,
+      "step": 10252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7450719004579762e-05,
+      "loss": 0.6298,
+      "step": 10253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7450243538082517e-05,
+      "loss": 0.5856,
+      "step": 10254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7449768033728404e-05,
+      "loss": 0.5633,
+      "step": 10255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7449292491519837e-05,
+      "loss": 0.5375,
+      "step": 10256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7448816911459234e-05,
+      "loss": 0.5716,
+      "step": 10257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744834129354901e-05,
+      "loss": 0.5809,
+      "step": 10258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7447865637791585e-05,
+      "loss": 0.5868,
+      "step": 10259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744738994418937e-05,
+      "loss": 0.664,
+      "step": 10260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744691421274479e-05,
+      "loss": 0.6223,
+      "step": 10261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744643844346025e-05,
+      "loss": 0.5901,
+      "step": 10262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7445962636338182e-05,
+      "loss": 0.6265,
+      "step": 10263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7445486791380998e-05,
+      "loss": 0.578,
+      "step": 10264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744501090859111e-05,
+      "loss": 0.6529,
+      "step": 10265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7444534987970944e-05,
+      "loss": 0.5314,
+      "step": 10266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7444059029522916e-05,
+      "loss": 0.5811,
+      "step": 10267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7443583033249445e-05,
+      "loss": 0.6398,
+      "step": 10268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744310699915295e-05,
+      "loss": 0.594,
+      "step": 10269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.744263092723584e-05,
+      "loss": 0.6384,
+      "step": 10270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7442154817500552e-05,
+      "loss": 0.6578,
+      "step": 10271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7441678669949487e-05,
+      "loss": 0.6022,
+      "step": 10272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7441202484585083e-05,
+      "loss": 0.6353,
+      "step": 10273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7440726261409745e-05,
+      "loss": 0.6703,
+      "step": 10274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7440250000425898e-05,
+      "loss": 0.6162,
+      "step": 10275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743977370163596e-05,
+      "loss": 0.6286,
+      "step": 10276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7439297365042356e-05,
+      "loss": 0.6105,
+      "step": 10277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7438820990647503e-05,
+      "loss": 0.6188,
+      "step": 10278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7438344578453822e-05,
+      "loss": 0.6005,
+      "step": 10279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743786812846373e-05,
+      "loss": 0.6769,
+      "step": 10280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7437391640679657e-05,
+      "loss": 0.5987,
+      "step": 10281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7436915115104015e-05,
+      "loss": 0.5853,
+      "step": 10282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743643855173923e-05,
+      "loss": 0.6175,
+      "step": 10283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7435961950587726e-05,
+      "loss": 0.572,
+      "step": 10284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7435485311651917e-05,
+      "loss": 0.6169,
+      "step": 10285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7435008634934234e-05,
+      "loss": 0.5632,
+      "step": 10286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7434531920437092e-05,
+      "loss": 0.6308,
+      "step": 10287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7434055168162918e-05,
+      "loss": 0.5744,
+      "step": 10288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743357837811413e-05,
+      "loss": 0.6245,
+      "step": 10289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7433101550293155e-05,
+      "loss": 0.5457,
+      "step": 10290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7432624684702412e-05,
+      "loss": 0.5925,
+      "step": 10291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743214778134433e-05,
+      "loss": 0.5572,
+      "step": 10292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7431670840221328e-05,
+      "loss": 0.6384,
+      "step": 10293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7431193861335825e-05,
+      "loss": 0.6204,
+      "step": 10294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7430716844690253e-05,
+      "loss": 0.6255,
+      "step": 10295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.743023979028703e-05,
+      "loss": 0.6578,
+      "step": 10296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742976269812859e-05,
+      "loss": 0.6372,
+      "step": 10297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7429285568217343e-05,
+      "loss": 0.5864,
+      "step": 10298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7428808400555725e-05,
+      "loss": 0.5476,
+      "step": 10299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7428331195146147e-05,
+      "loss": 0.5788,
+      "step": 10300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742785395199105e-05,
+      "loss": 0.6858,
+      "step": 10301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7427376671092853e-05,
+      "loss": 0.5719,
+      "step": 10302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7426899352453975e-05,
+      "loss": 0.5896,
+      "step": 10303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742642199607685e-05,
+      "loss": 0.6258,
+      "step": 10304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.74259446019639e-05,
+      "loss": 0.5647,
+      "step": 10305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742546717011755e-05,
+      "loss": 0.5533,
+      "step": 10306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7424989700540225e-05,
+      "loss": 0.5706,
+      "step": 10307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7424512193234353e-05,
+      "loss": 0.6121,
+      "step": 10308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7424034648202362e-05,
+      "loss": 0.605,
+      "step": 10309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7423557065446674e-05,
+      "loss": 0.637,
+      "step": 10310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7423079444969723e-05,
+      "loss": 0.6123,
+      "step": 10311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7422601786773927e-05,
+      "loss": 0.6889,
+      "step": 10312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742212409086172e-05,
+      "loss": 0.5879,
+      "step": 10313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7421646357235524e-05,
+      "loss": 0.5883,
+      "step": 10314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.742116858589777e-05,
+      "loss": 0.6095,
+      "step": 10315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7420690776850887e-05,
+      "loss": 0.6563,
+      "step": 10316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7420212930097304e-05,
+      "loss": 0.6533,
+      "step": 10317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7419735045639437e-05,
+      "loss": 0.5803,
+      "step": 10318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7419257123479728e-05,
+      "loss": 0.5891,
+      "step": 10319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.74187791636206e-05,
+      "loss": 0.6115,
+      "step": 10320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7418301166064487e-05,
+      "loss": 0.6737,
+      "step": 10321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7417823130813807e-05,
+      "loss": 0.6007,
+      "step": 10322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7417345057871002e-05,
+      "loss": 0.6292,
+      "step": 10323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741686694723849e-05,
+      "loss": 0.6365,
+      "step": 10324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7416388798918703e-05,
+      "loss": 0.6011,
+      "step": 10325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741591061291408e-05,
+      "loss": 0.6873,
+      "step": 10326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7415432389227035e-05,
+      "loss": 0.5995,
+      "step": 10327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7414954127860012e-05,
+      "loss": 0.6074,
+      "step": 10328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7414475828815433e-05,
+      "loss": 0.5782,
+      "step": 10329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741399749209573e-05,
+      "loss": 0.6255,
+      "step": 10330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7413519117703336e-05,
+      "loss": 0.6184,
+      "step": 10331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741304070564068e-05,
+      "loss": 0.5987,
+      "step": 10332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741256225591019e-05,
+      "loss": 0.6114,
+      "step": 10333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7412083768514303e-05,
+      "loss": 0.6077,
+      "step": 10334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7411605243455448e-05,
+      "loss": 0.6016,
+      "step": 10335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7411126680736057e-05,
+      "loss": 0.5886,
+      "step": 10336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7410648080358558e-05,
+      "loss": 0.6255,
+      "step": 10337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.741016944232539e-05,
+      "loss": 0.6618,
+      "step": 10338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740969076663898e-05,
+      "loss": 0.6037,
+      "step": 10339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7409212053301756e-05,
+      "loss": 0.594,
+      "step": 10340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740873330231616e-05,
+      "loss": 0.5133,
+      "step": 10341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740825451368462e-05,
+      "loss": 0.57,
+      "step": 10342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7407775687409566e-05,
+      "loss": 0.5934,
+      "step": 10343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740729682349344e-05,
+      "loss": 0.5907,
+      "step": 10344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7406817921938667e-05,
+      "loss": 0.6288,
+      "step": 10345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740633898274768e-05,
+      "loss": 0.6222,
+      "step": 10346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740586000592292e-05,
+      "loss": 0.621,
+      "step": 10347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7405380991466816e-05,
+      "loss": 0.5904,
+      "step": 10348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.74049019393818e-05,
+      "loss": 0.5991,
+      "step": 10349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740442284967031e-05,
+      "loss": 0.6157,
+      "step": 10350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7403943722334778e-05,
+      "loss": 0.6446,
+      "step": 10351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740346455737764e-05,
+      "loss": 0.612,
+      "step": 10352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740298535480133e-05,
+      "loss": 0.6127,
+      "step": 10353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7402506114608285e-05,
+      "loss": 0.6016,
+      "step": 10354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7402026836800937e-05,
+      "loss": 0.6163,
+      "step": 10355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7401547521381723e-05,
+      "loss": 0.5287,
+      "step": 10356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.7401068168353078e-05,
+      "loss": 0.5666,
+      "step": 10357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740058877771744e-05,
+      "loss": 0.6108,
+      "step": 10358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.740010934947724e-05,
+      "loss": 0.5985,
+      "step": 10359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739962988363492e-05,
+      "loss": 0.6113,
+      "step": 10360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7399150380192915e-05,
+      "loss": 0.6342,
+      "step": 10361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7398670839153658e-05,
+      "loss": 0.641,
+      "step": 10362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7398191260519587e-05,
+      "loss": 0.6323,
+      "step": 10363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739771164429314e-05,
+      "loss": 0.5932,
+      "step": 10364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7397231990476752e-05,
+      "loss": 0.648,
+      "step": 10365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7396752299072864e-05,
+      "loss": 0.607,
+      "step": 10366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739627257008391e-05,
+      "loss": 0.6603,
+      "step": 10367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7395792803512332e-05,
+      "loss": 0.6126,
+      "step": 10368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739531299936056e-05,
+      "loss": 0.594,
+      "step": 10369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7394833157631038e-05,
+      "loss": 0.6181,
+      "step": 10370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7394353278326206e-05,
+      "loss": 0.6462,
+      "step": 10371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7393873361448497e-05,
+      "loss": 0.5574,
+      "step": 10372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739339340700035e-05,
+      "loss": 0.6314,
+      "step": 10373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.739291341498421e-05,
+      "loss": 0.6301,
+      "step": 10374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7392433385402512e-05,
+      "loss": 0.6414,
+      "step": 10375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7391953318257692e-05,
+      "loss": 0.6766,
+      "step": 10376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7391473213552196e-05,
+      "loss": 0.5992,
+      "step": 10377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7390993071288455e-05,
+      "loss": 0.6415,
+      "step": 10378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7390512891468917e-05,
+      "loss": 0.5934,
+      "step": 10379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7390032674096018e-05,
+      "loss": 0.5708,
+      "step": 10380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.73895524191722e-05,
+      "loss": 0.6077,
+      "step": 10381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.73890721266999e-05,
+      "loss": 0.5611,
+      "step": 10382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7388591796681562e-05,
+      "loss": 0.6331,
+      "step": 10383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7388111429119624e-05,
+      "loss": 0.6338,
+      "step": 10384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7387631024016526e-05,
+      "loss": 0.5914,
+      "step": 10385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7387150581374716e-05,
+      "loss": 0.6439,
+      "step": 10386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7386670101196627e-05,
+      "loss": 0.5679,
+      "step": 10387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7386189583484707e-05,
+      "loss": 0.6144,
+      "step": 10388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7385709028241393e-05,
+      "loss": 0.62,
+      "step": 10389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7385228435469126e-05,
+      "loss": 0.5941,
+      "step": 10390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.738474780517035e-05,
+      "loss": 0.6022,
+      "step": 10391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7384267137347516e-05,
+      "loss": 0.6255,
+      "step": 10392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7383786432003052e-05,
+      "loss": 0.5978,
+      "step": 10393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7383305689139406e-05,
+      "loss": 0.5624,
+      "step": 10394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.738282490875902e-05,
+      "loss": 0.6074,
+      "step": 10395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7382344090864343e-05,
+      "loss": 0.5735,
+      "step": 10396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.738186323545781e-05,
+      "loss": 0.6344,
+      "step": 10397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.738138234254187e-05,
+      "loss": 0.6352,
+      "step": 10398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7380901412118965e-05,
+      "loss": 0.602,
+      "step": 10399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.738042044419154e-05,
+      "loss": 0.598,
+      "step": 10400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7379939438762036e-05,
+      "loss": 0.5608,
+      "step": 10401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7379458395832896e-05,
+      "loss": 0.6428,
+      "step": 10402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7378977315406563e-05,
+      "loss": 0.5534,
+      "step": 10403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7378496197485493e-05,
+      "loss": 0.5842,
+      "step": 10404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737801504207212e-05,
+      "loss": 0.5832,
+      "step": 10405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737753384916889e-05,
+      "loss": 0.6044,
+      "step": 10406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7377052618778255e-05,
+      "loss": 0.6528,
+      "step": 10407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7376571350902653e-05,
+      "loss": 0.6097,
+      "step": 10408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737609004554453e-05,
+      "loss": 0.6164,
+      "step": 10409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737560870270633e-05,
+      "loss": 0.5924,
+      "step": 10410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737512732239051e-05,
+      "loss": 0.6554,
+      "step": 10411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7374645904599506e-05,
+      "loss": 0.6505,
+      "step": 10412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737416444933576e-05,
+      "loss": 0.6079,
+      "step": 10413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737368295660173e-05,
+      "loss": 0.6605,
+      "step": 10414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7373201426399856e-05,
+      "loss": 0.6212,
+      "step": 10415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737271985873259e-05,
+      "loss": 0.5774,
+      "step": 10416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7372238253602373e-05,
+      "loss": 0.651,
+      "step": 10417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7371756611011653e-05,
+      "loss": 0.61,
+      "step": 10418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737127493096288e-05,
+      "loss": 0.6313,
+      "step": 10419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.73707932134585e-05,
+      "loss": 0.6436,
+      "step": 10420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.737031145850096e-05,
+      "loss": 0.5891,
+      "step": 10421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7369829666092707e-05,
+      "loss": 0.5905,
+      "step": 10422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7369347836236196e-05,
+      "loss": 0.6076,
+      "step": 10423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.736886596893387e-05,
+      "loss": 0.6092,
+      "step": 10424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7368384064188175e-05,
+      "loss": 0.6106,
+      "step": 10425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7367902122001563e-05,
+      "loss": 0.5879,
+      "step": 10426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7367420142376484e-05,
+      "loss": 0.6474,
+      "step": 10427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7366938125315386e-05,
+      "loss": 0.6431,
+      "step": 10428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7366456070820716e-05,
+      "loss": 0.6343,
+      "step": 10429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7365973978894926e-05,
+      "loss": 0.5393,
+      "step": 10430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7365491849540467e-05,
+      "loss": 0.5942,
+      "step": 10431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7365009682759783e-05,
+      "loss": 0.5189,
+      "step": 10432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.736452747855533e-05,
+      "loss": 0.6222,
+      "step": 10433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7364045236929557e-05,
+      "loss": 0.6252,
+      "step": 10434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.736356295788491e-05,
+      "loss": 0.6415,
+      "step": 10435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7363080641423846e-05,
+      "loss": 0.5957,
+      "step": 10436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7362598287548814e-05,
+      "loss": 0.6191,
+      "step": 10437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.736211589626226e-05,
+      "loss": 0.6292,
+      "step": 10438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7361633467566642e-05,
+      "loss": 0.5748,
+      "step": 10439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.736115100146441e-05,
+      "loss": 0.6003,
+      "step": 10440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7360668497958012e-05,
+      "loss": 0.5695,
+      "step": 10441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.73601859570499e-05,
+      "loss": 0.6432,
+      "step": 10442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7359703378742526e-05,
+      "loss": 0.6244,
+      "step": 10443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7359220763038348e-05,
+      "loss": 0.6188,
+      "step": 10444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7358738109939816e-05,
+      "loss": 0.645,
+      "step": 10445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7358255419449377e-05,
+      "loss": 0.5423,
+      "step": 10446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7357772691569484e-05,
+      "loss": 0.6301,
+      "step": 10447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7357289926302597e-05,
+      "loss": 0.591,
+      "step": 10448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7356807123651166e-05,
+      "loss": 0.5965,
+      "step": 10449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7356324283617643e-05,
+      "loss": 0.6076,
+      "step": 10450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7355841406204482e-05,
+      "loss": 0.5872,
+      "step": 10451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7355358491414135e-05,
+      "loss": 0.6023,
+      "step": 10452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.735487553924906e-05,
+      "loss": 0.6433,
+      "step": 10453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7354392549711704e-05,
+      "loss": 0.563,
+      "step": 10454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7353909522804532e-05,
+      "loss": 0.5689,
+      "step": 10455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7353426458529987e-05,
+      "loss": 0.5514,
+      "step": 10456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7352943356890528e-05,
+      "loss": 0.582,
+      "step": 10457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7352460217888615e-05,
+      "loss": 0.5857,
+      "step": 10458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7351977041526695e-05,
+      "loss": 0.6587,
+      "step": 10459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7351493827807226e-05,
+      "loss": 0.594,
+      "step": 10460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7351010576732666e-05,
+      "loss": 0.5914,
+      "step": 10461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7350527288305468e-05,
+      "loss": 0.6238,
+      "step": 10462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7350043962528084e-05,
+      "loss": 0.6291,
+      "step": 10463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7349560599402978e-05,
+      "loss": 0.629,
+      "step": 10464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7349077198932603e-05,
+      "loss": 0.6463,
+      "step": 10465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7348593761119415e-05,
+      "loss": 0.579,
+      "step": 10466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7348110285965866e-05,
+      "loss": 0.6223,
+      "step": 10467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.734762677347442e-05,
+      "loss": 0.6054,
+      "step": 10468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.734714322364753e-05,
+      "loss": 0.6578,
+      "step": 10469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7346659636487654e-05,
+      "loss": 0.5795,
+      "step": 10470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7346176011997247e-05,
+      "loss": 0.693,
+      "step": 10471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.734569235017877e-05,
+      "loss": 0.5912,
+      "step": 10472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7345208651034676e-05,
+      "loss": 0.5832,
+      "step": 10473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.734472491456743e-05,
+      "loss": 0.6417,
+      "step": 10474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7344241140779483e-05,
+      "loss": 0.6861,
+      "step": 10475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7343757329673298e-05,
+      "loss": 0.6784,
+      "step": 10476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7343273481251327e-05,
+      "loss": 0.6193,
+      "step": 10477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7342789595516033e-05,
+      "loss": 0.571,
+      "step": 10478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.734230567246988e-05,
+      "loss": 0.6501,
+      "step": 10479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7341821712115318e-05,
+      "loss": 0.5836,
+      "step": 10480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7341337714454812e-05,
+      "loss": 0.6318,
+      "step": 10481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7340853679490813e-05,
+      "loss": 0.5625,
+      "step": 10482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7340369607225797e-05,
+      "loss": 0.588,
+      "step": 10483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7339885497662203e-05,
+      "loss": 0.633,
+      "step": 10484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7339401350802508e-05,
+      "loss": 0.5618,
+      "step": 10485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7338917166649163e-05,
+      "loss": 0.6093,
+      "step": 10486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733843294520463e-05,
+      "loss": 0.6199,
+      "step": 10487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733794868647137e-05,
+      "loss": 0.5891,
+      "step": 10488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7337464390451844e-05,
+      "loss": 0.6198,
+      "step": 10489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7336980057148512e-05,
+      "loss": 0.6556,
+      "step": 10490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7336495686563838e-05,
+      "loss": 0.6607,
+      "step": 10491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733601127870028e-05,
+      "loss": 0.6221,
+      "step": 10492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7335526833560297e-05,
+      "loss": 0.6097,
+      "step": 10493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733504235114636e-05,
+      "loss": 0.6132,
+      "step": 10494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7334557831460918e-05,
+      "loss": 0.6139,
+      "step": 10495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733407327450644e-05,
+      "loss": 0.6325,
+      "step": 10496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7333588680285394e-05,
+      "loss": 0.5711,
+      "step": 10497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733310404880023e-05,
+      "loss": 0.6147,
+      "step": 10498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733261938005342e-05,
+      "loss": 0.6533,
+      "step": 10499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7332134674047423e-05,
+      "loss": 0.656,
+      "step": 10500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.73316499307847e-05,
+      "loss": 0.6385,
+      "step": 10501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.733116515026772e-05,
+      "loss": 0.5733,
+      "step": 10502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7330680332498943e-05,
+      "loss": 0.6415,
+      "step": 10503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7330195477480833e-05,
+      "loss": 0.5842,
+      "step": 10504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7329710585215848e-05,
+      "loss": 0.6737,
+      "step": 10505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.7329225655706458e-05,
+      "loss": 0.5803,
+      "step": 10506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732874068895513e-05,
+      "loss": 0.6286,
+      "step": 10507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7328255684964324e-05,
+      "loss": 0.6029,
+      "step": 10508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.73277706437365e-05,
+      "loss": 0.5767,
+      "step": 10509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732728556527413e-05,
+      "loss": 0.6274,
+      "step": 10510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7326800449579676e-05,
+      "loss": 0.6093,
+      "step": 10511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7326315296655605e-05,
+      "loss": 0.5977,
+      "step": 10512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732583010650438e-05,
+      "loss": 0.5725,
+      "step": 10513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7325344879128464e-05,
+      "loss": 0.7154,
+      "step": 10514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7324859614530323e-05,
+      "loss": 0.608,
+      "step": 10515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7324374312712433e-05,
+      "loss": 0.603,
+      "step": 10516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7323888973677247e-05,
+      "loss": 0.6075,
+      "step": 10517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7323403597427234e-05,
+      "loss": 0.617,
+      "step": 10518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732291818396487e-05,
+      "loss": 0.5987,
+      "step": 10519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732243273329261e-05,
+      "loss": 0.5873,
+      "step": 10520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7321947245412922e-05,
+      "loss": 0.645,
+      "step": 10521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7321461720328277e-05,
+      "loss": 0.6113,
+      "step": 10522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7320976158041145e-05,
+      "loss": 0.5663,
+      "step": 10523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7320490558553987e-05,
+      "loss": 0.6219,
+      "step": 10524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.732000492186927e-05,
+      "loss": 0.6178,
+      "step": 10525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7319519247989466e-05,
+      "loss": 0.5677,
+      "step": 10526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.731903353691704e-05,
+      "loss": 0.6116,
+      "step": 10527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7318547788654463e-05,
+      "loss": 0.5865,
+      "step": 10528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7318062003204202e-05,
+      "loss": 0.6251,
+      "step": 10529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.731757618056872e-05,
+      "loss": 0.5708,
+      "step": 10530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.73170903207505e-05,
+      "loss": 0.5947,
+      "step": 10531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.731660442375199e-05,
+      "loss": 0.5786,
+      "step": 10532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7316118489575676e-05,
+      "loss": 0.6232,
+      "step": 10533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7315632518224022e-05,
+      "loss": 0.5776,
+      "step": 10534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7315146509699493e-05,
+      "loss": 0.5876,
+      "step": 10535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7314660464004567e-05,
+      "loss": 0.5921,
+      "step": 10536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7314174381141706e-05,
+      "loss": 0.5266,
+      "step": 10537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7313688261113383e-05,
+      "loss": 0.6194,
+      "step": 10538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7313202103922067e-05,
+      "loss": 0.6032,
+      "step": 10539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.731271590957023e-05,
+      "loss": 0.5853,
+      "step": 10540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7312229678060343e-05,
+      "loss": 0.6205,
+      "step": 10541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7311743409394876e-05,
+      "loss": 0.5719,
+      "step": 10542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7311257103576297e-05,
+      "loss": 0.6775,
+      "step": 10543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.731077076060708e-05,
+      "loss": 0.589,
+      "step": 10544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.73102843804897e-05,
+      "loss": 0.6247,
+      "step": 10545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7309797963226616e-05,
+      "loss": 0.628,
+      "step": 10546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7309311508820315e-05,
+      "loss": 0.6114,
+      "step": 10547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730882501727326e-05,
+      "loss": 0.5762,
+      "step": 10548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7308338488587922e-05,
+      "loss": 0.6549,
+      "step": 10549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730785192276678e-05,
+      "loss": 0.5386,
+      "step": 10550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7307365319812296e-05,
+      "loss": 0.6069,
+      "step": 10551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7306878679726954e-05,
+      "loss": 0.5683,
+      "step": 10552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730639200251322e-05,
+      "loss": 0.6367,
+      "step": 10553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7305905288173572e-05,
+      "loss": 0.6116,
+      "step": 10554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730541853671047e-05,
+      "loss": 0.6315,
+      "step": 10555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7304931748126407e-05,
+      "loss": 0.5941,
+      "step": 10556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7304444922423843e-05,
+      "loss": 0.6155,
+      "step": 10557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7303958059605252e-05,
+      "loss": 0.6211,
+      "step": 10558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7303471159673112e-05,
+      "loss": 0.5515,
+      "step": 10559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.73029842226299e-05,
+      "loss": 0.6523,
+      "step": 10560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730249724847808e-05,
+      "loss": 0.6155,
+      "step": 10561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730201023722014e-05,
+      "loss": 0.5323,
+      "step": 10562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7301523188858545e-05,
+      "loss": 0.6117,
+      "step": 10563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.730103610339577e-05,
+      "loss": 0.6364,
+      "step": 10564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7300548980834296e-05,
+      "loss": 0.6237,
+      "step": 10565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7300061821176593e-05,
+      "loss": 0.6113,
+      "step": 10566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7299574624425135e-05,
+      "loss": 0.575,
+      "step": 10567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7299087390582406e-05,
+      "loss": 0.6377,
+      "step": 10568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7298600119650872e-05,
+      "loss": 0.621,
+      "step": 10569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7298112811633014e-05,
+      "loss": 0.5828,
+      "step": 10570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729762546653131e-05,
+      "loss": 0.5305,
+      "step": 10571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729713808434823e-05,
+      "loss": 0.5577,
+      "step": 10572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7296650665086255e-05,
+      "loss": 0.6211,
+      "step": 10573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7296163208747866e-05,
+      "loss": 0.5941,
+      "step": 10574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729567571533553e-05,
+      "loss": 0.5706,
+      "step": 10575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729518818485173e-05,
+      "loss": 0.6455,
+      "step": 10576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7294700617298942e-05,
+      "loss": 0.5454,
+      "step": 10577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7294213012679643e-05,
+      "loss": 0.6286,
+      "step": 10578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7293725370996313e-05,
+      "loss": 0.6424,
+      "step": 10579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729323769225143e-05,
+      "loss": 0.6207,
+      "step": 10580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7292749976447468e-05,
+      "loss": 0.7102,
+      "step": 10581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729226222358691e-05,
+      "loss": 0.6378,
+      "step": 10582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729177443367223e-05,
+      "loss": 0.611,
+      "step": 10583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.729128660670591e-05,
+      "loss": 0.6044,
+      "step": 10584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7290798742690423e-05,
+      "loss": 0.6413,
+      "step": 10585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7290310841628257e-05,
+      "loss": 0.5839,
+      "step": 10586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7289822903521885e-05,
+      "loss": 0.6265,
+      "step": 10587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7289334928373786e-05,
+      "loss": 0.6103,
+      "step": 10588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7288846916186447e-05,
+      "loss": 0.5698,
+      "step": 10589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7288358866962335e-05,
+      "loss": 0.6165,
+      "step": 10590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7287870780703944e-05,
+      "loss": 0.6372,
+      "step": 10591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7287382657413743e-05,
+      "loss": 0.6199,
+      "step": 10592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7286894497094217e-05,
+      "loss": 0.6247,
+      "step": 10593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7286406299747847e-05,
+      "loss": 0.648,
+      "step": 10594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7285918065377112e-05,
+      "loss": 0.6306,
+      "step": 10595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7285429793984495e-05,
+      "loss": 0.5906,
+      "step": 10596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7284941485572472e-05,
+      "loss": 0.5925,
+      "step": 10597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.728445314014353e-05,
+      "loss": 0.5966,
+      "step": 10598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.728396475770015e-05,
+      "loss": 0.6023,
+      "step": 10599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.728347633824481e-05,
+      "loss": 0.6055,
+      "step": 10600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7282987881779992e-05,
+      "loss": 0.5913,
+      "step": 10601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7282499388308184e-05,
+      "loss": 0.6647,
+      "step": 10602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.728201085783186e-05,
+      "loss": 0.6281,
+      "step": 10603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.728152229035351e-05,
+      "loss": 0.5754,
+      "step": 10604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7281033685875608e-05,
+      "loss": 0.5802,
+      "step": 10605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7280545044400642e-05,
+      "loss": 0.5962,
+      "step": 10606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.72800563659311e-05,
+      "loss": 0.6157,
+      "step": 10607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7279567650469454e-05,
+      "loss": 0.6256,
+      "step": 10608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7279078898018197e-05,
+      "loss": 0.5939,
+      "step": 10609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7278590108579805e-05,
+      "loss": 0.666,
+      "step": 10610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7278101282156764e-05,
+      "loss": 0.5287,
+      "step": 10611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7277612418751563e-05,
+      "loss": 0.605,
+      "step": 10612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7277123518366678e-05,
+      "loss": 0.5256,
+      "step": 10613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.72766345810046e-05,
+      "loss": 0.5964,
+      "step": 10614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7276145606667808e-05,
+      "loss": 0.5913,
+      "step": 10615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7275656595358792e-05,
+      "loss": 0.5562,
+      "step": 10616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7275167547080034e-05,
+      "loss": 0.6141,
+      "step": 10617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7274678461834017e-05,
+      "loss": 0.633,
+      "step": 10618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727418933962323e-05,
+      "loss": 0.6846,
+      "step": 10619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7273700180450155e-05,
+      "loss": 0.633,
+      "step": 10620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7273210984317283e-05,
+      "loss": 0.6462,
+      "step": 10621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727272175122709e-05,
+      "loss": 0.6632,
+      "step": 10622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727223248118207e-05,
+      "loss": 0.6329,
+      "step": 10623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727174317418471e-05,
+      "loss": 0.5582,
+      "step": 10624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727125383023749e-05,
+      "loss": 0.6384,
+      "step": 10625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.72707644493429e-05,
+      "loss": 0.5956,
+      "step": 10626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.727027503150343e-05,
+      "loss": 0.6128,
+      "step": 10627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.726978557672156e-05,
+      "loss": 0.6361,
+      "step": 10628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.726929608499978e-05,
+      "loss": 0.5594,
+      "step": 10629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7268806556340577e-05,
+      "loss": 0.61,
+      "step": 10630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7268316990746445e-05,
+      "loss": 0.6148,
+      "step": 10631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.726782738821986e-05,
+      "loss": 0.6221,
+      "step": 10632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7267337748763316e-05,
+      "loss": 0.5524,
+      "step": 10633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7266848072379304e-05,
+      "loss": 0.6347,
+      "step": 10634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7266358359070307e-05,
+      "loss": 0.6525,
+      "step": 10635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7265868608838815e-05,
+      "loss": 0.5957,
+      "step": 10636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7265378821687315e-05,
+      "loss": 0.5934,
+      "step": 10637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7264888997618297e-05,
+      "loss": 0.6507,
+      "step": 10638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7264399136634252e-05,
+      "loss": 0.6247,
+      "step": 10639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.726390923873767e-05,
+      "loss": 0.5839,
+      "step": 10640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7263419303931037e-05,
+      "loss": 0.6128,
+      "step": 10641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7262929332216842e-05,
+      "loss": 0.5509,
+      "step": 10642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7262439323597576e-05,
+      "loss": 0.6247,
+      "step": 10643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7261949278075728e-05,
+      "loss": 0.5778,
+      "step": 10644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7261459195653788e-05,
+      "loss": 0.603,
+      "step": 10645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.726096907633425e-05,
+      "loss": 0.6759,
+      "step": 10646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.72604789201196e-05,
+      "loss": 0.6165,
+      "step": 10647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7259988727012334e-05,
+      "loss": 0.529,
+      "step": 10648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7259498497014933e-05,
+      "loss": 0.612,
+      "step": 10649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7259008230129897e-05,
+      "loss": 0.5747,
+      "step": 10650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7258517926359718e-05,
+      "loss": 0.6223,
+      "step": 10651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7258027585706878e-05,
+      "loss": 0.5602,
+      "step": 10652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7257537208173878e-05,
+      "loss": 0.6287,
+      "step": 10653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7257046793763205e-05,
+      "loss": 0.5742,
+      "step": 10654
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7256556342477348e-05,
+      "loss": 0.6174,
+      "step": 10655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.725606585431881e-05,
+      "loss": 0.5758,
+      "step": 10656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7255575329290072e-05,
+      "loss": 0.6351,
+      "step": 10657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.725508476739363e-05,
+      "loss": 0.6649,
+      "step": 10658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7254594168631977e-05,
+      "loss": 0.5477,
+      "step": 10659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7254103533007608e-05,
+      "loss": 0.6634,
+      "step": 10660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7253612860523017e-05,
+      "loss": 0.5911,
+      "step": 10661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7253122151180693e-05,
+      "loss": 0.5993,
+      "step": 10662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.725263140498313e-05,
+      "loss": 0.5438,
+      "step": 10663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7252140621932823e-05,
+      "loss": 0.5866,
+      "step": 10664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7251649802032265e-05,
+      "loss": 0.5797,
+      "step": 10665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.725115894528395e-05,
+      "loss": 0.5384,
+      "step": 10666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.725066805169038e-05,
+      "loss": 0.5636,
+      "step": 10667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7250177121254034e-05,
+      "loss": 0.6011,
+      "step": 10668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7249686153977417e-05,
+      "loss": 0.6356,
+      "step": 10669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7249195149863026e-05,
+      "loss": 0.5847,
+      "step": 10670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7248704108913347e-05,
+      "loss": 0.6873,
+      "step": 10671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7248213031130877e-05,
+      "loss": 0.5863,
+      "step": 10672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7247721916518118e-05,
+      "loss": 0.6548,
+      "step": 10673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724723076507756e-05,
+      "loss": 0.5639,
+      "step": 10674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7246739576811702e-05,
+      "loss": 0.5369,
+      "step": 10675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7246248351723034e-05,
+      "loss": 0.5681,
+      "step": 10676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724575708981406e-05,
+      "loss": 0.6436,
+      "step": 10677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7245265791087268e-05,
+      "loss": 0.5729,
+      "step": 10678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724477445554516e-05,
+      "loss": 0.6157,
+      "step": 10679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7244283083190233e-05,
+      "loss": 0.5669,
+      "step": 10680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724379167402498e-05,
+      "loss": 0.6227,
+      "step": 10681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7243300228051906e-05,
+      "loss": 0.566,
+      "step": 10682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7242808745273494e-05,
+      "loss": 0.6239,
+      "step": 10683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7242317225692257e-05,
+      "loss": 0.6401,
+      "step": 10684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724182566931068e-05,
+      "loss": 0.6137,
+      "step": 10685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7241334076131267e-05,
+      "loss": 0.644,
+      "step": 10686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7240842446156513e-05,
+      "loss": 0.5894,
+      "step": 10687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.724035077938892e-05,
+      "loss": 0.5749,
+      "step": 10688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7239859075830986e-05,
+      "loss": 0.6196,
+      "step": 10689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7239367335485206e-05,
+      "loss": 0.6297,
+      "step": 10690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723887555835408e-05,
+      "loss": 0.6029,
+      "step": 10691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7238383744440108e-05,
+      "loss": 0.5907,
+      "step": 10692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723789189374579e-05,
+      "loss": 0.6556,
+      "step": 10693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7237400006273623e-05,
+      "loss": 0.5935,
+      "step": 10694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72369080820261e-05,
+      "loss": 0.6086,
+      "step": 10695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7236416121005738e-05,
+      "loss": 0.5702,
+      "step": 10696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723592412321502e-05,
+      "loss": 0.5913,
+      "step": 10697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723543208865645e-05,
+      "loss": 0.5793,
+      "step": 10698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7234940017332536e-05,
+      "loss": 0.58,
+      "step": 10699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7234447909245773e-05,
+      "loss": 0.6242,
+      "step": 10700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723395576439866e-05,
+      "loss": 0.5991,
+      "step": 10701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72334635827937e-05,
+      "loss": 0.6595,
+      "step": 10702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7232971364433394e-05,
+      "loss": 0.5998,
+      "step": 10703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7232479109320238e-05,
+      "loss": 0.6097,
+      "step": 10704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723198681745674e-05,
+      "loss": 0.5801,
+      "step": 10705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72314944888454e-05,
+      "loss": 0.5989,
+      "step": 10706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7231002123488717e-05,
+      "loss": 0.5761,
+      "step": 10707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7230509721389194e-05,
+      "loss": 0.6293,
+      "step": 10708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.723001728254934e-05,
+      "loss": 0.6209,
+      "step": 10709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7229524806971644e-05,
+      "loss": 0.6408,
+      "step": 10710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7229032294658617e-05,
+      "loss": 0.566,
+      "step": 10711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722853974561276e-05,
+      "loss": 0.627,
+      "step": 10712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7228047159836574e-05,
+      "loss": 0.6404,
+      "step": 10713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7227554537332564e-05,
+      "loss": 0.6284,
+      "step": 10714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7227061878103234e-05,
+      "loss": 0.6166,
+      "step": 10715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722656918215108e-05,
+      "loss": 0.5661,
+      "step": 10716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722607644947862e-05,
+      "loss": 0.6157,
+      "step": 10717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722558368008835e-05,
+      "loss": 0.6143,
+      "step": 10718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722509087398277e-05,
+      "loss": 0.6214,
+      "step": 10719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7224598031164384e-05,
+      "loss": 0.5968,
+      "step": 10720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7224105151635703e-05,
+      "loss": 0.6393,
+      "step": 10721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7223612235399226e-05,
+      "loss": 0.5773,
+      "step": 10722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722311928245746e-05,
+      "loss": 0.5674,
+      "step": 10723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7222626292812915e-05,
+      "loss": 0.6116,
+      "step": 10724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7222133266468084e-05,
+      "loss": 0.6532,
+      "step": 10725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722164020342548e-05,
+      "loss": 0.5874,
+      "step": 10726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.722114710368761e-05,
+      "loss": 0.5543,
+      "step": 10727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7220653967256976e-05,
+      "loss": 0.6348,
+      "step": 10728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7220160794136084e-05,
+      "loss": 0.6002,
+      "step": 10729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721966758432744e-05,
+      "loss": 0.6355,
+      "step": 10730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721917433783355e-05,
+      "loss": 0.6371,
+      "step": 10731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721868105465692e-05,
+      "loss": 0.6544,
+      "step": 10732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7218187734800057e-05,
+      "loss": 0.655,
+      "step": 10733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7217694378265473e-05,
+      "loss": 0.6203,
+      "step": 10734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7217200985055667e-05,
+      "loss": 0.6111,
+      "step": 10735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721670755517315e-05,
+      "loss": 0.6545,
+      "step": 10736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7216214088620427e-05,
+      "loss": 0.6092,
+      "step": 10737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7215720585400008e-05,
+      "loss": 0.6105,
+      "step": 10738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72152270455144e-05,
+      "loss": 0.6474,
+      "step": 10739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7214733468966107e-05,
+      "loss": 0.5579,
+      "step": 10740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7214239855757643e-05,
+      "loss": 0.6499,
+      "step": 10741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7213746205891514e-05,
+      "loss": 0.5698,
+      "step": 10742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7213252519370227e-05,
+      "loss": 0.6263,
+      "step": 10743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721275879619629e-05,
+      "loss": 0.5755,
+      "step": 10744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7212265036372217e-05,
+      "loss": 0.5882,
+      "step": 10745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721177123990051e-05,
+      "loss": 0.653,
+      "step": 10746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7211277406783683e-05,
+      "loss": 0.5919,
+      "step": 10747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7210783537024243e-05,
+      "loss": 0.6671,
+      "step": 10748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72102896306247e-05,
+      "loss": 0.629,
+      "step": 10749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7209795687587566e-05,
+      "loss": 0.5873,
+      "step": 10750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7209301707915345e-05,
+      "loss": 0.6097,
+      "step": 10751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7208807691610553e-05,
+      "loss": 0.5488,
+      "step": 10752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7208313638675695e-05,
+      "loss": 0.6379,
+      "step": 10753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7207819549113286e-05,
+      "loss": 0.5884,
+      "step": 10754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7207325422925833e-05,
+      "loss": 0.6333,
+      "step": 10755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7206831260115853e-05,
+      "loss": 0.6131,
+      "step": 10756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7206337060685848e-05,
+      "loss": 0.5829,
+      "step": 10757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.720584282463834e-05,
+      "loss": 0.5782,
+      "step": 10758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7205348551975828e-05,
+      "loss": 0.5634,
+      "step": 10759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7204854242700834e-05,
+      "loss": 0.6248,
+      "step": 10760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7204359896815864e-05,
+      "loss": 0.6647,
+      "step": 10761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.720386551432343e-05,
+      "loss": 0.6143,
+      "step": 10762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7203371095226052e-05,
+      "loss": 0.6351,
+      "step": 10763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7202876639526228e-05,
+      "loss": 0.6355,
+      "step": 10764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7202382147226483e-05,
+      "loss": 0.6259,
+      "step": 10765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.720188761832932e-05,
+      "loss": 0.5717,
+      "step": 10766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7201393052837263e-05,
+      "loss": 0.6333,
+      "step": 10767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7200898450752816e-05,
+      "loss": 0.5449,
+      "step": 10768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7200403812078493e-05,
+      "loss": 0.6076,
+      "step": 10769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7199909136816816e-05,
+      "loss": 0.5621,
+      "step": 10770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7199414424970285e-05,
+      "loss": 0.6917,
+      "step": 10771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7198919676541425e-05,
+      "loss": 0.6144,
+      "step": 10772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7198424891532745e-05,
+      "loss": 0.6119,
+      "step": 10773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7197930069946757e-05,
+      "loss": 0.6144,
+      "step": 10774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7197435211785984e-05,
+      "loss": 0.5893,
+      "step": 10775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.719694031705293e-05,
+      "loss": 0.611,
+      "step": 10776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7196445385750117e-05,
+      "loss": 0.5911,
+      "step": 10777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7195950417880055e-05,
+      "loss": 0.5568,
+      "step": 10778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7195455413445264e-05,
+      "loss": 0.5968,
+      "step": 10779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.719496037244826e-05,
+      "loss": 0.6243,
+      "step": 10780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7194465294891546e-05,
+      "loss": 0.6143,
+      "step": 10781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7193970180777654e-05,
+      "loss": 0.5996,
+      "step": 10782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.719347503010909e-05,
+      "loss": 0.6265,
+      "step": 10783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7192979842888373e-05,
+      "loss": 0.5836,
+      "step": 10784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7192484619118016e-05,
+      "loss": 0.6147,
+      "step": 10785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7191989358800545e-05,
+      "loss": 0.582,
+      "step": 10786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7191494061938464e-05,
+      "loss": 0.6027,
+      "step": 10787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.71909987285343e-05,
+      "loss": 0.6042,
+      "step": 10788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7190503358590564e-05,
+      "loss": 0.6269,
+      "step": 10789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7190007952109772e-05,
+      "loss": 0.6131,
+      "step": 10790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7189512509094447e-05,
+      "loss": 0.6382,
+      "step": 10791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7189017029547102e-05,
+      "loss": 0.5936,
+      "step": 10792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7188521513470252e-05,
+      "loss": 0.6356,
+      "step": 10793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7188025960866427e-05,
+      "loss": 0.595,
+      "step": 10794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7187530371738132e-05,
+      "loss": 0.6503,
+      "step": 10795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.718703474608789e-05,
+      "loss": 0.6105,
+      "step": 10796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7186539083918223e-05,
+      "loss": 0.6076,
+      "step": 10797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7186043385231644e-05,
+      "loss": 0.6645,
+      "step": 10798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7185547650030673e-05,
+      "loss": 0.6228,
+      "step": 10799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.718505187831783e-05,
+      "loss": 0.6736,
+      "step": 10800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7184556070095637e-05,
+      "loss": 0.613,
+      "step": 10801
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.718406022536661e-05,
+      "loss": 0.6136,
+      "step": 10802
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7183564344133268e-05,
+      "loss": 0.6169,
+      "step": 10803
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7183068426398128e-05,
+      "loss": 0.5694,
+      "step": 10804
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7182572472163718e-05,
+      "loss": 0.7056,
+      "step": 10805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.718207648143255e-05,
+      "loss": 0.6243,
+      "step": 10806
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7181580454207158e-05,
+      "loss": 0.6431,
+      "step": 10807
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7181084390490042e-05,
+      "loss": 0.5852,
+      "step": 10808
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7180588290283737e-05,
+      "loss": 0.5784,
+      "step": 10809
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.718009215359076e-05,
+      "loss": 0.5725,
+      "step": 10810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7179595980413633e-05,
+      "loss": 0.5864,
+      "step": 10811
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7179099770754876e-05,
+      "loss": 0.6592,
+      "step": 10812
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7178603524617008e-05,
+      "loss": 0.6104,
+      "step": 10813
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7178107242002554e-05,
+      "loss": 0.6939,
+      "step": 10814
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7177610922914035e-05,
+      "loss": 0.5799,
+      "step": 10815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717711456735397e-05,
+      "loss": 0.5955,
+      "step": 10816
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717661817532489e-05,
+      "loss": 0.5738,
+      "step": 10817
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717612174682931e-05,
+      "loss": 0.5955,
+      "step": 10818
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7175625281869753e-05,
+      "loss": 0.5938,
+      "step": 10819
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717512878044874e-05,
+      "loss": 0.6229,
+      "step": 10820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7174632242568797e-05,
+      "loss": 0.6653,
+      "step": 10821
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717413566823245e-05,
+      "loss": 0.6269,
+      "step": 10822
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7173639057442215e-05,
+      "loss": 0.5828,
+      "step": 10823
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717314241020062e-05,
+      "loss": 0.6228,
+      "step": 10824
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717264572651019e-05,
+      "loss": 0.619,
+      "step": 10825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.717214900637344e-05,
+      "loss": 0.6161,
+      "step": 10826
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7171652249792905e-05,
+      "loss": 0.6199,
+      "step": 10827
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7171155456771105e-05,
+      "loss": 0.5849,
+      "step": 10828
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7170658627310564e-05,
+      "loss": 0.5538,
+      "step": 10829
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7170161761413807e-05,
+      "loss": 0.6514,
+      "step": 10830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7169664859083357e-05,
+      "loss": 0.6157,
+      "step": 10831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716916792032174e-05,
+      "loss": 0.616,
+      "step": 10832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716867094513148e-05,
+      "loss": 0.6851,
+      "step": 10833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7168173933515105e-05,
+      "loss": 0.5892,
+      "step": 10834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7167676885475137e-05,
+      "loss": 0.5647,
+      "step": 10835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7167179801014108e-05,
+      "loss": 0.614,
+      "step": 10836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7166682680134538e-05,
+      "loss": 0.6126,
+      "step": 10837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716618552283895e-05,
+      "loss": 0.6216,
+      "step": 10838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7165688329129876e-05,
+      "loss": 0.6183,
+      "step": 10839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7165191099009845e-05,
+      "loss": 0.6466,
+      "step": 10840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7164693832481378e-05,
+      "loss": 0.5823,
+      "step": 10841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7164196529547004e-05,
+      "loss": 0.6128,
+      "step": 10842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7163699190209248e-05,
+      "loss": 0.6097,
+      "step": 10843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7163201814470637e-05,
+      "loss": 0.5862,
+      "step": 10844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.71627044023337e-05,
+      "loss": 0.582,
+      "step": 10845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7162206953800965e-05,
+      "loss": 0.607,
+      "step": 10846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716170946887496e-05,
+      "loss": 0.6193,
+      "step": 10847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7161211947558212e-05,
+      "loss": 0.6671,
+      "step": 10848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.716071438985325e-05,
+      "loss": 0.611,
+      "step": 10849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7160216795762598e-05,
+      "loss": 0.6078,
+      "step": 10850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715971916528879e-05,
+      "loss": 0.6368,
+      "step": 10851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715922149843435e-05,
+      "loss": 0.5932,
+      "step": 10852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7158723795201807e-05,
+      "loss": 0.6751,
+      "step": 10853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7158226055593695e-05,
+      "loss": 0.6433,
+      "step": 10854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7157728279612542e-05,
+      "loss": 0.5382,
+      "step": 10855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715723046726087e-05,
+      "loss": 0.5641,
+      "step": 10856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7156732618541218e-05,
+      "loss": 0.5791,
+      "step": 10857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7156234733456112e-05,
+      "loss": 0.5688,
+      "step": 10858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7155736812008078e-05,
+      "loss": 0.6305,
+      "step": 10859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715523885419965e-05,
+      "loss": 0.7027,
+      "step": 10860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715474086003336e-05,
+      "loss": 0.6755,
+      "step": 10861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715424282951174e-05,
+      "loss": 0.6691,
+      "step": 10862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715374476263731e-05,
+      "loss": 0.5936,
+      "step": 10863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715324665941261e-05,
+      "loss": 0.5177,
+      "step": 10864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715274851984017e-05,
+      "loss": 0.6127,
+      "step": 10865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715225034392252e-05,
+      "loss": 0.6661,
+      "step": 10866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715175213166219e-05,
+      "loss": 0.6126,
+      "step": 10867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7151253883061715e-05,
+      "loss": 0.5652,
+      "step": 10868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.715075559812362e-05,
+      "loss": 0.6434,
+      "step": 10869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7150257276850445e-05,
+      "loss": 0.6078,
+      "step": 10870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714975891924472e-05,
+      "loss": 0.6549,
+      "step": 10871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7149260525308976e-05,
+      "loss": 0.6321,
+      "step": 10872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714876209504574e-05,
+      "loss": 0.5783,
+      "step": 10873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7148263628457557e-05,
+      "loss": 0.6533,
+      "step": 10874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714776512554695e-05,
+      "loss": 0.6096,
+      "step": 10875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7147266586316455e-05,
+      "loss": 0.6093,
+      "step": 10876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714676801076861e-05,
+      "loss": 0.5789,
+      "step": 10877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7146269398905935e-05,
+      "loss": 0.5835,
+      "step": 10878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714577075073098e-05,
+      "loss": 0.5876,
+      "step": 10879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7145272066246267e-05,
+      "loss": 0.6409,
+      "step": 10880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7144773345454332e-05,
+      "loss": 0.5834,
+      "step": 10881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7144274588357715e-05,
+      "loss": 0.6445,
+      "step": 10882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714377579495895e-05,
+      "loss": 0.6097,
+      "step": 10883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.714327696526056e-05,
+      "loss": 0.6161,
+      "step": 10884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7142778099265093e-05,
+      "loss": 0.6327,
+      "step": 10885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7142279196975075e-05,
+      "loss": 0.6006,
+      "step": 10886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7141780258393046e-05,
+      "loss": 0.6306,
+      "step": 10887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7141281283521544e-05,
+      "loss": 0.5954,
+      "step": 10888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7140782272363093e-05,
+      "loss": 0.6484,
+      "step": 10889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7140283224920244e-05,
+      "loss": 0.6329,
+      "step": 10890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7139784141195518e-05,
+      "loss": 0.5954,
+      "step": 10891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713928502119146e-05,
+      "loss": 0.629,
+      "step": 10892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7138785864910606e-05,
+      "loss": 0.6527,
+      "step": 10893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7138286672355485e-05,
+      "loss": 0.606,
+      "step": 10894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7137787443528645e-05,
+      "loss": 0.6423,
+      "step": 10895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7137288178432613e-05,
+      "loss": 0.622,
+      "step": 10896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713678887706993e-05,
+      "loss": 0.6186,
+      "step": 10897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7136289539443138e-05,
+      "loss": 0.6028,
+      "step": 10898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7135790165554764e-05,
+      "loss": 0.5763,
+      "step": 10899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713529075540735e-05,
+      "loss": 0.6269,
+      "step": 10900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7134791309003436e-05,
+      "loss": 0.6535,
+      "step": 10901
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7134291826345555e-05,
+      "loss": 0.5504,
+      "step": 10902
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713379230743625e-05,
+      "loss": 0.6093,
+      "step": 10903
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7133292752278054e-05,
+      "loss": 0.6018,
+      "step": 10904
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7132793160873512e-05,
+      "loss": 0.5949,
+      "step": 10905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713229353322516e-05,
+      "loss": 0.601,
+      "step": 10906
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713179386933553e-05,
+      "loss": 0.5923,
+      "step": 10907
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713129416920717e-05,
+      "loss": 0.6052,
+      "step": 10908
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713079443284262e-05,
+      "loss": 0.7037,
+      "step": 10909
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.713029466024441e-05,
+      "loss": 0.6833,
+      "step": 10910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7129794851415086e-05,
+      "loss": 0.6128,
+      "step": 10911
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7129295006357187e-05,
+      "loss": 0.595,
+      "step": 10912
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712879512507325e-05,
+      "loss": 0.6265,
+      "step": 10913
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712829520756582e-05,
+      "loss": 0.5923,
+      "step": 10914
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7127795253837433e-05,
+      "loss": 0.662,
+      "step": 10915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7127295263890632e-05,
+      "loss": 0.5656,
+      "step": 10916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7126795237727955e-05,
+      "loss": 0.6336,
+      "step": 10917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7126295175351943e-05,
+      "loss": 0.6058,
+      "step": 10918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712579507676514e-05,
+      "loss": 0.6087,
+      "step": 10919
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7125294941970084e-05,
+      "loss": 0.6232,
+      "step": 10920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712479477096932e-05,
+      "loss": 0.6045,
+      "step": 10921
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7124294563765385e-05,
+      "loss": 0.6151,
+      "step": 10922
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7123794320360824e-05,
+      "loss": 0.5523,
+      "step": 10923
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7123294040758178e-05,
+      "loss": 0.5611,
+      "step": 10924
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7122793724959988e-05,
+      "loss": 0.5825,
+      "step": 10925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7122293372968798e-05,
+      "loss": 0.6794,
+      "step": 10926
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712179298478715e-05,
+      "loss": 0.5201,
+      "step": 10927
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7121292560417587e-05,
+      "loss": 0.6203,
+      "step": 10928
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7120792099862645e-05,
+      "loss": 0.6058,
+      "step": 10929
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.712029160312488e-05,
+      "loss": 0.5977,
+      "step": 10930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711979107020682e-05,
+      "loss": 0.5856,
+      "step": 10931
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7119290501111024e-05,
+      "loss": 0.5685,
+      "step": 10932
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7118789895840025e-05,
+      "loss": 0.5606,
+      "step": 10933
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7118289254396373e-05,
+      "loss": 0.5793,
+      "step": 10934
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7117788576782604e-05,
+      "loss": 0.6132,
+      "step": 10935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7117287863001267e-05,
+      "loss": 0.6518,
+      "step": 10936
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711678711305491e-05,
+      "loss": 0.5728,
+      "step": 10937
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711628632694607e-05,
+      "loss": 0.6136,
+      "step": 10938
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7115785504677296e-05,
+      "loss": 0.653,
+      "step": 10939
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711528464625113e-05,
+      "loss": 0.6294,
+      "step": 10940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7114783751670123e-05,
+      "loss": 0.5587,
+      "step": 10941
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711428282093681e-05,
+      "loss": 0.5992,
+      "step": 10942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711378185405375e-05,
+      "loss": 0.6214,
+      "step": 10943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7113280851023478e-05,
+      "loss": 0.6126,
+      "step": 10944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711277981184854e-05,
+      "loss": 0.6111,
+      "step": 10945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.711227873653149e-05,
+      "loss": 0.5757,
+      "step": 10946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7111777625074863e-05,
+      "loss": 0.6072,
+      "step": 10947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7111276477481215e-05,
+      "loss": 0.6594,
+      "step": 10948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7110775293753087e-05,
+      "loss": 0.5587,
+      "step": 10949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7110274073893027e-05,
+      "loss": 0.5723,
+      "step": 10950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7109772817903586e-05,
+      "loss": 0.5912,
+      "step": 10951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7109271525787302e-05,
+      "loss": 0.5989,
+      "step": 10952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710877019754673e-05,
+      "loss": 0.6933,
+      "step": 10953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7108268833184412e-05,
+      "loss": 0.653,
+      "step": 10954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.71077674327029e-05,
+      "loss": 0.6343,
+      "step": 10955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710726599610474e-05,
+      "loss": 0.6195,
+      "step": 10956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710676452339248e-05,
+      "loss": 0.6529,
+      "step": 10957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710626301456867e-05,
+      "loss": 0.573,
+      "step": 10958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7105761469635854e-05,
+      "loss": 0.5779,
+      "step": 10959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7105259888596584e-05,
+      "loss": 0.5755,
+      "step": 10960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7104758271453407e-05,
+      "loss": 0.607,
+      "step": 10961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7104256618208875e-05,
+      "loss": 0.579,
+      "step": 10962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710375492886553e-05,
+      "loss": 0.6005,
+      "step": 10963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710325320342593e-05,
+      "loss": 0.5371,
+      "step": 10964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7102751441892617e-05,
+      "loss": 0.6609,
+      "step": 10965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7102249644268148e-05,
+      "loss": 0.6691,
+      "step": 10966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7101747810555065e-05,
+      "loss": 0.6129,
+      "step": 10967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710124594075592e-05,
+      "loss": 0.6027,
+      "step": 10968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.710074403487327e-05,
+      "loss": 0.6169,
+      "step": 10969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7100242092909652e-05,
+      "loss": 0.6085,
+      "step": 10970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709974011486763e-05,
+      "loss": 0.6583,
+      "step": 10971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7099238100749752e-05,
+      "loss": 0.6506,
+      "step": 10972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709873605055856e-05,
+      "loss": 0.6207,
+      "step": 10973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7098233964296617e-05,
+      "loss": 0.6229,
+      "step": 10974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7097731841966466e-05,
+      "loss": 0.6241,
+      "step": 10975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709722968357066e-05,
+      "loss": 0.6783,
+      "step": 10976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709672748911175e-05,
+      "loss": 0.5888,
+      "step": 10977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709622525859229e-05,
+      "loss": 0.6096,
+      "step": 10978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7095722992014833e-05,
+      "loss": 0.6353,
+      "step": 10979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709522068938193e-05,
+      "loss": 0.5777,
+      "step": 10980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709471835069613e-05,
+      "loss": 0.6232,
+      "step": 10981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709421597595999e-05,
+      "loss": 0.5969,
+      "step": 10982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7093713565176064e-05,
+      "loss": 0.5728,
+      "step": 10983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70932111183469e-05,
+      "loss": 0.6353,
+      "step": 10984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709270863547505e-05,
+      "loss": 0.5579,
+      "step": 10985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709220611656307e-05,
+      "loss": 0.5949,
+      "step": 10986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7091703561613516e-05,
+      "loss": 0.6159,
+      "step": 10987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709120097062894e-05,
+      "loss": 0.5982,
+      "step": 10988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.709069834361189e-05,
+      "loss": 0.596,
+      "step": 10989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7090195680564933e-05,
+      "loss": 0.6033,
+      "step": 10990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708969298149061e-05,
+      "loss": 0.6325,
+      "step": 10991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708919024639148e-05,
+      "loss": 0.6227,
+      "step": 10992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70886874752701e-05,
+      "loss": 0.5251,
+      "step": 10993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7088184668129024e-05,
+      "loss": 0.6105,
+      "step": 10994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7087681824970802e-05,
+      "loss": 0.5636,
+      "step": 10995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7087178945797996e-05,
+      "loss": 0.5655,
+      "step": 10996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7086676030613158e-05,
+      "loss": 0.6629,
+      "step": 10997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7086173079418845e-05,
+      "loss": 0.5915,
+      "step": 10998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708567009221761e-05,
+      "loss": 0.6631,
+      "step": 10999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708516706901201e-05,
+      "loss": 0.6247,
+      "step": 11000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70846640098046e-05,
+      "loss": 0.5908,
+      "step": 11001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708416091459794e-05,
+      "loss": 0.6322,
+      "step": 11002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7083657783394582e-05,
+      "loss": 0.5933,
+      "step": 11003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7083154616197082e-05,
+      "loss": 0.6297,
+      "step": 11004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7082651413008004e-05,
+      "loss": 0.6073,
+      "step": 11005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70821481738299e-05,
+      "loss": 0.6786,
+      "step": 11006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.708164489866532e-05,
+      "loss": 0.6129,
+      "step": 11007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7081141587516835e-05,
+      "loss": 0.6135,
+      "step": 11008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7080638240386993e-05,
+      "loss": 0.5938,
+      "step": 11009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7080134857278355e-05,
+      "loss": 0.5722,
+      "step": 11010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7079631438193477e-05,
+      "loss": 0.5664,
+      "step": 11011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.707912798313492e-05,
+      "loss": 0.6241,
+      "step": 11012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.707862449210524e-05,
+      "loss": 0.6332,
+      "step": 11013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7078120965106995e-05,
+      "loss": 0.5616,
+      "step": 11014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7077617402142746e-05,
+      "loss": 0.6513,
+      "step": 11015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7077113803215047e-05,
+      "loss": 0.6387,
+      "step": 11016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7076610168326466e-05,
+      "loss": 0.6423,
+      "step": 11017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7076106497479554e-05,
+      "loss": 0.613,
+      "step": 11018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.707560279067687e-05,
+      "loss": 0.6087,
+      "step": 11019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7075099047920976e-05,
+      "loss": 0.6307,
+      "step": 11020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7074595269214437e-05,
+      "loss": 0.6235,
+      "step": 11021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70740914545598e-05,
+      "loss": 0.604,
+      "step": 11022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7073587603959638e-05,
+      "loss": 0.5805,
+      "step": 11023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7073083717416504e-05,
+      "loss": 0.5956,
+      "step": 11024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7072579794932963e-05,
+      "loss": 0.6125,
+      "step": 11025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.707207583651157e-05,
+      "loss": 0.6045,
+      "step": 11026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7071571842154886e-05,
+      "loss": 0.6399,
+      "step": 11027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7071067811865477e-05,
+      "loss": 0.6277,
+      "step": 11028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70705637456459e-05,
+      "loss": 0.5982,
+      "step": 11029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.707005964349872e-05,
+      "loss": 0.5919,
+      "step": 11030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7069555505426495e-05,
+      "loss": 0.5835,
+      "step": 11031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7069051331431793e-05,
+      "loss": 0.6197,
+      "step": 11032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7068547121517166e-05,
+      "loss": 0.5984,
+      "step": 11033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.706804287568518e-05,
+      "loss": 0.5676,
+      "step": 11034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70675385939384e-05,
+      "loss": 0.6219,
+      "step": 11035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7067034276279387e-05,
+      "loss": 0.5641,
+      "step": 11036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7066529922710705e-05,
+      "loss": 0.6602,
+      "step": 11037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7066025533234912e-05,
+      "loss": 0.593,
+      "step": 11038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7065521107854574e-05,
+      "loss": 0.5988,
+      "step": 11039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7065016646572253e-05,
+      "loss": 0.5744,
+      "step": 11040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7064512149390515e-05,
+      "loss": 0.5917,
+      "step": 11041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7064007616311923e-05,
+      "loss": 0.5951,
+      "step": 11042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7063503047339037e-05,
+      "loss": 0.6266,
+      "step": 11043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7062998442474427e-05,
+      "loss": 0.5375,
+      "step": 11044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.706249380172065e-05,
+      "loss": 0.6137,
+      "step": 11045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7061989125080275e-05,
+      "loss": 0.6017,
+      "step": 11046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7061484412555863e-05,
+      "loss": 0.6764,
+      "step": 11047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7060979664149983e-05,
+      "loss": 0.5974,
+      "step": 11048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7060474879865197e-05,
+      "loss": 0.6052,
+      "step": 11049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705997005970407e-05,
+      "loss": 0.6322,
+      "step": 11050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7059465203669166e-05,
+      "loss": 0.6237,
+      "step": 11051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7058960311763054e-05,
+      "loss": 0.6174,
+      "step": 11052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70584553839883e-05,
+      "loss": 0.6097,
+      "step": 11053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705795042034746e-05,
+      "loss": 0.6128,
+      "step": 11054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7057445420843116e-05,
+      "loss": 0.6414,
+      "step": 11055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705694038547782e-05,
+      "loss": 0.5951,
+      "step": 11056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705643531425414e-05,
+      "loss": 0.5559,
+      "step": 11057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705593020717465e-05,
+      "loss": 0.6009,
+      "step": 11058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7055425064241907e-05,
+      "loss": 0.6403,
+      "step": 11059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705491988545849e-05,
+      "loss": 0.6283,
+      "step": 11060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7054414670826952e-05,
+      "loss": 0.5799,
+      "step": 11061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7053909420349876e-05,
+      "loss": 0.5857,
+      "step": 11062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7053404134029814e-05,
+      "loss": 0.5983,
+      "step": 11063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705289881186934e-05,
+      "loss": 0.5434,
+      "step": 11064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7052393453871025e-05,
+      "loss": 0.6727,
+      "step": 11065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705188806003743e-05,
+      "loss": 0.5811,
+      "step": 11066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7051382630371126e-05,
+      "loss": 0.5842,
+      "step": 11067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7050877164874685e-05,
+      "loss": 0.6302,
+      "step": 11068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.705037166355067e-05,
+      "loss": 0.5561,
+      "step": 11069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7049866126401654e-05,
+      "loss": 0.5628,
+      "step": 11070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70493605534302e-05,
+      "loss": 0.6461,
+      "step": 11071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704885494463888e-05,
+      "loss": 0.6664,
+      "step": 11072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704834930003027e-05,
+      "loss": 0.5866,
+      "step": 11073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7047843619606926e-05,
+      "loss": 0.5896,
+      "step": 11074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704733790337143e-05,
+      "loss": 0.6108,
+      "step": 11075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704683215132634e-05,
+      "loss": 0.6074,
+      "step": 11076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704632636347424e-05,
+      "loss": 0.6344,
+      "step": 11077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704582053981769e-05,
+      "loss": 0.6186,
+      "step": 11078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7045314680359258e-05,
+      "loss": 0.6399,
+      "step": 11079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7044808785101522e-05,
+      "loss": 0.5614,
+      "step": 11080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704430285404705e-05,
+      "loss": 0.6189,
+      "step": 11081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7043796887198406e-05,
+      "loss": 0.5927,
+      "step": 11082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7043290884558173e-05,
+      "loss": 0.6258,
+      "step": 11083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7042784846128918e-05,
+      "loss": 0.6194,
+      "step": 11084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7042278771913204e-05,
+      "loss": 0.6644,
+      "step": 11085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7041772661913616e-05,
+      "loss": 0.6889,
+      "step": 11086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7041266516132716e-05,
+      "loss": 0.5952,
+      "step": 11087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.704076033457308e-05,
+      "loss": 0.6169,
+      "step": 11088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7040254117237276e-05,
+      "loss": 0.5637,
+      "step": 11089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.703974786412788e-05,
+      "loss": 0.6152,
+      "step": 11090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7039241575247464e-05,
+      "loss": 0.6017,
+      "step": 11091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.70387352505986e-05,
+      "loss": 0.5935,
+      "step": 11092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.703822889018386e-05,
+      "loss": 0.6185,
+      "step": 11093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.7037722494005818e-05,
+      "loss": 0.6362,
+      "step": 11094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7037216062067048e-05,
+      "loss": 0.5443,
+      "step": 11095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703670959437012e-05,
+      "loss": 0.567,
+      "step": 11096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7036203090917612e-05,
+      "loss": 0.6457,
+      "step": 11097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7035696551712094e-05,
+      "loss": 0.5694,
+      "step": 11098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703518997675614e-05,
+      "loss": 0.6399,
+      "step": 11099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703468336605233e-05,
+      "loss": 0.5768,
+      "step": 11100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7034176719603232e-05,
+      "loss": 0.6165,
+      "step": 11101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703367003741142e-05,
+      "loss": 0.6261,
+      "step": 11102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7033163319479473e-05,
+      "loss": 0.5926,
+      "step": 11103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703265656580996e-05,
+      "loss": 0.6327,
+      "step": 11104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703214977640546e-05,
+      "loss": 0.5901,
+      "step": 11105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7031642951268554e-05,
+      "loss": 0.6315,
+      "step": 11106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7031136090401802e-05,
+      "loss": 0.5413,
+      "step": 11107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.703062919380779e-05,
+      "loss": 0.6021,
+      "step": 11108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7030122261489098e-05,
+      "loss": 0.6283,
+      "step": 11109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7029615293448288e-05,
+      "loss": 0.641,
+      "step": 11110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7029108289687953e-05,
+      "loss": 0.554,
+      "step": 11111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7028601250210654e-05,
+      "loss": 0.5776,
+      "step": 11112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7028094175018973e-05,
+      "loss": 0.6211,
+      "step": 11113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7027587064115492e-05,
+      "loss": 0.625,
+      "step": 11114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.702707991750278e-05,
+      "loss": 0.6664,
+      "step": 11115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7026572735183416e-05,
+      "loss": 0.5978,
+      "step": 11116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7026065517159978e-05,
+      "loss": 0.5966,
+      "step": 11117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7025558263435046e-05,
+      "loss": 0.6469,
+      "step": 11118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7025050974011193e-05,
+      "loss": 0.6056,
+      "step": 11119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7024543648891e-05,
+      "loss": 0.6221,
+      "step": 11120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.702403628807704e-05,
+      "loss": 0.6161,
+      "step": 11121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.70235288915719e-05,
+      "loss": 0.6223,
+      "step": 11122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7023021459378148e-05,
+      "loss": 0.6212,
+      "step": 11123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7022513991498368e-05,
+      "loss": 0.6071,
+      "step": 11124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7022006487935142e-05,
+      "loss": 0.7007,
+      "step": 11125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.702149894869104e-05,
+      "loss": 0.6021,
+      "step": 11126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7020991373768647e-05,
+      "loss": 0.5956,
+      "step": 11127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.702048376317054e-05,
+      "loss": 0.6071,
+      "step": 11128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.70199761168993e-05,
+      "loss": 0.5085,
+      "step": 11129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7019468434957508e-05,
+      "loss": 0.6092,
+      "step": 11130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7018960717347736e-05,
+      "loss": 0.5735,
+      "step": 11131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.701845296407257e-05,
+      "loss": 0.5799,
+      "step": 11132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7017945175134594e-05,
+      "loss": 0.6627,
+      "step": 11133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7017437350536378e-05,
+      "loss": 0.6246,
+      "step": 11134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.701692949028051e-05,
+      "loss": 0.5942,
+      "step": 11135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7016421594369568e-05,
+      "loss": 0.6154,
+      "step": 11136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.701591366280613e-05,
+      "loss": 0.6135,
+      "step": 11137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7015405695592784e-05,
+      "loss": 0.593,
+      "step": 11138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7014897692732103e-05,
+      "loss": 0.6314,
+      "step": 11139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7014389654226677e-05,
+      "loss": 0.6216,
+      "step": 11140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.701388158007908e-05,
+      "loss": 0.5937,
+      "step": 11141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7013373470291898e-05,
+      "loss": 0.6223,
+      "step": 11142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7012865324867712e-05,
+      "loss": 0.6161,
+      "step": 11143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.70123571438091e-05,
+      "loss": 0.5572,
+      "step": 11144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7011848927118652e-05,
+      "loss": 0.6446,
+      "step": 11145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7011340674798945e-05,
+      "loss": 0.6212,
+      "step": 11146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.701083238685256e-05,
+      "loss": 0.5995,
+      "step": 11147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7010324063282083e-05,
+      "loss": 0.647,
+      "step": 11148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7009815704090102e-05,
+      "loss": 0.5629,
+      "step": 11149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7009307309279186e-05,
+      "loss": 0.6253,
+      "step": 11150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700879887885193e-05,
+      "loss": 0.6108,
+      "step": 11151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7008290412810916e-05,
+      "loss": 0.6319,
+      "step": 11152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7007781911158727e-05,
+      "loss": 0.6131,
+      "step": 11153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7007273373897943e-05,
+      "loss": 0.5966,
+      "step": 11154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7006764801031153e-05,
+      "loss": 0.5999,
+      "step": 11155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7006256192560936e-05,
+      "loss": 0.588,
+      "step": 11156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700574754848988e-05,
+      "loss": 0.5956,
+      "step": 11157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700523886882057e-05,
+      "loss": 0.6364,
+      "step": 11158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700473015355559e-05,
+      "loss": 0.586,
+      "step": 11159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7004221402697525e-05,
+      "loss": 0.6285,
+      "step": 11160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700371261624896e-05,
+      "loss": 0.5985,
+      "step": 11161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700320379421248e-05,
+      "loss": 0.6506,
+      "step": 11162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700269493659067e-05,
+      "loss": 0.6199,
+      "step": 11163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700218604338611e-05,
+      "loss": 0.5765,
+      "step": 11164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.7001677114601404e-05,
+      "loss": 0.6673,
+      "step": 11165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700116815023912e-05,
+      "loss": 0.5605,
+      "step": 11166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700065915030185e-05,
+      "loss": 0.6297,
+      "step": 11167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.700015011479218e-05,
+      "loss": 0.5934,
+      "step": 11168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.69996410437127e-05,
+      "loss": 0.5812,
+      "step": 11169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699913193706599e-05,
+      "loss": 0.5867,
+      "step": 11170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699862279485464e-05,
+      "loss": 0.6078,
+      "step": 11171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6998113617081244e-05,
+      "loss": 0.5795,
+      "step": 11172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699760440374838e-05,
+      "loss": 0.5876,
+      "step": 11173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699709515485864e-05,
+      "loss": 0.6012,
+      "step": 11174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699658587041461e-05,
+      "loss": 0.5742,
+      "step": 11175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6996076550418874e-05,
+      "loss": 0.6333,
+      "step": 11176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6995567194874028e-05,
+      "loss": 0.6069,
+      "step": 11177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6995057803782654e-05,
+      "loss": 0.6508,
+      "step": 11178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6994548377147345e-05,
+      "loss": 0.6476,
+      "step": 11179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6994038914970686e-05,
+      "loss": 0.5874,
+      "step": 11180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699352941725527e-05,
+      "loss": 0.654,
+      "step": 11181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6993019884003682e-05,
+      "loss": 0.5645,
+      "step": 11182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6992510315218508e-05,
+      "loss": 0.6594,
+      "step": 11183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6992000710902345e-05,
+      "loss": 0.6248,
+      "step": 11184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6991491071057774e-05,
+      "loss": 0.5782,
+      "step": 11185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6990981395687394e-05,
+      "loss": 0.5974,
+      "step": 11186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.699047168479379e-05,
+      "loss": 0.6088,
+      "step": 11187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.698996193837955e-05,
+      "loss": 0.6134,
+      "step": 11188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.698945215644727e-05,
+      "loss": 0.6601,
+      "step": 11189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.698894233899953e-05,
+      "loss": 0.6373,
+      "step": 11190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6988432486038932e-05,
+      "loss": 0.5965,
+      "step": 11191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6987922597568064e-05,
+      "loss": 0.6227,
+      "step": 11192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.698741267358951e-05,
+      "loss": 0.6211,
+      "step": 11193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6986902714105865e-05,
+      "loss": 0.5994,
+      "step": 11194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6986392719119726e-05,
+      "loss": 0.6132,
+      "step": 11195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6985882688633677e-05,
+      "loss": 0.7319,
+      "step": 11196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6985372622650313e-05,
+      "loss": 0.6285,
+      "step": 11197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6984862521172223e-05,
+      "loss": 0.6135,
+      "step": 11198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6984352384202003e-05,
+      "loss": 0.5561,
+      "step": 11199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6983842211742245e-05,
+      "loss": 0.5581,
+      "step": 11200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6983332003795536e-05,
+      "loss": 0.6309,
+      "step": 11201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.698282176036447e-05,
+      "loss": 0.6721,
+      "step": 11202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6982311481451645e-05,
+      "loss": 0.6205,
+      "step": 11203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6981801167059648e-05,
+      "loss": 0.6104,
+      "step": 11204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6981290817191077e-05,
+      "loss": 0.6347,
+      "step": 11205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6980780431848524e-05,
+      "loss": 0.6077,
+      "step": 11206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6980270011034578e-05,
+      "loss": 0.598,
+      "step": 11207
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6979759554751836e-05,
+      "loss": 0.5863,
+      "step": 11208
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6979249063002892e-05,
+      "loss": 0.6082,
+      "step": 11209
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6978738535790343e-05,
+      "loss": 0.6082,
+      "step": 11210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6978227973116776e-05,
+      "loss": 0.5528,
+      "step": 11211
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6977717374984787e-05,
+      "loss": 0.6234,
+      "step": 11212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6977206741396975e-05,
+      "loss": 0.6225,
+      "step": 11213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6976696072355935e-05,
+      "loss": 0.5881,
+      "step": 11214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6976185367864255e-05,
+      "loss": 0.5817,
+      "step": 11215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6975674627924538e-05,
+      "loss": 0.5684,
+      "step": 11216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.697516385253937e-05,
+      "loss": 0.6147,
+      "step": 11217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6974653041711358e-05,
+      "loss": 0.6834,
+      "step": 11218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6974142195443087e-05,
+      "loss": 0.6393,
+      "step": 11219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.697363131373716e-05,
+      "loss": 0.593,
+      "step": 11220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6973120396596164e-05,
+      "loss": 0.6221,
+      "step": 11221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6972609444022706e-05,
+      "loss": 0.6376,
+      "step": 11222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.697209845601938e-05,
+      "loss": 0.5885,
+      "step": 11223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6971587432588772e-05,
+      "loss": 0.5735,
+      "step": 11224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6971076373733493e-05,
+      "loss": 0.6345,
+      "step": 11225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.697056527945613e-05,
+      "loss": 0.6176,
+      "step": 11226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6970054149759284e-05,
+      "loss": 0.6751,
+      "step": 11227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6969542984645554e-05,
+      "loss": 0.6063,
+      "step": 11228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6969031784117533e-05,
+      "loss": 0.644,
+      "step": 11229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.696852054817782e-05,
+      "loss": 0.5645,
+      "step": 11230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6968009276829014e-05,
+      "loss": 0.6843,
+      "step": 11231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.696749797007371e-05,
+      "loss": 0.6213,
+      "step": 11232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6966986627914513e-05,
+      "loss": 0.6621,
+      "step": 11233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6966475250354014e-05,
+      "loss": 0.5729,
+      "step": 11234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6965963837394812e-05,
+      "loss": 0.6159,
+      "step": 11235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.696545238903951e-05,
+      "loss": 0.6104,
+      "step": 11236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6964940905290705e-05,
+      "loss": 0.5829,
+      "step": 11237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6964429386150993e-05,
+      "loss": 0.6016,
+      "step": 11238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.696391783162298e-05,
+      "loss": 0.5956,
+      "step": 11239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.6963406241709258e-05,
+      "loss": 0.6347,
+      "step": 11240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.696289461641243e-05,
+      "loss": 0.6116,
+      "step": 11241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6962382955735094e-05,
+      "loss": 0.5749,
+      "step": 11242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6961871259679853e-05,
+      "loss": 0.6262,
+      "step": 11243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6961359528249303e-05,
+      "loss": 0.624,
+      "step": 11244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.696084776144605e-05,
+      "loss": 0.6263,
+      "step": 11245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.696033595927269e-05,
+      "loss": 0.5982,
+      "step": 11246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6959824121731823e-05,
+      "loss": 0.5726,
+      "step": 11247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695931224882605e-05,
+      "loss": 0.6047,
+      "step": 11248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6958800340557976e-05,
+      "loss": 0.5448,
+      "step": 11249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6958288396930197e-05,
+      "loss": 0.6305,
+      "step": 11250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695777641794532e-05,
+      "loss": 0.6484,
+      "step": 11251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6957264403605944e-05,
+      "loss": 0.6418,
+      "step": 11252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6956752353914664e-05,
+      "loss": 0.6014,
+      "step": 11253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6956240268874093e-05,
+      "loss": 0.6606,
+      "step": 11254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6955728148486828e-05,
+      "loss": 0.6693,
+      "step": 11255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695521599275547e-05,
+      "loss": 0.5824,
+      "step": 11256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695470380168262e-05,
+      "loss": 0.6215,
+      "step": 11257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6954191575270886e-05,
+      "loss": 0.6151,
+      "step": 11258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695367931352287e-05,
+      "loss": 0.6651,
+      "step": 11259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6953167016441172e-05,
+      "loss": 0.592,
+      "step": 11260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6952654684028394e-05,
+      "loss": 0.6386,
+      "step": 11261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6952142316287142e-05,
+      "loss": 0.5777,
+      "step": 11262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695162991322002e-05,
+      "loss": 0.6542,
+      "step": 11263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695111747482963e-05,
+      "loss": 0.5388,
+      "step": 11264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6950605001118575e-05,
+      "loss": 0.6151,
+      "step": 11265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.695009249208946e-05,
+      "loss": 0.6265,
+      "step": 11266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694957994774489e-05,
+      "loss": 0.613,
+      "step": 11267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694906736808747e-05,
+      "loss": 0.5861,
+      "step": 11268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6948554753119806e-05,
+      "loss": 0.5907,
+      "step": 11269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6948042102844497e-05,
+      "loss": 0.6592,
+      "step": 11270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6947529417264153e-05,
+      "loss": 0.6109,
+      "step": 11271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6947016696381374e-05,
+      "loss": 0.601,
+      "step": 11272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6946503940198775e-05,
+      "loss": 0.6046,
+      "step": 11273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694599114871895e-05,
+      "loss": 0.6255,
+      "step": 11274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694547832194451e-05,
+      "loss": 0.6142,
+      "step": 11275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6944965459878062e-05,
+      "loss": 0.6028,
+      "step": 11276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694445256252221e-05,
+      "loss": 0.5953,
+      "step": 11277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694393962987956e-05,
+      "loss": 0.5962,
+      "step": 11278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6943426661952723e-05,
+      "loss": 0.5835,
+      "step": 11279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6942913658744295e-05,
+      "loss": 0.6163,
+      "step": 11280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6942400620256893e-05,
+      "loss": 0.5936,
+      "step": 11281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694188754649312e-05,
+      "loss": 0.5999,
+      "step": 11282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6941374437455586e-05,
+      "loss": 0.6492,
+      "step": 11283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6940861293146894e-05,
+      "loss": 0.568,
+      "step": 11284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.694034811356965e-05,
+      "loss": 0.6634,
+      "step": 11285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6939834898726466e-05,
+      "loss": 0.6393,
+      "step": 11286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693932164861995e-05,
+      "loss": 0.6479,
+      "step": 11287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6938808363252712e-05,
+      "loss": 0.6565,
+      "step": 11288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693829504262735e-05,
+      "loss": 0.6211,
+      "step": 11289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693778168674648e-05,
+      "loss": 0.6195,
+      "step": 11290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6937268295612713e-05,
+      "loss": 0.6163,
+      "step": 11291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693675486922865e-05,
+      "loss": 0.6203,
+      "step": 11292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6936241407596904e-05,
+      "loss": 0.5576,
+      "step": 11293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6935727910720086e-05,
+      "loss": 0.6051,
+      "step": 11294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6935214378600805e-05,
+      "loss": 0.6195,
+      "step": 11295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6934700811241667e-05,
+      "loss": 0.536,
+      "step": 11296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693418720864528e-05,
+      "loss": 0.5153,
+      "step": 11297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6933673570814263e-05,
+      "loss": 0.6111,
+      "step": 11298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6933159897751217e-05,
+      "loss": 0.6254,
+      "step": 11299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6932646189458756e-05,
+      "loss": 0.5844,
+      "step": 11300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.693213244593949e-05,
+      "loss": 0.5871,
+      "step": 11301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6931618667196023e-05,
+      "loss": 0.634,
+      "step": 11302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6931104853230978e-05,
+      "loss": 0.6627,
+      "step": 11303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6930591004046956e-05,
+      "loss": 0.6078,
+      "step": 11304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6930077119646575e-05,
+      "loss": 0.5889,
+      "step": 11305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6929563200032438e-05,
+      "loss": 0.5993,
+      "step": 11306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6929049245207165e-05,
+      "loss": 0.5981,
+      "step": 11307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.692853525517336e-05,
+      "loss": 0.6297,
+      "step": 11308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.692802122993364e-05,
+      "loss": 0.6363,
+      "step": 11309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6927507169490618e-05,
+      "loss": 0.6572,
+      "step": 11310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.69269930738469e-05,
+      "loss": 0.6732,
+      "step": 11311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6926478943005102e-05,
+      "loss": 0.5806,
+      "step": 11312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6925964776967834e-05,
+      "loss": 0.6016,
+      "step": 11313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6925450575737716e-05,
+      "loss": 0.6212,
+      "step": 11314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.692493633931735e-05,
+      "loss": 0.6177,
+      "step": 11315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6924422067709354e-05,
+      "loss": 0.6527,
+      "step": 11316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6923907760916346e-05,
+      "loss": 0.6134,
+      "step": 11317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6923393418940933e-05,
+      "loss": 0.6137,
+      "step": 11318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.692287904178573e-05,
+      "loss": 0.605,
+      "step": 11319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6922364629453352e-05,
+      "loss": 0.582,
+      "step": 11320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.692185018194641e-05,
+      "loss": 0.6079,
+      "step": 11321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6921335699267524e-05,
+      "loss": 0.5916,
+      "step": 11322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.69208211814193e-05,
+      "loss": 0.6585,
+      "step": 11323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6920306628404356e-05,
+      "loss": 0.6203,
+      "step": 11324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6919792040225315e-05,
+      "loss": 0.6275,
+      "step": 11325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6919277416884778e-05,
+      "loss": 0.6174,
+      "step": 11326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6918762758385366e-05,
+      "loss": 0.624,
+      "step": 11327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6918248064729693e-05,
+      "loss": 0.6176,
+      "step": 11328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6917733335920378e-05,
+      "loss": 0.5665,
+      "step": 11329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6917218571960035e-05,
+      "loss": 0.6157,
+      "step": 11330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6916703772851274e-05,
+      "loss": 0.627,
+      "step": 11331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.691618893859672e-05,
+      "loss": 0.6538,
+      "step": 11332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6915674069198984e-05,
+      "loss": 0.6338,
+      "step": 11333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6915159164660682e-05,
+      "loss": 0.5984,
+      "step": 11334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6914644224984432e-05,
+      "loss": 0.5693,
+      "step": 11335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.691412925017285e-05,
+      "loss": 0.6196,
+      "step": 11336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6913614240228547e-05,
+      "loss": 0.6115,
+      "step": 11337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.691309919515415e-05,
+      "loss": 0.6341,
+      "step": 11338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6912584114952272e-05,
+      "loss": 0.5814,
+      "step": 11339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.691206899962553e-05,
+      "loss": 0.6412,
+      "step": 11340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6911553849176538e-05,
+      "loss": 0.5894,
+      "step": 11341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6911038663607917e-05,
+      "loss": 0.6784,
+      "step": 11342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6910523442922286e-05,
+      "loss": 0.5558,
+      "step": 11343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.691000818712226e-05,
+      "loss": 0.6314,
+      "step": 11344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6909492896210463e-05,
+      "loss": 0.6534,
+      "step": 11345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6908977570189506e-05,
+      "loss": 0.6058,
+      "step": 11346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690846220906201e-05,
+      "loss": 0.6255,
+      "step": 11347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6907946812830594e-05,
+      "loss": 0.5997,
+      "step": 11348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690743138149788e-05,
+      "loss": 0.6045,
+      "step": 11349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690691591506648e-05,
+      "loss": 0.6871,
+      "step": 11350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690640041353902e-05,
+      "loss": 0.6423,
+      "step": 11351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690588487691812e-05,
+      "loss": 0.5987,
+      "step": 11352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6905369305206392e-05,
+      "loss": 0.6562,
+      "step": 11353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690485369840646e-05,
+      "loss": 0.6331,
+      "step": 11354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6904338056520947e-05,
+      "loss": 0.5809,
+      "step": 11355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6903822379552468e-05,
+      "loss": 0.57,
+      "step": 11356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6903306667503648e-05,
+      "loss": 0.633,
+      "step": 11357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6902790920377103e-05,
+      "loss": 0.6515,
+      "step": 11358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6902275138175455e-05,
+      "loss": 0.6887,
+      "step": 11359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6901759320901328e-05,
+      "loss": 0.627,
+      "step": 11360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6901243468557336e-05,
+      "loss": 0.5872,
+      "step": 11361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.690072758114611e-05,
+      "loss": 0.6046,
+      "step": 11362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6900211658670267e-05,
+      "loss": 0.6006,
+      "step": 11363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689969570113242e-05,
+      "loss": 0.6451,
+      "step": 11364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689917970853521e-05,
+      "loss": 0.5711,
+      "step": 11365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6898663680881238e-05,
+      "loss": 0.645,
+      "step": 11366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689814761817314e-05,
+      "loss": 0.6608,
+      "step": 11367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689763152041353e-05,
+      "loss": 0.6057,
+      "step": 11368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689711538760504e-05,
+      "loss": 0.6169,
+      "step": 11369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689659921975028e-05,
+      "loss": 0.5929,
+      "step": 11370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6896083016851882e-05,
+      "loss": 0.605,
+      "step": 11371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689556677891247e-05,
+      "loss": 0.6581,
+      "step": 11372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689505050593466e-05,
+      "loss": 0.5606,
+      "step": 11373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689453419792108e-05,
+      "loss": 0.5697,
+      "step": 11374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6894017854874352e-05,
+      "loss": 0.5776,
+      "step": 11375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6893501476797102e-05,
+      "loss": 0.6009,
+      "step": 11376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6892985063691953e-05,
+      "loss": 0.6428,
+      "step": 11377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6892468615561527e-05,
+      "loss": 0.6327,
+      "step": 11378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689195213240845e-05,
+      "loss": 0.6362,
+      "step": 11379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6891435614235342e-05,
+      "loss": 0.5861,
+      "step": 11380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6890919061044834e-05,
+      "loss": 0.5795,
+      "step": 11381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.689040247283955e-05,
+      "loss": 0.6472,
+      "step": 11382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6889885849622108e-05,
+      "loss": 0.5975,
+      "step": 11383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6889369191395144e-05,
+      "loss": 0.6352,
+      "step": 11384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.6888852498161275e-05,
+      "loss": 0.5655,
+      "step": 11385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.688833576992313e-05,
+      "loss": 0.6271,
+      "step": 11386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.688781900668333e-05,
+      "loss": 0.6409,
+      "step": 11387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.688730220844451e-05,
+      "loss": 0.5936,
+      "step": 11388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6886785375209285e-05,
+      "loss": 0.5899,
+      "step": 11389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.688626850698029e-05,
+      "loss": 0.5863,
+      "step": 11390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6885751603760144e-05,
+      "loss": 0.6285,
+      "step": 11391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6885234665551485e-05,
+      "loss": 0.5765,
+      "step": 11392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6884717692356926e-05,
+      "loss": 0.5887,
+      "step": 11393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68842006841791e-05,
+      "loss": 0.6384,
+      "step": 11394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.688368364102064e-05,
+      "loss": 0.5575,
+      "step": 11395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6883166562884163e-05,
+      "loss": 0.5971,
+      "step": 11396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68826494497723e-05,
+      "loss": 0.64,
+      "step": 11397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.688213230168768e-05,
+      "loss": 0.6033,
+      "step": 11398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6881615118632933e-05,
+      "loss": 0.664,
+      "step": 11399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.688109790061068e-05,
+      "loss": 0.5778,
+      "step": 11400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6880580647623557e-05,
+      "loss": 0.5968,
+      "step": 11401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6880063359674188e-05,
+      "loss": 0.6433,
+      "step": 11402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68795460367652e-05,
+      "loss": 0.6261,
+      "step": 11403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6879028678899227e-05,
+      "loss": 0.6597,
+      "step": 11404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.687851128607889e-05,
+      "loss": 0.5953,
+      "step": 11405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6877993858306828e-05,
+      "loss": 0.5854,
+      "step": 11406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.687747639558566e-05,
+      "loss": 0.5918,
+      "step": 11407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6876958897918024e-05,
+      "loss": 0.595,
+      "step": 11408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6876441365306545e-05,
+      "loss": 0.5759,
+      "step": 11409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6875923797753856e-05,
+      "loss": 0.591,
+      "step": 11410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.687540619526258e-05,
+      "loss": 0.6087,
+      "step": 11411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6874888557835354e-05,
+      "loss": 0.6291,
+      "step": 11412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6874370885474803e-05,
+      "loss": 0.6823,
+      "step": 11413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6873853178183564e-05,
+      "loss": 0.5739,
+      "step": 11414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6873335435964262e-05,
+      "loss": 0.6229,
+      "step": 11415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6872817658819528e-05,
+      "loss": 0.62,
+      "step": 11416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6872299846752e-05,
+      "loss": 0.6108,
+      "step": 11417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6871781999764302e-05,
+      "loss": 0.6045,
+      "step": 11418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6871264117859065e-05,
+      "loss": 0.6332,
+      "step": 11419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.687074620103892e-05,
+      "loss": 0.6318,
+      "step": 11420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6870228249306502e-05,
+      "loss": 0.6116,
+      "step": 11421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.686971026266445e-05,
+      "loss": 0.5942,
+      "step": 11422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.686919224111538e-05,
+      "loss": 0.6342,
+      "step": 11423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6868674184661936e-05,
+      "loss": 0.5657,
+      "step": 11424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6868156093306746e-05,
+      "loss": 0.5683,
+      "step": 11425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6867637967052446e-05,
+      "loss": 0.6091,
+      "step": 11426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6867119805901663e-05,
+      "loss": 0.5788,
+      "step": 11427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6866601609857035e-05,
+      "loss": 0.6082,
+      "step": 11428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6866083378921193e-05,
+      "loss": 0.5953,
+      "step": 11429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6865565113096772e-05,
+      "loss": 0.6225,
+      "step": 11430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68650468123864e-05,
+      "loss": 0.5569,
+      "step": 11431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6864528476792716e-05,
+      "loss": 0.6421,
+      "step": 11432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6864010106318356e-05,
+      "loss": 0.6308,
+      "step": 11433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6863491700965945e-05,
+      "loss": 0.5702,
+      "step": 11434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6862973260738128e-05,
+      "loss": 0.6001,
+      "step": 11435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.686245478563753e-05,
+      "loss": 0.6158,
+      "step": 11436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6861936275666792e-05,
+      "loss": 0.6221,
+      "step": 11437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6861417730828545e-05,
+      "loss": 0.5488,
+      "step": 11438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6860899151125423e-05,
+      "loss": 0.5656,
+      "step": 11439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6860380536560065e-05,
+      "loss": 0.6542,
+      "step": 11440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6859861887135106e-05,
+      "loss": 0.5811,
+      "step": 11441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6859343202853177e-05,
+      "loss": 0.5872,
+      "step": 11442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6858824483716915e-05,
+      "loss": 0.5742,
+      "step": 11443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.685830572972896e-05,
+      "loss": 0.5631,
+      "step": 11444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6857786940891948e-05,
+      "loss": 0.6177,
+      "step": 11445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6857268117208504e-05,
+      "loss": 0.5876,
+      "step": 11446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.685674925868128e-05,
+      "loss": 0.581,
+      "step": 11447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68562303653129e-05,
+      "loss": 0.5987,
+      "step": 11448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6855711437106008e-05,
+      "loss": 0.6164,
+      "step": 11449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6855192474063238e-05,
+      "loss": 0.5849,
+      "step": 11450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6854673476187228e-05,
+      "loss": 0.6115,
+      "step": 11451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6854154443480616e-05,
+      "loss": 0.5589,
+      "step": 11452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6853635375946032e-05,
+      "loss": 0.5576,
+      "step": 11453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6853116273586126e-05,
+      "loss": 0.587,
+      "step": 11454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6852597136403526e-05,
+      "loss": 0.6411,
+      "step": 11455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6852077964400876e-05,
+      "loss": 0.6125,
+      "step": 11456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6851558757580808e-05,
+      "loss": 0.6166,
+      "step": 11457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6851039515945964e-05,
+      "loss": 0.6625,
+      "step": 11458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.685052023949898e-05,
+      "loss": 0.5119,
+      "step": 11459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6850000928242502e-05,
+      "loss": 0.5994,
+      "step": 11460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6849481582179158e-05,
+      "loss": 0.6391,
+      "step": 11461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6848962201311593e-05,
+      "loss": 0.6738,
+      "step": 11462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6848442785642447e-05,
+      "loss": 0.6484,
+      "step": 11463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6847923335174354e-05,
+      "loss": 0.627,
+      "step": 11464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684740384990996e-05,
+      "loss": 0.6034,
+      "step": 11465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68468843298519e-05,
+      "loss": 0.6301,
+      "step": 11466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684636477500282e-05,
+      "loss": 0.6359,
+      "step": 11467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684584518536535e-05,
+      "loss": 0.5855,
+      "step": 11468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684532556094214e-05,
+      "loss": 0.5819,
+      "step": 11469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6844805901735824e-05,
+      "loss": 0.5928,
+      "step": 11470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6844286207749043e-05,
+      "loss": 0.5964,
+      "step": 11471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6843766478984443e-05,
+      "loss": 0.5702,
+      "step": 11472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6843246715444655e-05,
+      "loss": 0.5403,
+      "step": 11473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6842726917132332e-05,
+      "loss": 0.6337,
+      "step": 11474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684220708405011e-05,
+      "loss": 0.5685,
+      "step": 11475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684168721620063e-05,
+      "loss": 0.6345,
+      "step": 11476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684116731358653e-05,
+      "loss": 0.65,
+      "step": 11477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6840647376210454e-05,
+      "loss": 0.644,
+      "step": 11478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.684012740407505e-05,
+      "loss": 0.6028,
+      "step": 11479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6839607397182954e-05,
+      "loss": 0.6209,
+      "step": 11480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6839087355536808e-05,
+      "loss": 0.6335,
+      "step": 11481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683856727913926e-05,
+      "loss": 0.569,
+      "step": 11482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6838047167992945e-05,
+      "loss": 0.6058,
+      "step": 11483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6837527022100512e-05,
+      "loss": 0.619,
+      "step": 11484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68370068414646e-05,
+      "loss": 0.626,
+      "step": 11485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6836486626087854e-05,
+      "loss": 0.5916,
+      "step": 11486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683596637597292e-05,
+      "loss": 0.5716,
+      "step": 11487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683544609112244e-05,
+      "loss": 0.6189,
+      "step": 11488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683492577153905e-05,
+      "loss": 0.6202,
+      "step": 11489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6834405417225404e-05,
+      "loss": 0.6022,
+      "step": 11490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683388502818414e-05,
+      "loss": 0.6309,
+      "step": 11491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683336460441791e-05,
+      "loss": 0.593,
+      "step": 11492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6832844145929353e-05,
+      "loss": 0.5632,
+      "step": 11493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683232365272111e-05,
+      "loss": 0.6202,
+      "step": 11494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683180312479583e-05,
+      "loss": 0.624,
+      "step": 11495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683128256215616e-05,
+      "loss": 0.6576,
+      "step": 11496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6830761964804742e-05,
+      "loss": 0.6254,
+      "step": 11497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.683024133274422e-05,
+      "loss": 0.6176,
+      "step": 11498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6829720665977243e-05,
+      "loss": 0.6461,
+      "step": 11499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6829199964506455e-05,
+      "loss": 0.6204,
+      "step": 11500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6828679228334502e-05,
+      "loss": 0.6079,
+      "step": 11501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6828158457464027e-05,
+      "loss": 0.5719,
+      "step": 11502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.682763765189768e-05,
+      "loss": 0.5913,
+      "step": 11503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.682711681163811e-05,
+      "loss": 0.6233,
+      "step": 11504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6826595936687958e-05,
+      "loss": 0.6261,
+      "step": 11505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6826075027049875e-05,
+      "loss": 0.5643,
+      "step": 11506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6825554082726504e-05,
+      "loss": 0.5962,
+      "step": 11507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.682503310372049e-05,
+      "loss": 0.5703,
+      "step": 11508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.682451209003449e-05,
+      "loss": 0.6085,
+      "step": 11509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6823991041671143e-05,
+      "loss": 0.6368,
+      "step": 11510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.68234699586331e-05,
+      "loss": 0.5876,
+      "step": 11511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6822948840923006e-05,
+      "loss": 0.5717,
+      "step": 11512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6822427688543512e-05,
+      "loss": 0.6109,
+      "step": 11513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6821906501497266e-05,
+      "loss": 0.6194,
+      "step": 11514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6821385279786912e-05,
+      "loss": 0.6667,
+      "step": 11515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6820864023415103e-05,
+      "loss": 0.5758,
+      "step": 11516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6820342732384487e-05,
+      "loss": 0.5655,
+      "step": 11517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.681982140669771e-05,
+      "loss": 0.5932,
+      "step": 11518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6819300046357425e-05,
+      "loss": 0.6049,
+      "step": 11519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6818778651366282e-05,
+      "loss": 0.6578,
+      "step": 11520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6818257221726924e-05,
+      "loss": 0.5883,
+      "step": 11521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6817735757442007e-05,
+      "loss": 0.5496,
+      "step": 11522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6817214258514175e-05,
+      "loss": 0.6112,
+      "step": 11523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6816692724946083e-05,
+      "loss": 0.6167,
+      "step": 11524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6816171156740378e-05,
+      "loss": 0.6088,
+      "step": 11525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.681564955389971e-05,
+      "loss": 0.6235,
+      "step": 11526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.681512791642673e-05,
+      "loss": 0.601,
+      "step": 11527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6814606244324092e-05,
+      "loss": 0.5749,
+      "step": 11528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6814084537594442e-05,
+      "loss": 0.5741,
+      "step": 11529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6813562796240434e-05,
+      "loss": 0.6467,
+      "step": 11530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6813041020264716e-05,
+      "loss": 0.6407,
+      "step": 11531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6812519209669946e-05,
+      "loss": 0.6073,
+      "step": 11532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6811997364458768e-05,
+      "loss": 0.6136,
+      "step": 11533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.6811475484633836e-05,
+      "loss": 0.6308,
+      "step": 11534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.68109535701978e-05,
+      "loss": 0.5763,
+      "step": 11535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.681043162115332e-05,
+      "loss": 0.6218,
+      "step": 11536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6809909637503035e-05,
+      "loss": 0.5635,
+      "step": 11537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680938761924961e-05,
+      "loss": 0.6027,
+      "step": 11538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680886556639569e-05,
+      "loss": 0.6005,
+      "step": 11539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680834347894393e-05,
+      "loss": 0.6122,
+      "step": 11540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6807821356896986e-05,
+      "loss": 0.5433,
+      "step": 11541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6807299200257503e-05,
+      "loss": 0.6664,
+      "step": 11542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6806777009028145e-05,
+      "loss": 0.5961,
+      "step": 11543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6806254783211555e-05,
+      "loss": 0.6225,
+      "step": 11544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680573252281039e-05,
+      "loss": 0.6702,
+      "step": 11545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680521022782731e-05,
+      "loss": 0.6151,
+      "step": 11546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6804687898264962e-05,
+      "loss": 0.6325,
+      "step": 11547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6804165534126002e-05,
+      "loss": 0.6463,
+      "step": 11548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680364313541308e-05,
+      "loss": 0.6104,
+      "step": 11549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.680312070212886e-05,
+      "loss": 0.6562,
+      "step": 11550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6802598234275986e-05,
+      "loss": 0.6166,
+      "step": 11551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6802075731857125e-05,
+      "loss": 0.5902,
+      "step": 11552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6801553194874922e-05,
+      "loss": 0.5874,
+      "step": 11553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6801030623332033e-05,
+      "loss": 0.6587,
+      "step": 11554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6800508017231118e-05,
+      "loss": 0.6391,
+      "step": 11555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679998537657483e-05,
+      "loss": 0.621,
+      "step": 11556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6799462701365825e-05,
+      "loss": 0.5911,
+      "step": 11557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679893999160676e-05,
+      "loss": 0.6021,
+      "step": 11558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6798417247300286e-05,
+      "loss": 0.5672,
+      "step": 11559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6797894468449065e-05,
+      "loss": 0.5589,
+      "step": 11560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679737165505575e-05,
+      "loss": 0.698,
+      "step": 11561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6796848807123e-05,
+      "loss": 0.5905,
+      "step": 11562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6796325924653473e-05,
+      "loss": 0.5975,
+      "step": 11563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679580300764982e-05,
+      "loss": 0.5855,
+      "step": 11564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6795280056114703e-05,
+      "loss": 0.578,
+      "step": 11565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679475707005078e-05,
+      "loss": 0.6461,
+      "step": 11566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6794234049460706e-05,
+      "loss": 0.599,
+      "step": 11567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6793710994347138e-05,
+      "loss": 0.6247,
+      "step": 11568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679318790471273e-05,
+      "loss": 0.5887,
+      "step": 11569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679266478056015e-05,
+      "loss": 0.5996,
+      "step": 11570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679214162189205e-05,
+      "loss": 0.6512,
+      "step": 11571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6791618428711092e-05,
+      "loss": 0.6286,
+      "step": 11572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6791095201019928e-05,
+      "loss": 0.5809,
+      "step": 11573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679057193882122e-05,
+      "loss": 0.6529,
+      "step": 11574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.679004864211763e-05,
+      "loss": 0.5881,
+      "step": 11575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6789525310911813e-05,
+      "loss": 0.6204,
+      "step": 11576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6789001945206428e-05,
+      "loss": 0.5681,
+      "step": 11577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.678847854500414e-05,
+      "loss": 0.6213,
+      "step": 11578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6787955110307603e-05,
+      "loss": 0.5896,
+      "step": 11579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6787431641119474e-05,
+      "loss": 0.6529,
+      "step": 11580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6786908137442418e-05,
+      "loss": 0.576,
+      "step": 11581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.67863845992791e-05,
+      "loss": 0.5611,
+      "step": 11582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6785861026632167e-05,
+      "loss": 0.5832,
+      "step": 11583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.678533741950429e-05,
+      "loss": 0.6396,
+      "step": 11584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.678481377789813e-05,
+      "loss": 0.611,
+      "step": 11585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.678429010181634e-05,
+      "loss": 0.621,
+      "step": 11586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6783766391261585e-05,
+      "loss": 0.6065,
+      "step": 11587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6783242646236527e-05,
+      "loss": 0.6959,
+      "step": 11588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6782718866743824e-05,
+      "loss": 0.6141,
+      "step": 11589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6782195052786144e-05,
+      "loss": 0.6333,
+      "step": 11590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6781671204366143e-05,
+      "loss": 0.6334,
+      "step": 11591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6781147321486486e-05,
+      "loss": 0.6113,
+      "step": 11592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6780623404149834e-05,
+      "loss": 0.673,
+      "step": 11593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6780099452358846e-05,
+      "loss": 0.5881,
+      "step": 11594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677957546611619e-05,
+      "loss": 0.652,
+      "step": 11595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6779051445424523e-05,
+      "loss": 0.5326,
+      "step": 11596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677852739028651e-05,
+      "loss": 0.5819,
+      "step": 11597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6778003300704817e-05,
+      "loss": 0.5889,
+      "step": 11598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.67774791766821e-05,
+      "loss": 0.637,
+      "step": 11599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677695501822103e-05,
+      "loss": 0.616,
+      "step": 11600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6776430825324265e-05,
+      "loss": 0.5248,
+      "step": 11601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6775906597994472e-05,
+      "loss": 0.6366,
+      "step": 11602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6775382336234312e-05,
+      "loss": 0.5863,
+      "step": 11603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677485804004645e-05,
+      "loss": 0.5937,
+      "step": 11604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677433370943355e-05,
+      "loss": 0.611,
+      "step": 11605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6773809344398277e-05,
+      "loss": 0.5872,
+      "step": 11606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6773284944943292e-05,
+      "loss": 0.5737,
+      "step": 11607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6772760511071268e-05,
+      "loss": 0.6739,
+      "step": 11608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677223604278486e-05,
+      "loss": 0.5672,
+      "step": 11609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6771711540086738e-05,
+      "loss": 0.5816,
+      "step": 11610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677118700297957e-05,
+      "loss": 0.5243,
+      "step": 11611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6770662431466015e-05,
+      "loss": 0.6202,
+      "step": 11612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.677013782554874e-05,
+      "loss": 0.6601,
+      "step": 11613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6769613185230417e-05,
+      "loss": 0.604,
+      "step": 11614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6769088510513702e-05,
+      "loss": 0.5791,
+      "step": 11615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6768563801401266e-05,
+      "loss": 0.5858,
+      "step": 11616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676803905789578e-05,
+      "loss": 0.59,
+      "step": 11617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.67675142799999e-05,
+      "loss": 0.5637,
+      "step": 11618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6766989467716297e-05,
+      "loss": 0.6029,
+      "step": 11619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676646462104764e-05,
+      "loss": 0.663,
+      "step": 11620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.67659397399966e-05,
+      "loss": 0.6434,
+      "step": 11621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6765414824565834e-05,
+      "loss": 0.6383,
+      "step": 11622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6764889874758015e-05,
+      "loss": 0.6613,
+      "step": 11623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676436489057581e-05,
+      "loss": 0.6011,
+      "step": 11624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6763839872021887e-05,
+      "loss": 0.6067,
+      "step": 11625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676331481909891e-05,
+      "loss": 0.5916,
+      "step": 11626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676278973180955e-05,
+      "loss": 0.6017,
+      "step": 11627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6762264610156472e-05,
+      "loss": 0.676,
+      "step": 11628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676173945414235e-05,
+      "loss": 0.5729,
+      "step": 11629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676121426376985e-05,
+      "loss": 0.5983,
+      "step": 11630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6760689039041643e-05,
+      "loss": 0.5925,
+      "step": 11631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.676016377996039e-05,
+      "loss": 0.6302,
+      "step": 11632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6759638486528768e-05,
+      "loss": 0.5547,
+      "step": 11633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6759113158749444e-05,
+      "loss": 0.6634,
+      "step": 11634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.675858779662508e-05,
+      "loss": 0.689,
+      "step": 11635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6758062400158362e-05,
+      "loss": 0.5867,
+      "step": 11636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6757536969351944e-05,
+      "loss": 0.6108,
+      "step": 11637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6757011504208504e-05,
+      "loss": 0.6738,
+      "step": 11638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6756486004730706e-05,
+      "loss": 0.5813,
+      "step": 11639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6755960470921226e-05,
+      "loss": 0.6338,
+      "step": 11640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6755434902782733e-05,
+      "loss": 0.6495,
+      "step": 11641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6754909300317895e-05,
+      "loss": 0.5653,
+      "step": 11642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6754383663529385e-05,
+      "loss": 0.6528,
+      "step": 11643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6753857992419876e-05,
+      "loss": 0.6546,
+      "step": 11644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6753332286992034e-05,
+      "loss": 0.618,
+      "step": 11645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.675280654724853e-05,
+      "loss": 0.6327,
+      "step": 11646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6752280773192045e-05,
+      "loss": 0.6165,
+      "step": 11647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.675175496482524e-05,
+      "loss": 0.5751,
+      "step": 11648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6751229122150793e-05,
+      "loss": 0.5856,
+      "step": 11649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6750703245171373e-05,
+      "loss": 0.5806,
+      "step": 11650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.675017733388965e-05,
+      "loss": 0.6461,
+      "step": 11651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6749651388308298e-05,
+      "loss": 0.6164,
+      "step": 11652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6749125408429997e-05,
+      "loss": 0.5813,
+      "step": 11653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674859939425741e-05,
+      "loss": 0.5943,
+      "step": 11654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674807334579321e-05,
+      "loss": 0.6333,
+      "step": 11655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6747547263040075e-05,
+      "loss": 0.5734,
+      "step": 11656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674702114600068e-05,
+      "loss": 0.6142,
+      "step": 11657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674649499467769e-05,
+      "loss": 0.5781,
+      "step": 11658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6745968809073783e-05,
+      "loss": 0.59,
+      "step": 11659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6745442589191634e-05,
+      "loss": 0.572,
+      "step": 11660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6744916335033918e-05,
+      "loss": 0.6135,
+      "step": 11661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6744390046603303e-05,
+      "loss": 0.6043,
+      "step": 11662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6743863723902474e-05,
+      "loss": 0.581,
+      "step": 11663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674333736693409e-05,
+      "loss": 0.6207,
+      "step": 11664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674281097570084e-05,
+      "loss": 0.5854,
+      "step": 11665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.674228455020539e-05,
+      "loss": 0.5573,
+      "step": 11666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6741758090450418e-05,
+      "loss": 0.6608,
+      "step": 11667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.67412315964386e-05,
+      "loss": 0.6277,
+      "step": 11668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6740705068172604e-05,
+      "loss": 0.6291,
+      "step": 11669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6740178505655114e-05,
+      "loss": 0.6135,
+      "step": 11670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6739651908888807e-05,
+      "loss": 0.6088,
+      "step": 11671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6739125277876353e-05,
+      "loss": 0.577,
+      "step": 11672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.673859861262043e-05,
+      "loss": 0.5962,
+      "step": 11673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6738071913123715e-05,
+      "loss": 0.6377,
+      "step": 11674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.673754517938888e-05,
+      "loss": 0.6216,
+      "step": 11675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6737018411418605e-05,
+      "loss": 0.6011,
+      "step": 11676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.673649160921557e-05,
+      "loss": 0.6042,
+      "step": 11677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6735964772782444e-05,
+      "loss": 0.5874,
+      "step": 11678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6735437902121914e-05,
+      "loss": 0.6548,
+      "step": 11679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6734910997236648e-05,
+      "loss": 0.5828,
+      "step": 11680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.6734384058129326e-05,
+      "loss": 0.6841,
+      "step": 11681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6733857084802626e-05,
+      "loss": 0.6026,
+      "step": 11682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.673333007725923e-05,
+      "loss": 0.593,
+      "step": 11683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.673280303550181e-05,
+      "loss": 0.6502,
+      "step": 11684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6732275959533045e-05,
+      "loss": 0.6,
+      "step": 11685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6731748849355612e-05,
+      "loss": 0.7071,
+      "step": 11686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6731221704972194e-05,
+      "loss": 0.5971,
+      "step": 11687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6730694526385466e-05,
+      "loss": 0.5797,
+      "step": 11688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.673016731359811e-05,
+      "loss": 0.6671,
+      "step": 11689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.67296400666128e-05,
+      "loss": 0.6003,
+      "step": 11690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6729112785432224e-05,
+      "loss": 0.6387,
+      "step": 11691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.672858547005905e-05,
+      "loss": 0.5927,
+      "step": 11692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6728058120495965e-05,
+      "loss": 0.6359,
+      "step": 11693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6727530736745643e-05,
+      "loss": 0.625,
+      "step": 11694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.672700331881077e-05,
+      "loss": 0.6412,
+      "step": 11695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.672647586669402e-05,
+      "loss": 0.6297,
+      "step": 11696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6725948380398082e-05,
+      "loss": 0.6796,
+      "step": 11697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6725420859925624e-05,
+      "loss": 0.6223,
+      "step": 11698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6724893305279338e-05,
+      "loss": 0.5691,
+      "step": 11699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6724365716461896e-05,
+      "loss": 0.5499,
+      "step": 11700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6723838093475982e-05,
+      "loss": 0.5795,
+      "step": 11701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6723310436324283e-05,
+      "loss": 0.6231,
+      "step": 11702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6722782745009468e-05,
+      "loss": 0.6056,
+      "step": 11703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.672225501953423e-05,
+      "loss": 0.6533,
+      "step": 11704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6721727259901245e-05,
+      "loss": 0.6017,
+      "step": 11705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.672119946611319e-05,
+      "loss": 0.562,
+      "step": 11706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6720671638172757e-05,
+      "loss": 0.5825,
+      "step": 11707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6720143776082625e-05,
+      "loss": 0.6007,
+      "step": 11708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.671961587984547e-05,
+      "loss": 0.57,
+      "step": 11709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6719087949463983e-05,
+      "loss": 0.5827,
+      "step": 11710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.671855998494084e-05,
+      "loss": 0.5888,
+      "step": 11711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6718031986278726e-05,
+      "loss": 0.626,
+      "step": 11712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6717503953480322e-05,
+      "loss": 0.6382,
+      "step": 11713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6716975886548316e-05,
+      "loss": 0.565,
+      "step": 11714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6716447785485385e-05,
+      "loss": 0.6039,
+      "step": 11715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6715919650294216e-05,
+      "loss": 0.6153,
+      "step": 11716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.67153914809775e-05,
+      "loss": 0.6089,
+      "step": 11717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6714863277537904e-05,
+      "loss": 0.6366,
+      "step": 11718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6714335039978125e-05,
+      "loss": 0.6317,
+      "step": 11719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.671380676830084e-05,
+      "loss": 0.6726,
+      "step": 11720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6713278462508742e-05,
+      "loss": 0.5523,
+      "step": 11721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.671275012260451e-05,
+      "loss": 0.56,
+      "step": 11722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6712221748590825e-05,
+      "loss": 0.621,
+      "step": 11723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6711693340470377e-05,
+      "loss": 0.6308,
+      "step": 11724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6711164898245852e-05,
+      "loss": 0.5975,
+      "step": 11725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.671063642191993e-05,
+      "loss": 0.597,
+      "step": 11726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6710107911495296e-05,
+      "loss": 0.6306,
+      "step": 11727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6709579366974642e-05,
+      "loss": 0.5811,
+      "step": 11728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.670905078836065e-05,
+      "loss": 0.5264,
+      "step": 11729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6708522175656007e-05,
+      "loss": 0.6408,
+      "step": 11730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.67079935288634e-05,
+      "loss": 0.6268,
+      "step": 11731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6707464847985507e-05,
+      "loss": 0.6039,
+      "step": 11732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6706936133025026e-05,
+      "loss": 0.571,
+      "step": 11733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6706407383984635e-05,
+      "loss": 0.6458,
+      "step": 11734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6705878600867028e-05,
+      "loss": 0.5777,
+      "step": 11735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6705349783674884e-05,
+      "loss": 0.6641,
+      "step": 11736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6704820932410895e-05,
+      "loss": 0.664,
+      "step": 11737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.670429204707775e-05,
+      "loss": 0.6265,
+      "step": 11738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.670376312767813e-05,
+      "loss": 0.5525,
+      "step": 11739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6703234174214727e-05,
+      "loss": 0.6009,
+      "step": 11740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.670270518669023e-05,
+      "loss": 0.6256,
+      "step": 11741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6702176165107323e-05,
+      "loss": 0.6333,
+      "step": 11742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6701647109468696e-05,
+      "loss": 0.5636,
+      "step": 11743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.670111801977704e-05,
+      "loss": 0.6327,
+      "step": 11744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6700588896035038e-05,
+      "loss": 0.5753,
+      "step": 11745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6700059738245383e-05,
+      "loss": 0.618,
+      "step": 11746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6699530546410763e-05,
+      "loss": 0.6031,
+      "step": 11747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6699001320533862e-05,
+      "loss": 0.6168,
+      "step": 11748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.669847206061738e-05,
+      "loss": 0.6241,
+      "step": 11749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6697942766663998e-05,
+      "loss": 0.6442,
+      "step": 11750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6697413438676404e-05,
+      "loss": 0.5927,
+      "step": 11751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.669688407665729e-05,
+      "loss": 0.6102,
+      "step": 11752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.669635468060935e-05,
+      "loss": 0.6744,
+      "step": 11753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6695825250535268e-05,
+      "loss": 0.6219,
+      "step": 11754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.669529578643774e-05,
+      "loss": 0.6379,
+      "step": 11755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6694766288319453e-05,
+      "loss": 0.5721,
+      "step": 11756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6694236756183096e-05,
+      "loss": 0.5682,
+      "step": 11757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6693707190031362e-05,
+      "loss": 0.6021,
+      "step": 11758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6693177589866943e-05,
+      "loss": 0.5929,
+      "step": 11759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6692647955692528e-05,
+      "loss": 0.598,
+      "step": 11760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6692118287510807e-05,
+      "loss": 0.6704,
+      "step": 11761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6691588585324475e-05,
+      "loss": 0.6018,
+      "step": 11762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.669105884913622e-05,
+      "loss": 0.641,
+      "step": 11763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6690529078948734e-05,
+      "loss": 0.614,
+      "step": 11764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6689999274764713e-05,
+      "loss": 0.6677,
+      "step": 11765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6689469436586848e-05,
+      "loss": 0.6477,
+      "step": 11766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6688939564417825e-05,
+      "loss": 0.6397,
+      "step": 11767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6688409658260344e-05,
+      "loss": 0.6136,
+      "step": 11768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6687879718117093e-05,
+      "loss": 0.6307,
+      "step": 11769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6687349743990772e-05,
+      "loss": 0.6346,
+      "step": 11770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6686819735884064e-05,
+      "loss": 0.686,
+      "step": 11771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6686289693799666e-05,
+      "loss": 0.608,
+      "step": 11772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6685759617740275e-05,
+      "loss": 0.6758,
+      "step": 11773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6685229507708577e-05,
+      "loss": 0.5699,
+      "step": 11774
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6684699363707274e-05,
+      "loss": 0.6268,
+      "step": 11775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6684169185739052e-05,
+      "loss": 0.5607,
+      "step": 11776
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.668363897380661e-05,
+      "loss": 0.6099,
+      "step": 11777
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6683108727912645e-05,
+      "loss": 0.6097,
+      "step": 11778
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6682578448059844e-05,
+      "loss": 0.6378,
+      "step": 11779
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6682048134250902e-05,
+      "loss": 0.5469,
+      "step": 11780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6681517786488522e-05,
+      "loss": 0.6166,
+      "step": 11781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.668098740477539e-05,
+      "loss": 0.6036,
+      "step": 11782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6680456989114207e-05,
+      "loss": 0.6085,
+      "step": 11783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667992653950766e-05,
+      "loss": 0.6261,
+      "step": 11784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6679396055958456e-05,
+      "loss": 0.5963,
+      "step": 11785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667886553846928e-05,
+      "loss": 0.5895,
+      "step": 11786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667833498704284e-05,
+      "loss": 0.6306,
+      "step": 11787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6677804401681817e-05,
+      "loss": 0.6093,
+      "step": 11788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6677273782388917e-05,
+      "loss": 0.5902,
+      "step": 11789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6676743129166832e-05,
+      "loss": 0.6359,
+      "step": 11790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667621244201826e-05,
+      "loss": 0.5975,
+      "step": 11791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6675681720945894e-05,
+      "loss": 0.6108,
+      "step": 11792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6675150965952442e-05,
+      "loss": 0.5958,
+      "step": 11793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6674620177040585e-05,
+      "loss": 0.6157,
+      "step": 11794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667408935421303e-05,
+      "loss": 0.6185,
+      "step": 11795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6673558497472476e-05,
+      "loss": 0.5492,
+      "step": 11796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.667302760682162e-05,
+      "loss": 0.6265,
+      "step": 11797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6672496682263147e-05,
+      "loss": 0.6249,
+      "step": 11798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6671965723799767e-05,
+      "loss": 0.5544,
+      "step": 11799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6671434731434177e-05,
+      "loss": 0.5897,
+      "step": 11800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6670903705169076e-05,
+      "loss": 0.6516,
+      "step": 11801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6670372645007152e-05,
+      "loss": 0.6503,
+      "step": 11802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6669841550951117e-05,
+      "loss": 0.5607,
+      "step": 11803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6669310423003663e-05,
+      "loss": 0.6252,
+      "step": 11804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6668779261167486e-05,
+      "loss": 0.6226,
+      "step": 11805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6668248065445295e-05,
+      "loss": 0.655,
+      "step": 11806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6667716835839776e-05,
+      "loss": 0.6163,
+      "step": 11807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.666718557235364e-05,
+      "loss": 0.6385,
+      "step": 11808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6666654274989583e-05,
+      "loss": 0.6152,
+      "step": 11809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.66661229437503e-05,
+      "loss": 0.5829,
+      "step": 11810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.666559157863849e-05,
+      "loss": 0.585,
+      "step": 11811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6665060179656865e-05,
+      "loss": 0.6118,
+      "step": 11812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6664528746808112e-05,
+      "loss": 0.5171,
+      "step": 11813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6663997280094934e-05,
+      "loss": 0.6138,
+      "step": 11814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6663465779520042e-05,
+      "loss": 0.5663,
+      "step": 11815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6662934245086123e-05,
+      "loss": 0.6398,
+      "step": 11816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6662402676795886e-05,
+      "loss": 0.6161,
+      "step": 11817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.666187107465203e-05,
+      "loss": 0.5712,
+      "step": 11818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6661339438657255e-05,
+      "loss": 0.5648,
+      "step": 11819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6660807768814263e-05,
+      "loss": 0.5979,
+      "step": 11820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.666027606512576e-05,
+      "loss": 0.5681,
+      "step": 11821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6659744327594437e-05,
+      "loss": 0.5681,
+      "step": 11822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.665921255622301e-05,
+      "loss": 0.5886,
+      "step": 11823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.665868075101417e-05,
+      "loss": 0.5857,
+      "step": 11824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6658148911970623e-05,
+      "loss": 0.6319,
+      "step": 11825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.665761703909507e-05,
+      "loss": 0.5618,
+      "step": 11826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6657085132390222e-05,
+      "loss": 0.6006,
+      "step": 11827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.6656553191858767e-05,
+      "loss": 0.6068,
+      "step": 11828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6656021217503416e-05,
+      "loss": 0.5971,
+      "step": 11829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6655489209326876e-05,
+      "loss": 0.6109,
+      "step": 11830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6654957167331845e-05,
+      "loss": 0.6171,
+      "step": 11831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.665442509152103e-05,
+      "loss": 0.5543,
+      "step": 11832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6653892981897128e-05,
+      "loss": 0.5846,
+      "step": 11833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6653360838462847e-05,
+      "loss": 0.6111,
+      "step": 11834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6652828661220892e-05,
+      "loss": 0.6456,
+      "step": 11835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6652296450173968e-05,
+      "loss": 0.6076,
+      "step": 11836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6651764205324778e-05,
+      "loss": 0.6336,
+      "step": 11837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6651231926676022e-05,
+      "loss": 0.6075,
+      "step": 11838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.665069961423041e-05,
+      "loss": 0.5857,
+      "step": 11839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6650167267990645e-05,
+      "loss": 0.5863,
+      "step": 11840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6649634887959435e-05,
+      "loss": 0.6262,
+      "step": 11841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6649102474139484e-05,
+      "loss": 0.5681,
+      "step": 11842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6648570026533493e-05,
+      "loss": 0.58,
+      "step": 11843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6648037545144172e-05,
+      "loss": 0.5796,
+      "step": 11844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6647505029974222e-05,
+      "loss": 0.5743,
+      "step": 11845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6646972481026357e-05,
+      "loss": 0.6105,
+      "step": 11846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6646439898303276e-05,
+      "loss": 0.6417,
+      "step": 11847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6645907281807687e-05,
+      "loss": 0.5859,
+      "step": 11848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6645374631542297e-05,
+      "loss": 0.6156,
+      "step": 11849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6644841947509814e-05,
+      "loss": 0.6546,
+      "step": 11850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.664430922971294e-05,
+      "loss": 0.6339,
+      "step": 11851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6643776478154387e-05,
+      "loss": 0.5943,
+      "step": 11852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6643243692836858e-05,
+      "loss": 0.602,
+      "step": 11853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.664271087376306e-05,
+      "loss": 0.6616,
+      "step": 11854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.664217802093571e-05,
+      "loss": 0.6228,
+      "step": 11855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6641645134357504e-05,
+      "loss": 0.6075,
+      "step": 11856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6641112214031153e-05,
+      "loss": 0.6546,
+      "step": 11857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6640579259959372e-05,
+      "loss": 0.6306,
+      "step": 11858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6640046272144858e-05,
+      "loss": 0.5694,
+      "step": 11859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663951325059032e-05,
+      "loss": 0.6294,
+      "step": 11860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663898019529848e-05,
+      "loss": 0.625,
+      "step": 11861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6638447106272028e-05,
+      "loss": 0.5736,
+      "step": 11862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6637913983513687e-05,
+      "loss": 0.5923,
+      "step": 11863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663738082702616e-05,
+      "loss": 0.6228,
+      "step": 11864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663684763681216e-05,
+      "loss": 0.6018,
+      "step": 11865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6636314412874393e-05,
+      "loss": 0.6695,
+      "step": 11866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6635781155215568e-05,
+      "loss": 0.6053,
+      "step": 11867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6635247863838394e-05,
+      "loss": 0.6146,
+      "step": 11868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6634714538745582e-05,
+      "loss": 0.5965,
+      "step": 11869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6634181179939843e-05,
+      "loss": 0.6138,
+      "step": 11870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663364778742389e-05,
+      "loss": 0.6552,
+      "step": 11871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6633114361200426e-05,
+      "loss": 0.6551,
+      "step": 11872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6632580901272163e-05,
+      "loss": 0.6015,
+      "step": 11873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6632047407641817e-05,
+      "loss": 0.5947,
+      "step": 11874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6631513880312096e-05,
+      "loss": 0.5759,
+      "step": 11875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.663098031928571e-05,
+      "loss": 0.6457,
+      "step": 11876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6630446724565375e-05,
+      "loss": 0.6264,
+      "step": 11877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6629913096153795e-05,
+      "loss": 0.5921,
+      "step": 11878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6629379434053685e-05,
+      "loss": 0.619,
+      "step": 11879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6628845738267756e-05,
+      "loss": 0.5975,
+      "step": 11880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.662831200879872e-05,
+      "loss": 0.6009,
+      "step": 11881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6627778245649292e-05,
+      "loss": 0.5819,
+      "step": 11882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6627244448822182e-05,
+      "loss": 0.5829,
+      "step": 11883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6626710618320098e-05,
+      "loss": 0.6638,
+      "step": 11884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.662617675414576e-05,
+      "loss": 0.6499,
+      "step": 11885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6625642856301877e-05,
+      "loss": 0.6309,
+      "step": 11886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.662510892479116e-05,
+      "loss": 0.6113,
+      "step": 11887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6624574959616327e-05,
+      "loss": 0.6128,
+      "step": 11888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.662404096078009e-05,
+      "loss": 0.5511,
+      "step": 11889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.662350692828516e-05,
+      "loss": 0.5743,
+      "step": 11890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6622972862134247e-05,
+      "loss": 0.6851,
+      "step": 11891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6622438762330072e-05,
+      "loss": 0.5982,
+      "step": 11892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6621904628875348e-05,
+      "loss": 0.5804,
+      "step": 11893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6621370461772786e-05,
+      "loss": 0.5932,
+      "step": 11894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.66208362610251e-05,
+      "loss": 0.631,
+      "step": 11895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6620302026635008e-05,
+      "loss": 0.5937,
+      "step": 11896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661976775860522e-05,
+      "loss": 0.5859,
+      "step": 11897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6619233456938458e-05,
+      "loss": 0.5592,
+      "step": 11898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6618699121637426e-05,
+      "loss": 0.6808,
+      "step": 11899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661816475270485e-05,
+      "loss": 0.5543,
+      "step": 11900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661763035014344e-05,
+      "loss": 0.5616,
+      "step": 11901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661709591395591e-05,
+      "loss": 0.6828,
+      "step": 11902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661656144414498e-05,
+      "loss": 0.6722,
+      "step": 11903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6616026940713364e-05,
+      "loss": 0.569,
+      "step": 11904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6615492403663775e-05,
+      "loss": 0.6369,
+      "step": 11905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6614957832998934e-05,
+      "loss": 0.6351,
+      "step": 11906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6614423228721553e-05,
+      "loss": 0.5811,
+      "step": 11907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6613888590834352e-05,
+      "loss": 0.5667,
+      "step": 11908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6613353919340048e-05,
+      "loss": 0.5823,
+      "step": 11909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.661281921424135e-05,
+      "loss": 0.6022,
+      "step": 11910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6612284475540986e-05,
+      "loss": 0.5803,
+      "step": 11911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6611749703241667e-05,
+      "loss": 0.6056,
+      "step": 11912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6611214897346107e-05,
+      "loss": 0.5958,
+      "step": 11913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6610680057857035e-05,
+      "loss": 0.5689,
+      "step": 11914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6610145184777156e-05,
+      "loss": 0.6379,
+      "step": 11915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.66096102781092e-05,
+      "loss": 0.7038,
+      "step": 11916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6609075337855874e-05,
+      "loss": 0.6462,
+      "step": 11917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.66085403640199e-05,
+      "loss": 0.6133,
+      "step": 11918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6608005356604e-05,
+      "loss": 0.6389,
+      "step": 11919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6607470315610888e-05,
+      "loss": 0.6608,
+      "step": 11920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6606935241043284e-05,
+      "loss": 0.6586,
+      "step": 11921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6606400132903905e-05,
+      "loss": 0.5931,
+      "step": 11922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6605864991195473e-05,
+      "loss": 0.5648,
+      "step": 11923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6605329815920707e-05,
+      "loss": 0.6562,
+      "step": 11924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6604794607082328e-05,
+      "loss": 0.5966,
+      "step": 11925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.660425936468305e-05,
+      "loss": 0.5133,
+      "step": 11926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6603724088725597e-05,
+      "loss": 0.6667,
+      "step": 11927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.660318877921269e-05,
+      "loss": 0.649,
+      "step": 11928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6602653436147044e-05,
+      "loss": 0.6208,
+      "step": 11929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6602118059531384e-05,
+      "loss": 0.6204,
+      "step": 11930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.660158264936843e-05,
+      "loss": 0.6766,
+      "step": 11931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6601047205660895e-05,
+      "loss": 0.5864,
+      "step": 11932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6600511728411514e-05,
+      "loss": 0.6276,
+      "step": 11933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6599976217622996e-05,
+      "loss": 0.5952,
+      "step": 11934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6599440673298065e-05,
+      "loss": 0.5742,
+      "step": 11935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6598905095439442e-05,
+      "loss": 0.6198,
+      "step": 11936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6598369484049858e-05,
+      "loss": 0.6136,
+      "step": 11937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6597833839132016e-05,
+      "loss": 0.6436,
+      "step": 11938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6597298160688654e-05,
+      "loss": 0.5743,
+      "step": 11939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6596762448722486e-05,
+      "loss": 0.6452,
+      "step": 11940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6596226703236236e-05,
+      "loss": 0.5797,
+      "step": 11941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6595690924232625e-05,
+      "loss": 0.632,
+      "step": 11942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.659515511171438e-05,
+      "loss": 0.6406,
+      "step": 11943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.659461926568422e-05,
+      "loss": 0.618,
+      "step": 11944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6594083386144866e-05,
+      "loss": 0.5674,
+      "step": 11945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6593547473099043e-05,
+      "loss": 0.555,
+      "step": 11946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6593011526549475e-05,
+      "loss": 0.602,
+      "step": 11947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6592475546498884e-05,
+      "loss": 0.617,
+      "step": 11948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6591939532949996e-05,
+      "loss": 0.6785,
+      "step": 11949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6591403485905532e-05,
+      "loss": 0.5996,
+      "step": 11950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6590867405368217e-05,
+      "loss": 0.628,
+      "step": 11951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.659033129134077e-05,
+      "loss": 0.6581,
+      "step": 11952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6589795143825925e-05,
+      "loss": 0.5775,
+      "step": 11953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.65892589628264e-05,
+      "loss": 0.6523,
+      "step": 11954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6588722748344918e-05,
+      "loss": 0.6075,
+      "step": 11955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.658818650038421e-05,
+      "loss": 0.6013,
+      "step": 11956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6587650218946992e-05,
+      "loss": 0.5914,
+      "step": 11957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6587113904036e-05,
+      "loss": 0.5888,
+      "step": 11958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6586577555653947e-05,
+      "loss": 0.5824,
+      "step": 11959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6586041173803567e-05,
+      "loss": 0.6401,
+      "step": 11960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6585504758487586e-05,
+      "loss": 0.6493,
+      "step": 11961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.658496830970872e-05,
+      "loss": 0.6548,
+      "step": 11962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.658443182746971e-05,
+      "loss": 0.5115,
+      "step": 11963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6583895311773267e-05,
+      "loss": 0.6807,
+      "step": 11964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6583358762622126e-05,
+      "loss": 0.6179,
+      "step": 11965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6582822180019014e-05,
+      "loss": 0.6281,
+      "step": 11966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.658228556396665e-05,
+      "loss": 0.5899,
+      "step": 11967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6581748914467763e-05,
+      "loss": 0.5709,
+      "step": 11968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.658121223152509e-05,
+      "loss": 0.6144,
+      "step": 11969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6580675515141346e-05,
+      "loss": 0.6686,
+      "step": 11970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6580138765319263e-05,
+      "loss": 0.5566,
+      "step": 11971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6579601982061565e-05,
+      "loss": 0.6154,
+      "step": 11972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6579065165370985e-05,
+      "loss": 0.6686,
+      "step": 11973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.657852831525025e-05,
+      "loss": 0.6041,
+      "step": 11974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.6577991431702086e-05,
+      "loss": 0.6038,
+      "step": 11975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6577454514729216e-05,
+      "loss": 0.6283,
+      "step": 11976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6576917564334377e-05,
+      "loss": 0.617,
+      "step": 11977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6576380580520296e-05,
+      "loss": 0.6731,
+      "step": 11978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6575843563289692e-05,
+      "loss": 0.5991,
+      "step": 11979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657530651264531e-05,
+      "loss": 0.6318,
+      "step": 11980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6574769428589863e-05,
+      "loss": 0.5687,
+      "step": 11981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657423231112609e-05,
+      "loss": 0.6118,
+      "step": 11982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657369516025672e-05,
+      "loss": 0.6207,
+      "step": 11983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6573157975984472e-05,
+      "loss": 0.5696,
+      "step": 11984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6572620758312086e-05,
+      "loss": 0.5957,
+      "step": 11985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6572083507242293e-05,
+      "loss": 0.6158,
+      "step": 11986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657154622277782e-05,
+      "loss": 0.6436,
+      "step": 11987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.657100890492139e-05,
+      "loss": 0.6284,
+      "step": 11988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6570471553675745e-05,
+      "loss": 0.5709,
+      "step": 11989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6569934169043604e-05,
+      "loss": 0.5708,
+      "step": 11990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6569396751027705e-05,
+      "loss": 0.5976,
+      "step": 11991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656885929963078e-05,
+      "loss": 0.5835,
+      "step": 11992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6568321814855557e-05,
+      "loss": 0.6083,
+      "step": 11993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6567784296704766e-05,
+      "loss": 0.554,
+      "step": 11994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656724674518114e-05,
+      "loss": 0.6094,
+      "step": 11995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6566709160287412e-05,
+      "loss": 0.5849,
+      "step": 11996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6566171542026305e-05,
+      "loss": 0.6672,
+      "step": 11997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656563389040056e-05,
+      "loss": 0.5984,
+      "step": 11998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656509620541291e-05,
+      "loss": 0.6654,
+      "step": 11999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6564558487066082e-05,
+      "loss": 0.6023,
+      "step": 12000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6564020735362808e-05,
+      "loss": 0.5802,
+      "step": 12001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6563482950305822e-05,
+      "loss": 0.6066,
+      "step": 12002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656294513189786e-05,
+      "loss": 0.6192,
+      "step": 12003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656240728014165e-05,
+      "loss": 0.547,
+      "step": 12004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656186939503993e-05,
+      "loss": 0.5549,
+      "step": 12005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6561331476595422e-05,
+      "loss": 0.5987,
+      "step": 12006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6560793524810874e-05,
+      "loss": 0.6625,
+      "step": 12007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656025553968901e-05,
+      "loss": 0.6078,
+      "step": 12008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655971752123257e-05,
+      "loss": 0.5844,
+      "step": 12009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655917946944428e-05,
+      "loss": 0.6664,
+      "step": 12010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6558641384326878e-05,
+      "loss": 0.6439,
+      "step": 12011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.65581032658831e-05,
+      "loss": 0.572,
+      "step": 12012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6557565114115677e-05,
+      "loss": 0.5791,
+      "step": 12013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6557026929027348e-05,
+      "loss": 0.6733,
+      "step": 12014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6556488710620843e-05,
+      "loss": 0.6176,
+      "step": 12015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.65559504588989e-05,
+      "loss": 0.5593,
+      "step": 12016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655541217386425e-05,
+      "loss": 0.6198,
+      "step": 12017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6554873855519634e-05,
+      "loss": 0.6468,
+      "step": 12018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655433550386778e-05,
+      "loss": 0.5655,
+      "step": 12019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655379711891143e-05,
+      "loss": 0.6324,
+      "step": 12020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655325870065332e-05,
+      "loss": 0.601,
+      "step": 12021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655272024909618e-05,
+      "loss": 0.5988,
+      "step": 12022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6552181764242753e-05,
+      "loss": 0.5976,
+      "step": 12023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6551643246095767e-05,
+      "loss": 0.6108,
+      "step": 12024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6551104694657967e-05,
+      "loss": 0.6135,
+      "step": 12025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6550566109932084e-05,
+      "loss": 0.5973,
+      "step": 12026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6550027491920855e-05,
+      "loss": 0.646,
+      "step": 12027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.654948884062702e-05,
+      "loss": 0.5831,
+      "step": 12028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6548950156053312e-05,
+      "loss": 0.594,
+      "step": 12029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6548411438202472e-05,
+      "loss": 0.6284,
+      "step": 12030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6547872687077234e-05,
+      "loss": 0.649,
+      "step": 12031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.654733390268034e-05,
+      "loss": 0.639,
+      "step": 12032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6546795085014524e-05,
+      "loss": 0.619,
+      "step": 12033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6546256234082526e-05,
+      "loss": 0.5961,
+      "step": 12034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6545717349887077e-05,
+      "loss": 0.5233,
+      "step": 12035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6545178432430925e-05,
+      "loss": 0.5802,
+      "step": 12036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6544639481716806e-05,
+      "loss": 0.6298,
+      "step": 12037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6544100497747456e-05,
+      "loss": 0.5724,
+      "step": 12038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6543561480525615e-05,
+      "loss": 0.6353,
+      "step": 12039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.654302243005402e-05,
+      "loss": 0.5861,
+      "step": 12040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6542483346335414e-05,
+      "loss": 0.6379,
+      "step": 12041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6541944229372533e-05,
+      "loss": 0.5697,
+      "step": 12042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6541405079168118e-05,
+      "loss": 0.5432,
+      "step": 12043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6540865895724908e-05,
+      "loss": 0.619,
+      "step": 12044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6540326679045646e-05,
+      "loss": 0.5646,
+      "step": 12045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6539787429133063e-05,
+      "loss": 0.6426,
+      "step": 12046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6539248145989907e-05,
+      "loss": 0.6108,
+      "step": 12047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6538708829618915e-05,
+      "loss": 0.6615,
+      "step": 12048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653816948002283e-05,
+      "loss": 0.6368,
+      "step": 12049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6537630097204394e-05,
+      "loss": 0.5688,
+      "step": 12050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653709068116634e-05,
+      "loss": 0.6028,
+      "step": 12051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6536551231911417e-05,
+      "loss": 0.6225,
+      "step": 12052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6536011749442357e-05,
+      "loss": 0.5471,
+      "step": 12053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6535472233761914e-05,
+      "loss": 0.6006,
+      "step": 12054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653493268487282e-05,
+      "loss": 0.6921,
+      "step": 12055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653439310277782e-05,
+      "loss": 0.5843,
+      "step": 12056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6533853487479654e-05,
+      "loss": 0.6263,
+      "step": 12057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6533313838981064e-05,
+      "loss": 0.5899,
+      "step": 12058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6532774157284793e-05,
+      "loss": 0.5296,
+      "step": 12059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6532234442393586e-05,
+      "loss": 0.6155,
+      "step": 12060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653169469431018e-05,
+      "loss": 0.6053,
+      "step": 12061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653115491303732e-05,
+      "loss": 0.6802,
+      "step": 12062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6530615098577754e-05,
+      "loss": 0.633,
+      "step": 12063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6530075250934213e-05,
+      "loss": 0.6163,
+      "step": 12064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6529535370109455e-05,
+      "loss": 0.6161,
+      "step": 12065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652899545610621e-05,
+      "loss": 0.609,
+      "step": 12066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6528455508927227e-05,
+      "loss": 0.6167,
+      "step": 12067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6527915528575253e-05,
+      "loss": 0.5999,
+      "step": 12068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6527375515053026e-05,
+      "loss": 0.5559,
+      "step": 12069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6526835468363295e-05,
+      "loss": 0.5589,
+      "step": 12070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6526295388508796e-05,
+      "loss": 0.633,
+      "step": 12071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6525755275492283e-05,
+      "loss": 0.6072,
+      "step": 12072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6525215129316496e-05,
+      "loss": 0.5622,
+      "step": 12073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652467494998418e-05,
+      "loss": 0.6386,
+      "step": 12074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6524134737498078e-05,
+      "loss": 0.5894,
+      "step": 12075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652359449186094e-05,
+      "loss": 0.6139,
+      "step": 12076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6523054213075504e-05,
+      "loss": 0.508,
+      "step": 12077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652251390114452e-05,
+      "loss": 0.6385,
+      "step": 12078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6521973556070735e-05,
+      "loss": 0.5873,
+      "step": 12079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6521433177856893e-05,
+      "loss": 0.5913,
+      "step": 12080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6520892766505733e-05,
+      "loss": 0.5995,
+      "step": 12081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6520352322020016e-05,
+      "loss": 0.6027,
+      "step": 12082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651981184440247e-05,
+      "loss": 0.6065,
+      "step": 12083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6519271333655856e-05,
+      "loss": 0.5341,
+      "step": 12084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6518730789782916e-05,
+      "loss": 0.5863,
+      "step": 12085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6518190212786392e-05,
+      "loss": 0.5308,
+      "step": 12086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6517649602669034e-05,
+      "loss": 0.597,
+      "step": 12087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6517108959433596e-05,
+      "loss": 0.6019,
+      "step": 12088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651656828308281e-05,
+      "loss": 0.6509,
+      "step": 12089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651602757361944e-05,
+      "loss": 0.5707,
+      "step": 12090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6515486831046223e-05,
+      "loss": 0.6651,
+      "step": 12091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6514946055365908e-05,
+      "loss": 0.6751,
+      "step": 12092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6514405246581248e-05,
+      "loss": 0.6289,
+      "step": 12093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6513864404694984e-05,
+      "loss": 0.6261,
+      "step": 12094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6513323529709864e-05,
+      "loss": 0.6028,
+      "step": 12095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6512782621628643e-05,
+      "loss": 0.6612,
+      "step": 12096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6512241680454065e-05,
+      "loss": 0.587,
+      "step": 12097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651170070618888e-05,
+      "loss": 0.636,
+      "step": 12098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6511159698835842e-05,
+      "loss": 0.6772,
+      "step": 12099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6510618658397687e-05,
+      "loss": 0.5787,
+      "step": 12100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6510077584877175e-05,
+      "loss": 0.6602,
+      "step": 12101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6509536478277052e-05,
+      "loss": 0.6263,
+      "step": 12102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650899533860007e-05,
+      "loss": 0.6144,
+      "step": 12103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6508454165848974e-05,
+      "loss": 0.6085,
+      "step": 12104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6507912960026517e-05,
+      "loss": 0.5806,
+      "step": 12105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6507371721135446e-05,
+      "loss": 0.6237,
+      "step": 12106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6506830449178517e-05,
+      "loss": 0.5742,
+      "step": 12107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6506289144158477e-05,
+      "loss": 0.5957,
+      "step": 12108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6505747806078076e-05,
+      "loss": 0.5791,
+      "step": 12109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6505206434940064e-05,
+      "loss": 0.5863,
+      "step": 12110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650466503074719e-05,
+      "loss": 0.5344,
+      "step": 12111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6504123593502213e-05,
+      "loss": 0.5848,
+      "step": 12112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6503582123207874e-05,
+      "loss": 0.6135,
+      "step": 12113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6503040619866934e-05,
+      "loss": 0.6062,
+      "step": 12114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650249908348214e-05,
+      "loss": 0.6391,
+      "step": 12115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650195751405624e-05,
+      "loss": 0.578,
+      "step": 12116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6501415911591995e-05,
+      "loss": 0.548,
+      "step": 12117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650087427609215e-05,
+      "loss": 0.6284,
+      "step": 12118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6500332607559456e-05,
+      "loss": 0.6235,
+      "step": 12119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6499790905996673e-05,
+      "loss": 0.5817,
+      "step": 12120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6499249171406543e-05,
+      "loss": 0.6179,
+      "step": 12121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.649870740379183e-05,
+      "loss": 0.6165,
+      "step": 12122
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6498165603155277e-05,
+      "loss": 0.6136,
+      "step": 12123
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6497623769499646e-05,
+      "loss": 0.6249,
+      "step": 12124
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.649708190282768e-05,
+      "loss": 0.6584,
+      "step": 12125
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6496540003142146e-05,
+      "loss": 0.6379,
+      "step": 12126
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6495998070445783e-05,
+      "loss": 0.5871,
+      "step": 12127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6495456104741354e-05,
+      "loss": 0.6121,
+      "step": 12128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.649491410603161e-05,
+      "loss": 0.634,
+      "step": 12129
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6494372074319304e-05,
+      "loss": 0.6227,
+      "step": 12130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6493830009607193e-05,
+      "loss": 0.6897,
+      "step": 12131
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6493287911898027e-05,
+      "loss": 0.6212,
+      "step": 12132
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.649274578119457e-05,
+      "loss": 0.6188,
+      "step": 12133
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6492203617499564e-05,
+      "loss": 0.5978,
+      "step": 12134
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6491661420815773e-05,
+      "loss": 0.6567,
+      "step": 12135
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6491119191145948e-05,
+      "loss": 0.6146,
+      "step": 12136
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6490576928492846e-05,
+      "loss": 0.6277,
+      "step": 12137
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6490034632859223e-05,
+      "loss": 0.5341,
+      "step": 12138
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6489492304247828e-05,
+      "loss": 0.5939,
+      "step": 12139
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6488949942661427e-05,
+      "loss": 0.6254,
+      "step": 12140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648840754810277e-05,
+      "loss": 0.6493,
+      "step": 12141
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6487865120574618e-05,
+      "loss": 0.5555,
+      "step": 12142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6487322660079715e-05,
+      "loss": 0.6273,
+      "step": 12143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648678016662083e-05,
+      "loss": 0.6477,
+      "step": 12144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6486237640200715e-05,
+      "loss": 0.6045,
+      "step": 12145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648569508082213e-05,
+      "loss": 0.5775,
+      "step": 12146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6485152488487822e-05,
+      "loss": 0.5845,
+      "step": 12147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648460986320056e-05,
+      "loss": 0.6134,
+      "step": 12148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6484067204963095e-05,
+      "loss": 0.5835,
+      "step": 12149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648352451377819e-05,
+      "loss": 0.5896,
+      "step": 12150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6482981789648592e-05,
+      "loss": 0.6013,
+      "step": 12151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6482439032577063e-05,
+      "loss": 0.6495,
+      "step": 12152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648189624256637e-05,
+      "loss": 0.6019,
+      "step": 12153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.648135341961926e-05,
+      "loss": 0.5747,
+      "step": 12154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6480810563738496e-05,
+      "loss": 0.596,
+      "step": 12155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6480267674926835e-05,
+      "loss": 0.6294,
+      "step": 12156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6479724753187033e-05,
+      "loss": 0.6363,
+      "step": 12157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6479181798521857e-05,
+      "loss": 0.5562,
+      "step": 12158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.647863881093406e-05,
+      "loss": 0.6096,
+      "step": 12159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.64780957904264e-05,
+      "loss": 0.6782,
+      "step": 12160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6477552737001638e-05,
+      "loss": 0.6657,
+      "step": 12161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6477009650662534e-05,
+      "loss": 0.6739,
+      "step": 12162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6476466531411846e-05,
+      "loss": 0.618,
+      "step": 12163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6475923379252336e-05,
+      "loss": 0.6059,
+      "step": 12164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6475380194186765e-05,
+      "loss": 0.6516,
+      "step": 12165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6474836976217888e-05,
+      "loss": 0.581,
+      "step": 12166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.647429372534847e-05,
+      "loss": 0.6128,
+      "step": 12167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6473750441581265e-05,
+      "loss": 0.6163,
+      "step": 12168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6473207124919043e-05,
+      "loss": 0.64,
+      "step": 12169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.647266377536456e-05,
+      "loss": 0.5696,
+      "step": 12170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6472120392920575e-05,
+      "loss": 0.6265,
+      "step": 12171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.647157697758985e-05,
+      "loss": 0.6274,
+      "step": 12172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6471033529375147e-05,
+      "loss": 0.5827,
+      "step": 12173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.647049004827923e-05,
+      "loss": 0.6099,
+      "step": 12174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6469946534304855e-05,
+      "loss": 0.6446,
+      "step": 12175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6469402987454788e-05,
+      "loss": 0.5729,
+      "step": 12176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6468859407731793e-05,
+      "loss": 0.604,
+      "step": 12177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6468315795138624e-05,
+      "loss": 0.6138,
+      "step": 12178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646777214967805e-05,
+      "loss": 0.5888,
+      "step": 12179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646722847135283e-05,
+      "loss": 0.5695,
+      "step": 12180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646668476016573e-05,
+      "loss": 0.5696,
+      "step": 12181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646614101611951e-05,
+      "loss": 0.6075,
+      "step": 12182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6465597239216936e-05,
+      "loss": 0.6334,
+      "step": 12183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6465053429460765e-05,
+      "loss": 0.5899,
+      "step": 12184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6464509586853766e-05,
+      "loss": 0.5814,
+      "step": 12185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6463965711398704e-05,
+      "loss": 0.6078,
+      "step": 12186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6463421803098334e-05,
+      "loss": 0.6315,
+      "step": 12187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646287786195543e-05,
+      "loss": 0.6352,
+      "step": 12188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6462333887972747e-05,
+      "loss": 0.6368,
+      "step": 12189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6461789881153053e-05,
+      "loss": 0.6315,
+      "step": 12190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6461245841499112e-05,
+      "loss": 0.6343,
+      "step": 12191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.646070176901369e-05,
+      "loss": 0.7104,
+      "step": 12192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6460157663699548e-05,
+      "loss": 0.6293,
+      "step": 12193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6459613525559457e-05,
+      "loss": 0.6036,
+      "step": 12194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6459069354596176e-05,
+      "loss": 0.6618,
+      "step": 12195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.645852515081247e-05,
+      "loss": 0.6471,
+      "step": 12196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.645798091421111e-05,
+      "loss": 0.6366,
+      "step": 12197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6457436644794854e-05,
+      "loss": 0.5685,
+      "step": 12198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6456892342566476e-05,
+      "loss": 0.5782,
+      "step": 12199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6456348007528732e-05,
+      "loss": 0.5843,
+      "step": 12200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6455803639684392e-05,
+      "loss": 0.6183,
+      "step": 12201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6455259239036227e-05,
+      "loss": 0.5885,
+      "step": 12202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6454714805586997e-05,
+      "loss": 0.6441,
+      "step": 12203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6454170339339475e-05,
+      "loss": 0.6354,
+      "step": 12204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6453625840296423e-05,
+      "loss": 0.6238,
+      "step": 12205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6453081308460602e-05,
+      "loss": 0.5363,
+      "step": 12206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6452536743834787e-05,
+      "loss": 0.6623,
+      "step": 12207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6451992146421743e-05,
+      "loss": 0.6059,
+      "step": 12208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.645144751622424e-05,
+      "loss": 0.6113,
+      "step": 12209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6450902853245044e-05,
+      "loss": 0.5663,
+      "step": 12210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6450358157486918e-05,
+      "loss": 0.6107,
+      "step": 12211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6449813428952635e-05,
+      "loss": 0.6091,
+      "step": 12212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6449268667644956e-05,
+      "loss": 0.6004,
+      "step": 12213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6448723873566658e-05,
+      "loss": 0.6288,
+      "step": 12214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644817904672051e-05,
+      "loss": 0.6013,
+      "step": 12215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644763418710927e-05,
+      "loss": 0.609,
+      "step": 12216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6447089294735712e-05,
+      "loss": 0.5863,
+      "step": 12217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644654436960261e-05,
+      "loss": 0.6196,
+      "step": 12218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6445999411712723e-05,
+      "loss": 0.581,
+      "step": 12219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6445454421068827e-05,
+      "loss": 0.5584,
+      "step": 12220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6444909397673692e-05,
+      "loss": 0.5517,
+      "step": 12221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644436434153008e-05,
+      "loss": 0.6053,
+      "step": 12222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644381925264077e-05,
+      "loss": 0.6136,
+      "step": 12223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6443274131008524e-05,
+      "loss": 0.5952,
+      "step": 12224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6442728976636114e-05,
+      "loss": 0.6092,
+      "step": 12225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6442183789526315e-05,
+      "loss": 0.6313,
+      "step": 12226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.644163856968189e-05,
+      "loss": 0.6067,
+      "step": 12227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6441093317105617e-05,
+      "loss": 0.5776,
+      "step": 12228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6440548031800264e-05,
+      "loss": 0.6238,
+      "step": 12229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6440002713768594e-05,
+      "loss": 0.6022,
+      "step": 12230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6439457363013388e-05,
+      "loss": 0.6119,
+      "step": 12231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.643891197953741e-05,
+      "loss": 0.6166,
+      "step": 12232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6438366563343438e-05,
+      "loss": 0.6063,
+      "step": 12233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6437821114434237e-05,
+      "loss": 0.6055,
+      "step": 12234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6437275632812583e-05,
+      "loss": 0.6617,
+      "step": 12235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6436730118481248e-05,
+      "loss": 0.6492,
+      "step": 12236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6436184571443002e-05,
+      "loss": 0.6215,
+      "step": 12237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6435638991700614e-05,
+      "loss": 0.6088,
+      "step": 12238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.643509337925686e-05,
+      "loss": 0.5882,
+      "step": 12239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6434547734114516e-05,
+      "loss": 0.6023,
+      "step": 12240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6434002056276348e-05,
+      "loss": 0.5669,
+      "step": 12241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.643345634574513e-05,
+      "loss": 0.6581,
+      "step": 12242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6432910602523637e-05,
+      "loss": 0.5771,
+      "step": 12243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6432364826614642e-05,
+      "loss": 0.6088,
+      "step": 12244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6431819018020917e-05,
+      "loss": 0.6403,
+      "step": 12245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6431273176745235e-05,
+      "loss": 0.6215,
+      "step": 12246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6430727302790374e-05,
+      "loss": 0.6171,
+      "step": 12247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.64301813961591e-05,
+      "loss": 0.5926,
+      "step": 12248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6429635456854195e-05,
+      "loss": 0.531,
+      "step": 12249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6429089484878425e-05,
+      "loss": 0.5983,
+      "step": 12250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642854348023457e-05,
+      "loss": 0.6081,
+      "step": 12251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6427997442925405e-05,
+      "loss": 0.7124,
+      "step": 12252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.64274513729537e-05,
+      "loss": 0.6192,
+      "step": 12253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6426905270322234e-05,
+      "loss": 0.6239,
+      "step": 12254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642635913503378e-05,
+      "loss": 0.6437,
+      "step": 12255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642581296709111e-05,
+      "loss": 0.5668,
+      "step": 12256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6425266766497004e-05,
+      "loss": 0.5845,
+      "step": 12257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642472053325424e-05,
+      "loss": 0.564,
+      "step": 12258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6424174267365587e-05,
+      "loss": 0.6085,
+      "step": 12259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642362796883382e-05,
+      "loss": 0.5737,
+      "step": 12260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642308163766172e-05,
+      "loss": 0.5673,
+      "step": 12261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6422535273852064e-05,
+      "loss": 0.5794,
+      "step": 12262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.642198887740762e-05,
+      "loss": 0.6203,
+      "step": 12263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6421442448331174e-05,
+      "loss": 0.6304,
+      "step": 12264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.64208959866255e-05,
+      "loss": 0.6406,
+      "step": 12265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6420349492293367e-05,
+      "loss": 0.5872,
+      "step": 12266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.641980296533756e-05,
+      "loss": 0.5959,
+      "step": 12267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6419256405760854e-05,
+      "loss": 0.5729,
+      "step": 12268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.641870981356603e-05,
+      "loss": 0.5905,
+      "step": 12269
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6418163188755858e-05,
+      "loss": 0.6431,
+      "step": 12270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.641761653133312e-05,
+      "loss": 0.5906,
+      "step": 12271
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6417069841300592e-05,
+      "loss": 0.6266,
+      "step": 12272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6416523118661053e-05,
+      "loss": 0.578,
+      "step": 12273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.641597636341728e-05,
+      "loss": 0.6183,
+      "step": 12274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6415429575572055e-05,
+      "loss": 0.5756,
+      "step": 12275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6414882755128153e-05,
+      "loss": 0.5836,
+      "step": 12276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6414335902088353e-05,
+      "loss": 0.6414,
+      "step": 12277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6413789016455433e-05,
+      "loss": 0.5455,
+      "step": 12278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.641324209823217e-05,
+      "loss": 0.6712,
+      "step": 12279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.641269514742135e-05,
+      "loss": 0.5966,
+      "step": 12280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6412148164025744e-05,
+      "loss": 0.5872,
+      "step": 12281
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.641160114804814e-05,
+      "loss": 0.6138,
+      "step": 12282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6411054099491308e-05,
+      "loss": 0.6223,
+      "step": 12283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6410507018358037e-05,
+      "loss": 0.6129,
+      "step": 12284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.64099599046511e-05,
+      "loss": 0.5606,
+      "step": 12285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640941275837328e-05,
+      "loss": 0.5702,
+      "step": 12286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6408865579527355e-05,
+      "loss": 0.5792,
+      "step": 12287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640831836811611e-05,
+      "loss": 0.6121,
+      "step": 12288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640777112414232e-05,
+      "loss": 0.6159,
+      "step": 12289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6407223847608768e-05,
+      "loss": 0.6036,
+      "step": 12290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640667653851824e-05,
+      "loss": 0.6187,
+      "step": 12291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6406129196873508e-05,
+      "loss": 0.5546,
+      "step": 12292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6405581822677357e-05,
+      "loss": 0.5651,
+      "step": 12293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640503441593257e-05,
+      "loss": 0.5972,
+      "step": 12294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6404486976641928e-05,
+      "loss": 0.6893,
+      "step": 12295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640393950480821e-05,
+      "loss": 0.6365,
+      "step": 12296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.64033920004342e-05,
+      "loss": 0.5894,
+      "step": 12297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6402844463522685e-05,
+      "loss": 0.5841,
+      "step": 12298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640229689407644e-05,
+      "loss": 0.6252,
+      "step": 12299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6401749292098244e-05,
+      "loss": 0.616,
+      "step": 12300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640120165759089e-05,
+      "loss": 0.6063,
+      "step": 12301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6400653990557153e-05,
+      "loss": 0.6244,
+      "step": 12302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.640010629099982e-05,
+      "loss": 0.6181,
+      "step": 12303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6399558558921676e-05,
+      "loss": 0.5973,
+      "step": 12304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6399010794325496e-05,
+      "loss": 0.6631,
+      "step": 12305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639846299721407e-05,
+      "loss": 0.6491,
+      "step": 12306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6397915167590177e-05,
+      "loss": 0.5825,
+      "step": 12307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6397367305456608e-05,
+      "loss": 0.6223,
+      "step": 12308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6396819410816137e-05,
+      "loss": 0.6042,
+      "step": 12309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639627148367156e-05,
+      "loss": 0.5893,
+      "step": 12310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639572352402565e-05,
+      "loss": 0.6011,
+      "step": 12311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6395175531881193e-05,
+      "loss": 0.6128,
+      "step": 12312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639462750724098e-05,
+      "loss": 0.576,
+      "step": 12313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639407945010779e-05,
+      "loss": 0.5458,
+      "step": 12314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639353136048441e-05,
+      "loss": 0.625,
+      "step": 12315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6392983238373625e-05,
+      "loss": 0.5696,
+      "step": 12316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639243508377822e-05,
+      "loss": 0.5539,
+      "step": 12317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6391886896700984e-05,
+      "loss": 0.5634,
+      "step": 12318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6391338677144695e-05,
+      "loss": 0.5822,
+      "step": 12319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.639079042511214e-05,
+      "loss": 0.6243,
+      "step": 12320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6390242140606113e-05,
+      "loss": 0.5649,
+      "step": 12321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.638969382362939e-05,
+      "loss": 0.6039,
+      "step": 12322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6389145474184763e-05,
+      "loss": 0.58,
+      "step": 12323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6388597092275013e-05,
+      "loss": 0.6117,
+      "step": 12324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6388048677902935e-05,
+      "loss": 0.6021,
+      "step": 12325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6387500231071308e-05,
+      "loss": 0.6462,
+      "step": 12326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6386951751782924e-05,
+      "loss": 0.613,
+      "step": 12327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6386403240040566e-05,
+      "loss": 0.6265,
+      "step": 12328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6385854695847022e-05,
+      "loss": 0.6247,
+      "step": 12329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6385306119205085e-05,
+      "loss": 0.637,
+      "step": 12330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6384757510117534e-05,
+      "loss": 0.604,
+      "step": 12331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.638420886858716e-05,
+      "loss": 0.6165,
+      "step": 12332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.638366019461675e-05,
+      "loss": 0.6186,
+      "step": 12333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6383111488209095e-05,
+      "loss": 0.7095,
+      "step": 12334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.638256274936698e-05,
+      "loss": 0.6225,
+      "step": 12335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6382013978093195e-05,
+      "loss": 0.6694,
+      "step": 12336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6381465174390528e-05,
+      "loss": 0.6118,
+      "step": 12337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.638091633826177e-05,
+      "loss": 0.6355,
+      "step": 12338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6380367469709703e-05,
+      "loss": 0.5988,
+      "step": 12339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637981856873712e-05,
+      "loss": 0.6229,
+      "step": 12340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6379269635346813e-05,
+      "loss": 0.6105,
+      "step": 12341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6378720669541568e-05,
+      "loss": 0.6098,
+      "step": 12342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6378171671324174e-05,
+      "loss": 0.5653,
+      "step": 12343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6377622640697423e-05,
+      "loss": 0.5566,
+      "step": 12344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6377073577664106e-05,
+      "loss": 0.6886,
+      "step": 12345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6376524482227004e-05,
+      "loss": 0.617,
+      "step": 12346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637597535438892e-05,
+      "loss": 0.5838,
+      "step": 12347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6375426194152637e-05,
+      "loss": 0.6166,
+      "step": 12348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6374877001520944e-05,
+      "loss": 0.5708,
+      "step": 12349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637432777649664e-05,
+      "loss": 0.642,
+      "step": 12350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.63737785190825e-05,
+      "loss": 0.5806,
+      "step": 12351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637322922928133e-05,
+      "loss": 0.591,
+      "step": 12352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6372679907095917e-05,
+      "loss": 0.5507,
+      "step": 12353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637213055252905e-05,
+      "loss": 0.5566,
+      "step": 12354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637158116558352e-05,
+      "loss": 0.6213,
+      "step": 12355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.637103174626212e-05,
+      "loss": 0.5675,
+      "step": 12356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6370482294567643e-05,
+      "loss": 0.6198,
+      "step": 12357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636993281050288e-05,
+      "loss": 0.5994,
+      "step": 12358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6369383294070625e-05,
+      "loss": 0.654,
+      "step": 12359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6368833745273663e-05,
+      "loss": 0.5931,
+      "step": 12360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6368284164114795e-05,
+      "loss": 0.6088,
+      "step": 12361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636773455059681e-05,
+      "loss": 0.5959,
+      "step": 12362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.63671849047225e-05,
+      "loss": 0.6141,
+      "step": 12363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636663522649466e-05,
+      "loss": 0.5381,
+      "step": 12364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636608551591608e-05,
+      "loss": 0.5998,
+      "step": 12365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6365535772989556e-05,
+      "loss": 0.5945,
+      "step": 12366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636498599771788e-05,
+      "loss": 0.6554,
+      "step": 12367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6364436190103847e-05,
+      "loss": 0.5912,
+      "step": 12368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636388635015025e-05,
+      "loss": 0.569,
+      "step": 12369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636333647785988e-05,
+      "loss": 0.6658,
+      "step": 12370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6362786573235538e-05,
+      "loss": 0.6036,
+      "step": 12371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6362236636280012e-05,
+      "loss": 0.6027,
+      "step": 12372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.63616866669961e-05,
+      "loss": 0.5973,
+      "step": 12373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.636113666538659e-05,
+      "loss": 0.6275,
+      "step": 12374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6360586631454286e-05,
+      "loss": 0.6062,
+      "step": 12375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6360036565201976e-05,
+      "loss": 0.6122,
+      "step": 12376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6359486466632463e-05,
+      "loss": 0.6006,
+      "step": 12377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.635893633574853e-05,
+      "loss": 0.6293,
+      "step": 12378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6358386172552982e-05,
+      "loss": 0.6823,
+      "step": 12379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6357835977048612e-05,
+      "loss": 0.5546,
+      "step": 12380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6357285749238215e-05,
+      "loss": 0.6112,
+      "step": 12381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6356735489124587e-05,
+      "loss": 0.6347,
+      "step": 12382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6356185196710527e-05,
+      "loss": 0.5774,
+      "step": 12383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6355634871998826e-05,
+      "loss": 0.6199,
+      "step": 12384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.635508451499228e-05,
+      "loss": 0.6387,
+      "step": 12385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6354534125693693e-05,
+      "loss": 0.6075,
+      "step": 12386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6353983704105854e-05,
+      "loss": 0.6486,
+      "step": 12387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6353433250231566e-05,
+      "loss": 0.6438,
+      "step": 12388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.635288276407362e-05,
+      "loss": 0.6849,
+      "step": 12389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6352332245634818e-05,
+      "loss": 0.6416,
+      "step": 12390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6351781694917956e-05,
+      "loss": 0.6475,
+      "step": 12391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.635123111192583e-05,
+      "loss": 0.5673,
+      "step": 12392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.635068049666124e-05,
+      "loss": 0.6002,
+      "step": 12393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6350129849126983e-05,
+      "loss": 0.5753,
+      "step": 12394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6349579169325853e-05,
+      "loss": 0.6547,
+      "step": 12395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6349028457260652e-05,
+      "loss": 0.6162,
+      "step": 12396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6348477712934182e-05,
+      "loss": 0.605,
+      "step": 12397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6347926936349235e-05,
+      "loss": 0.6191,
+      "step": 12398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6347376127508613e-05,
+      "loss": 0.5794,
+      "step": 12399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.634682528641511e-05,
+      "loss": 0.5557,
+      "step": 12400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6346274413071533e-05,
+      "loss": 0.6484,
+      "step": 12401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6345723507480676e-05,
+      "loss": 0.5763,
+      "step": 12402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.634517256964534e-05,
+      "loss": 0.6293,
+      "step": 12403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6344621599568325e-05,
+      "loss": 0.5976,
+      "step": 12404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6344070597252428e-05,
+      "loss": 0.6541,
+      "step": 12405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6343519562700452e-05,
+      "loss": 0.665,
+      "step": 12406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6342968495915193e-05,
+      "loss": 0.6267,
+      "step": 12407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6342417396899454e-05,
+      "loss": 0.6139,
+      "step": 12408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.634186626565603e-05,
+      "loss": 0.5951,
+      "step": 12409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6341315102187735e-05,
+      "loss": 0.6022,
+      "step": 12410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6340763906497357e-05,
+      "loss": 0.5717,
+      "step": 12411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6340212678587702e-05,
+      "loss": 0.6294,
+      "step": 12412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6339661418461566e-05,
+      "loss": 0.5752,
+      "step": 12413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6339110126121757e-05,
+      "loss": 0.651,
+      "step": 12414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6338558801571074e-05,
+      "loss": 0.5501,
+      "step": 12415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6338007444812313e-05,
+      "loss": 0.6242,
+      "step": 12416
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6337456055848284e-05,
+      "loss": 0.6217,
+      "step": 12417
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6336904634681784e-05,
+      "loss": 0.5668,
+      "step": 12418
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.633635318131561e-05,
+      "loss": 0.6051,
+      "step": 12419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6335801695752578e-05,
+      "loss": 0.6203,
+      "step": 12420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.633525017799548e-05,
+      "loss": 0.6003,
+      "step": 12421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6334698628047116e-05,
+      "loss": 0.6252,
+      "step": 12422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.63341470459103e-05,
+      "loss": 0.5629,
+      "step": 12423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6333595431587822e-05,
+      "loss": 0.5894,
+      "step": 12424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.633304378508249e-05,
+      "loss": 0.6352,
+      "step": 12425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6332492106397114e-05,
+      "loss": 0.5926,
+      "step": 12426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6331940395534487e-05,
+      "loss": 0.5877,
+      "step": 12427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6331388652497416e-05,
+      "loss": 0.5583,
+      "step": 12428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6330836877288705e-05,
+      "loss": 0.5757,
+      "step": 12429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.633028506991116e-05,
+      "loss": 0.616,
+      "step": 12430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6329733230367582e-05,
+      "loss": 0.6421,
+      "step": 12431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6329181358660774e-05,
+      "loss": 0.6262,
+      "step": 12432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6328629454793542e-05,
+      "loss": 0.6098,
+      "step": 12433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6328077518768694e-05,
+      "loss": 0.5474,
+      "step": 12434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6327525550589028e-05,
+      "loss": 0.6044,
+      "step": 12435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.632697355025735e-05,
+      "loss": 0.6545,
+      "step": 12436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.632642151777647e-05,
+      "loss": 0.5446,
+      "step": 12437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6325869453149187e-05,
+      "loss": 0.5734,
+      "step": 12438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.632531735637831e-05,
+      "loss": 0.5923,
+      "step": 12439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6324765227466643e-05,
+      "loss": 0.5796,
+      "step": 12440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.632421306641699e-05,
+      "loss": 0.5879,
+      "step": 12441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6323660873232162e-05,
+      "loss": 0.5758,
+      "step": 12442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6323108647914958e-05,
+      "loss": 0.5827,
+      "step": 12443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6322556390468184e-05,
+      "loss": 0.6242,
+      "step": 12444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6322004100894653e-05,
+      "loss": 0.6038,
+      "step": 12445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6321451779197167e-05,
+      "loss": 0.5537,
+      "step": 12446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6320899425378535e-05,
+      "loss": 0.6521,
+      "step": 12447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6320347039441563e-05,
+      "loss": 0.5701,
+      "step": 12448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6319794621389055e-05,
+      "loss": 0.5874,
+      "step": 12449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6319242171223817e-05,
+      "loss": 0.5742,
+      "step": 12450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.631868968894866e-05,
+      "loss": 0.5709,
+      "step": 12451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6318137174566392e-05,
+      "loss": 0.5873,
+      "step": 12452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.631758462807982e-05,
+      "loss": 0.6024,
+      "step": 12453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6317032049491746e-05,
+      "loss": 0.5585,
+      "step": 12454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6316479438804983e-05,
+      "loss": 0.5885,
+      "step": 12455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.631592679602234e-05,
+      "loss": 0.6382,
+      "step": 12456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.631537412114662e-05,
+      "loss": 0.6314,
+      "step": 12457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6314821414180638e-05,
+      "loss": 0.5985,
+      "step": 12458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.63142686751272e-05,
+      "loss": 0.6577,
+      "step": 12459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6313715903989113e-05,
+      "loss": 0.5876,
+      "step": 12460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6313163100769186e-05,
+      "loss": 0.5938,
+      "step": 12461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6312610265470227e-05,
+      "loss": 0.6013,
+      "step": 12462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6312057398095045e-05,
+      "loss": 0.5768,
+      "step": 12463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6311504498646455e-05,
+      "loss": 0.5709,
+      "step": 12464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.631095156712726e-05,
+      "loss": 0.6275,
+      "step": 12465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6310398603540272e-05,
+      "loss": 0.6117,
+      "step": 12466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.63098456078883e-05,
+      "loss": 0.6424,
+      "step": 12467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6309292580174158e-05,
+      "loss": 0.5952,
+      "step": 12468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630873952040065e-05,
+      "loss": 0.6173,
+      "step": 12469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6308186428570586e-05,
+      "loss": 0.6166,
+      "step": 12470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630763330468678e-05,
+      "loss": 0.6318,
+      "step": 12471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6307080148752047e-05,
+      "loss": 0.6065,
+      "step": 12472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630652696076919e-05,
+      "loss": 0.5737,
+      "step": 12473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6305973740741018e-05,
+      "loss": 0.5921,
+      "step": 12474
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6305420488670352e-05,
+      "loss": 0.5857,
+      "step": 12475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630486720456e-05,
+      "loss": 0.6458,
+      "step": 12476
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6304313888412765e-05,
+      "loss": 0.5232,
+      "step": 12477
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630376054023147e-05,
+      "loss": 0.6663,
+      "step": 12478
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6303207160018916e-05,
+      "loss": 0.576,
+      "step": 12479
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6302653747777925e-05,
+      "loss": 0.6173,
+      "step": 12480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6302100303511303e-05,
+      "loss": 0.5675,
+      "step": 12481
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6301546827221862e-05,
+      "loss": 0.6515,
+      "step": 12482
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.630099331891242e-05,
+      "loss": 0.6279,
+      "step": 12483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6300439778585784e-05,
+      "loss": 0.6051,
+      "step": 12484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629988620624477e-05,
+      "loss": 0.5816,
+      "step": 12485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6299332601892186e-05,
+      "loss": 0.5723,
+      "step": 12486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6298778965530848e-05,
+      "loss": 0.6178,
+      "step": 12487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629822529716357e-05,
+      "loss": 0.5542,
+      "step": 12488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629767159679317e-05,
+      "loss": 0.5892,
+      "step": 12489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6297117864422446e-05,
+      "loss": 0.605,
+      "step": 12490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629656410005423e-05,
+      "loss": 0.5485,
+      "step": 12491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6296010303691322e-05,
+      "loss": 0.6218,
+      "step": 12492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6295456475336546e-05,
+      "loss": 0.5844,
+      "step": 12493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6294902614992714e-05,
+      "loss": 0.6304,
+      "step": 12494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6294348722662634e-05,
+      "loss": 0.6328,
+      "step": 12495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629379479834913e-05,
+      "loss": 0.6614,
+      "step": 12496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6293240842055004e-05,
+      "loss": 0.5855,
+      "step": 12497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629268685378308e-05,
+      "loss": 0.6335,
+      "step": 12498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6292132833536175e-05,
+      "loss": 0.6113,
+      "step": 12499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.62915787813171e-05,
+      "loss": 0.6073,
+      "step": 12500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6291024697128664e-05,
+      "loss": 0.5644,
+      "step": 12501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6290470580973695e-05,
+      "loss": 0.5818,
+      "step": 12502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6289916432855004e-05,
+      "loss": 0.5358,
+      "step": 12503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6289362252775404e-05,
+      "loss": 0.6296,
+      "step": 12504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628880804073771e-05,
+      "loss": 0.6339,
+      "step": 12505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6288253796744743e-05,
+      "loss": 0.5622,
+      "step": 12506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6287699520799316e-05,
+      "loss": 0.6439,
+      "step": 12507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6287145212904247e-05,
+      "loss": 0.6408,
+      "step": 12508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628659087306235e-05,
+      "loss": 0.6174,
+      "step": 12509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6286036501276448e-05,
+      "loss": 0.643,
+      "step": 12510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6285482097549348e-05,
+      "loss": 0.6124,
+      "step": 12511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6284927661883875e-05,
+      "loss": 0.6401,
+      "step": 12512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6284373194282845e-05,
+      "loss": 0.533,
+      "step": 12513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6283818694749075e-05,
+      "loss": 0.5447,
+      "step": 12514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628326416328538e-05,
+      "loss": 0.5606,
+      "step": 12515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628270959989458e-05,
+      "loss": 0.5957,
+      "step": 12516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628215500457949e-05,
+      "loss": 0.6387,
+      "step": 12517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6281600377342936e-05,
+      "loss": 0.6269,
+      "step": 12518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628104571818773e-05,
+      "loss": 0.6139,
+      "step": 12519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6280491027116686e-05,
+      "loss": 0.6303,
+      "step": 12520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627993630413263e-05,
+      "loss": 0.602,
+      "step": 12521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627938154923838e-05,
+      "loss": 0.6168,
+      "step": 12522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6278826762436753e-05,
+      "loss": 0.5498,
+      "step": 12523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6278271943730568e-05,
+      "loss": 0.6275,
+      "step": 12524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6277717093122648e-05,
+      "loss": 0.6253,
+      "step": 12525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6277162210615803e-05,
+      "loss": 0.6273,
+      "step": 12526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627660729621286e-05,
+      "loss": 0.651,
+      "step": 12527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627605234991664e-05,
+      "loss": 0.5759,
+      "step": 12528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627549737172996e-05,
+      "loss": 0.5874,
+      "step": 12529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6274942361655635e-05,
+      "loss": 0.6623,
+      "step": 12530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6274387319696495e-05,
+      "loss": 0.5647,
+      "step": 12531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6273832245855356e-05,
+      "loss": 0.6388,
+      "step": 12532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6273277140135034e-05,
+      "loss": 0.6133,
+      "step": 12533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6272722002538356e-05,
+      "loss": 0.6027,
+      "step": 12534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6272166833068142e-05,
+      "loss": 0.5985,
+      "step": 12535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627161163172721e-05,
+      "loss": 0.5987,
+      "step": 12536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627105639851838e-05,
+      "loss": 0.676,
+      "step": 12537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.627050113344448e-05,
+      "loss": 0.5978,
+      "step": 12538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626994583650832e-05,
+      "loss": 0.5038,
+      "step": 12539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6269390507712738e-05,
+      "loss": 0.624,
+      "step": 12540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626883514706054e-05,
+      "loss": 0.5959,
+      "step": 12541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626827975455456e-05,
+      "loss": 0.6291,
+      "step": 12542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6267724330197613e-05,
+      "loss": 0.5794,
+      "step": 12543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626716887399252e-05,
+      "loss": 0.5916,
+      "step": 12544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6266613385942112e-05,
+      "loss": 0.6206,
+      "step": 12545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6266057866049202e-05,
+      "loss": 0.5939,
+      "step": 12546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6265502314316618e-05,
+      "loss": 0.6212,
+      "step": 12547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6264946730747185e-05,
+      "loss": 0.6283,
+      "step": 12548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6264391115343717e-05,
+      "loss": 0.6447,
+      "step": 12549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6263835468109042e-05,
+      "loss": 0.5943,
+      "step": 12550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6263279789045993e-05,
+      "loss": 0.6097,
+      "step": 12551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626272407815738e-05,
+      "loss": 0.6839,
+      "step": 12552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626216833544603e-05,
+      "loss": 0.5572,
+      "step": 12553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6261612560914775e-05,
+      "loss": 0.6351,
+      "step": 12554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6261056754566425e-05,
+      "loss": 0.6504,
+      "step": 12555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.626050091640382e-05,
+      "loss": 0.6144,
+      "step": 12556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6259945046429772e-05,
+      "loss": 0.6308,
+      "step": 12557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6259389144647107e-05,
+      "loss": 0.6673,
+      "step": 12558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.625883321105866e-05,
+      "loss": 0.559,
+      "step": 12559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6258277245667243e-05,
+      "loss": 0.5819,
+      "step": 12560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6257721248475688e-05,
+      "loss": 0.5514,
+      "step": 12561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6257165219486818e-05,
+      "loss": 0.6466,
+      "step": 12562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6256609158703458e-05,
+      "loss": 0.5817,
+      "step": 12563
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6256053066128438e-05,
+      "loss": 0.5859,
+      "step": 12564
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6255496941764576e-05,
+      "loss": 0.6291,
+      "step": 12565
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6254940785614705e-05,
+      "loss": 0.626,
+      "step": 12566
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.625438459768165e-05,
+      "loss": 0.6079,
+      "step": 12567
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.625382837796823e-05,
+      "loss": 0.6009,
+      "step": 12568
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6253272126477278e-05,
+      "loss": 0.6043,
+      "step": 12569
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.625271584321162e-05,
+      "loss": 0.5688,
+      "step": 12570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6252159528174084e-05,
+      "loss": 0.5852,
+      "step": 12571
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.625160318136749e-05,
+      "loss": 0.5993,
+      "step": 12572
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6251046802794673e-05,
+      "loss": 0.5775,
+      "step": 12573
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6250490392458454e-05,
+      "loss": 0.6373,
+      "step": 12574
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6249933950361665e-05,
+      "loss": 0.636,
+      "step": 12575
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6249377476507127e-05,
+      "loss": 0.6315,
+      "step": 12576
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6248820970897673e-05,
+      "loss": 0.6337,
+      "step": 12577
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6248264433536135e-05,
+      "loss": 0.5451,
+      "step": 12578
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6247707864425332e-05,
+      "loss": 0.5341,
+      "step": 12579
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6247151263568097e-05,
+      "loss": 0.534,
+      "step": 12580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.624659463096725e-05,
+      "loss": 0.6362,
+      "step": 12581
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6246037966625635e-05,
+      "loss": 0.5893,
+      "step": 12582
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6245481270546067e-05,
+      "loss": 0.6437,
+      "step": 12583
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.624492454273138e-05,
+      "loss": 0.5802,
+      "step": 12584
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6244367783184403e-05,
+      "loss": 0.5786,
+      "step": 12585
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6243810991907968e-05,
+      "loss": 0.5613,
+      "step": 12586
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6243254168904897e-05,
+      "loss": 0.6421,
+      "step": 12587
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6242697314178023e-05,
+      "loss": 0.5567,
+      "step": 12588
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6242140427730176e-05,
+      "loss": 0.5524,
+      "step": 12589
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6241583509564186e-05,
+      "loss": 0.5904,
+      "step": 12590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6241026559682885e-05,
+      "loss": 0.6125,
+      "step": 12591
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6240469578089094e-05,
+      "loss": 0.6278,
+      "step": 12592
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6239912564785655e-05,
+      "loss": 0.6089,
+      "step": 12593
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.623935551977539e-05,
+      "loss": 0.5503,
+      "step": 12594
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.623879844306113e-05,
+      "loss": 0.6398,
+      "step": 12595
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6238241334645714e-05,
+      "loss": 0.6173,
+      "step": 12596
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6237684194531963e-05,
+      "loss": 0.5893,
+      "step": 12597
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6237127022722714e-05,
+      "loss": 0.623,
+      "step": 12598
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6236569819220795e-05,
+      "loss": 0.6613,
+      "step": 12599
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.623601258402904e-05,
+      "loss": 0.5776,
+      "step": 12600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6235455317150275e-05,
+      "loss": 0.6039,
+      "step": 12601
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6234898018587336e-05,
+      "loss": 0.5508,
+      "step": 12602
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6234340688343057e-05,
+      "loss": 0.5786,
+      "step": 12603
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6233783326420264e-05,
+      "loss": 0.5536,
+      "step": 12604
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6233225932821793e-05,
+      "loss": 0.6358,
+      "step": 12605
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6232668507550477e-05,
+      "loss": 0.5565,
+      "step": 12606
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6232111050609148e-05,
+      "loss": 0.5999,
+      "step": 12607
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6231553562000635e-05,
+      "loss": 0.6653,
+      "step": 12608
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6230996041727774e-05,
+      "loss": 0.6065,
+      "step": 12609
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.62304384897934e-05,
+      "loss": 0.65,
+      "step": 12610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.622988090620034e-05,
+      "loss": 0.666,
+      "step": 12611
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.622932329095143e-05,
+      "loss": 0.5933,
+      "step": 12612
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6228765644049508e-05,
+      "loss": 0.6304,
+      "step": 12613
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.62282079654974e-05,
+      "loss": 0.5679,
+      "step": 12614
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6227650255297943e-05,
+      "loss": 0.5935,
+      "step": 12615
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6227092513453975e-05,
+      "loss": 0.617,
+      "step": 12616
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6226534739968322e-05,
+      "loss": 0.5518,
+      "step": 12617
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6225976934843824e-05,
+      "loss": 0.6152,
+      "step": 12618
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6225419098083316e-05,
+      "loss": 0.5977,
+      "step": 12619
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6224861229689627e-05,
+      "loss": 0.6144,
+      "step": 12620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6224303329665597e-05,
+      "loss": 0.5438,
+      "step": 12621
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6223745398014058e-05,
+      "loss": 0.5955,
+      "step": 12622
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6223187434737846e-05,
+      "loss": 0.6273,
+      "step": 12623
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6222629439839797e-05,
+      "loss": 0.6021,
+      "step": 12624
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6222071413322743e-05,
+      "loss": 0.5804,
+      "step": 12625
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6221513355189527e-05,
+      "loss": 0.5991,
+      "step": 12626
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6220955265442975e-05,
+      "loss": 0.6572,
+      "step": 12627
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.622039714408593e-05,
+      "loss": 0.6265,
+      "step": 12628
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6219838991121225e-05,
+      "loss": 0.6505,
+      "step": 12629
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6219280806551695e-05,
+      "loss": 0.5794,
+      "step": 12630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6218722590380177e-05,
+      "loss": 0.5583,
+      "step": 12631
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621816434260951e-05,
+      "loss": 0.6542,
+      "step": 12632
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6217606063242528e-05,
+      "loss": 0.5922,
+      "step": 12633
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6217047752282072e-05,
+      "loss": 0.6782,
+      "step": 12634
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6216489409730975e-05,
+      "loss": 0.6181,
+      "step": 12635
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621593103559207e-05,
+      "loss": 0.6284,
+      "step": 12636
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.62153726298682e-05,
+      "loss": 0.6354,
+      "step": 12637
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6214814192562205e-05,
+      "loss": 0.6288,
+      "step": 12638
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621425572367692e-05,
+      "loss": 0.6122,
+      "step": 12639
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6213697223215176e-05,
+      "loss": 0.6357,
+      "step": 12640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621313869117982e-05,
+      "loss": 0.5954,
+      "step": 12641
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621258012757369e-05,
+      "loss": 0.618,
+      "step": 12642
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6212021532399615e-05,
+      "loss": 0.6786,
+      "step": 12643
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.621146290566044e-05,
+      "loss": 0.5719,
+      "step": 12644
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6210904247359006e-05,
+      "loss": 0.5576,
+      "step": 12645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6210345557498146e-05,
+      "loss": 0.5915,
+      "step": 12646
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.62097868360807e-05,
+      "loss": 0.5875,
+      "step": 12647
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6209228083109515e-05,
+      "loss": 0.6077,
+      "step": 12648
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6208669298587416e-05,
+      "loss": 0.6337,
+      "step": 12649
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6208110482517254e-05,
+      "loss": 0.6129,
+      "step": 12650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6207551634901863e-05,
+      "loss": 0.5686,
+      "step": 12651
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6206992755744087e-05,
+      "loss": 0.6496,
+      "step": 12652
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620643384504676e-05,
+      "loss": 0.6048,
+      "step": 12653
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620587490281273e-05,
+      "loss": 0.6026,
+      "step": 12654
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620531592904483e-05,
+      "loss": 0.5803,
+      "step": 12655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6204756923745896e-05,
+      "loss": 0.5903,
+      "step": 12656
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6204197886918778e-05,
+      "loss": 0.629,
+      "step": 12657
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6203638818566317e-05,
+      "loss": 0.6446,
+      "step": 12658
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620307971869135e-05,
+      "loss": 0.6335,
+      "step": 12659
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620252058729672e-05,
+      "loss": 0.583,
+      "step": 12660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620196142438526e-05,
+      "loss": 0.6133,
+      "step": 12661
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6201402229959822e-05,
+      "loss": 0.6467,
+      "step": 12662
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620084300402324e-05,
+      "loss": 0.5828,
+      "step": 12663
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.620028374657836e-05,
+      "loss": 0.6332,
+      "step": 12664
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6199724457628024e-05,
+      "loss": 0.5914,
+      "step": 12665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619916513717507e-05,
+      "loss": 0.6009,
+      "step": 12666
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619860578522235e-05,
+      "loss": 0.6242,
+      "step": 12667
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619804640177269e-05,
+      "loss": 0.6268,
+      "step": 12668
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6197486986828942e-05,
+      "loss": 0.5654,
+      "step": 12669
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619692754039395e-05,
+      "loss": 0.6424,
+      "step": 12670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6196368062470552e-05,
+      "loss": 0.6176,
+      "step": 12671
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6195808553061598e-05,
+      "loss": 0.627,
+      "step": 12672
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6195249012169923e-05,
+      "loss": 0.5945,
+      "step": 12673
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6194689439798374e-05,
+      "loss": 0.6449,
+      "step": 12674
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6194129835949793e-05,
+      "loss": 0.6179,
+      "step": 12675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6193570200627023e-05,
+      "loss": 0.5782,
+      "step": 12676
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619301053383291e-05,
+      "loss": 0.6034,
+      "step": 12677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6192450835570295e-05,
+      "loss": 0.6192,
+      "step": 12678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6191891105842027e-05,
+      "loss": 0.5851,
+      "step": 12679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619133134465095e-05,
+      "loss": 0.6444,
+      "step": 12680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.61907715519999e-05,
+      "loss": 0.593,
+      "step": 12681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.619021172789173e-05,
+      "loss": 0.631,
+      "step": 12682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6189651872329276e-05,
+      "loss": 0.526,
+      "step": 12683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618909198531539e-05,
+      "loss": 0.5894,
+      "step": 12684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618853206685292e-05,
+      "loss": 0.598,
+      "step": 12685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.61879721169447e-05,
+      "loss": 0.5912,
+      "step": 12686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6187412135593583e-05,
+      "loss": 0.6765,
+      "step": 12687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6186852122802414e-05,
+      "loss": 0.6534,
+      "step": 12688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6186292078574038e-05,
+      "loss": 0.5959,
+      "step": 12689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.61857320029113e-05,
+      "loss": 0.6309,
+      "step": 12690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6185171895817044e-05,
+      "loss": 0.5628,
+      "step": 12691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618461175729412e-05,
+      "loss": 0.5876,
+      "step": 12692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6184051587345374e-05,
+      "loss": 0.562,
+      "step": 12693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618349138597365e-05,
+      "loss": 0.5683,
+      "step": 12694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618293115318179e-05,
+      "loss": 0.5946,
+      "step": 12695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6182370888972653e-05,
+      "loss": 0.6298,
+      "step": 12696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6181810593349075e-05,
+      "loss": 0.6108,
+      "step": 12697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.618125026631391e-05,
+      "loss": 0.6171,
+      "step": 12698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6180689907870003e-05,
+      "loss": 0.6215,
+      "step": 12699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6180129518020194e-05,
+      "loss": 0.5713,
+      "step": 12700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.617956909676734e-05,
+      "loss": 0.5417,
+      "step": 12701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.617900864411429e-05,
+      "loss": 0.5839,
+      "step": 12702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6178448160063883e-05,
+      "loss": 0.6555,
+      "step": 12703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.617788764461897e-05,
+      "loss": 0.6105,
+      "step": 12704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6177327097782404e-05,
+      "loss": 0.5902,
+      "step": 12705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6176766519557028e-05,
+      "loss": 0.6548,
+      "step": 12706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6176205909945696e-05,
+      "loss": 0.6269,
+      "step": 12707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6175645268951248e-05,
+      "loss": 0.6146,
+      "step": 12708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.6175084596576543e-05,
+      "loss": 0.6175,
+      "step": 12709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.617452389282442e-05,
+      "loss": 0.607,
+      "step": 12710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6173963157697735e-05,
+      "loss": 0.6137,
+      "step": 12711
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6173402391199333e-05,
+      "loss": 0.5708,
+      "step": 12712
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.617284159333207e-05,
+      "loss": 0.6807,
+      "step": 12713
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6172280764098785e-05,
+      "loss": 0.595,
+      "step": 12714
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.617171990350234e-05,
+      "loss": 0.5791,
+      "step": 12715
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6171159011545574e-05,
+      "loss": 0.6777,
+      "step": 12716
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.617059808823134e-05,
+      "loss": 0.6034,
+      "step": 12717
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6170037133562497e-05,
+      "loss": 0.5894,
+      "step": 12718
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6169476147541883e-05,
+      "loss": 0.5957,
+      "step": 12719
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6168915130172357e-05,
+      "loss": 0.5731,
+      "step": 12720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6168354081456767e-05,
+      "loss": 0.6338,
+      "step": 12721
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.616779300139796e-05,
+      "loss": 0.5766,
+      "step": 12722
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6167231889998793e-05,
+      "loss": 0.5755,
+      "step": 12723
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6166670747262112e-05,
+      "loss": 0.5955,
+      "step": 12724
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6166109573190772e-05,
+      "loss": 0.6135,
+      "step": 12725
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6165548367787626e-05,
+      "loss": 0.6096,
+      "step": 12726
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.616498713105552e-05,
+      "loss": 0.6456,
+      "step": 12727
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.616442586299731e-05,
+      "loss": 0.6426,
+      "step": 12728
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6163864563615846e-05,
+      "loss": 0.5654,
+      "step": 12729
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6163303232913982e-05,
+      "loss": 0.5682,
+      "step": 12730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6162741870894573e-05,
+      "loss": 0.6528,
+      "step": 12731
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6162180477560463e-05,
+      "loss": 0.5646,
+      "step": 12732
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6161619052914507e-05,
+      "loss": 0.6584,
+      "step": 12733
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6161057596959564e-05,
+      "loss": 0.5456,
+      "step": 12734
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6160496109698486e-05,
+      "loss": 0.6089,
+      "step": 12735
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6159934591134113e-05,
+      "loss": 0.5735,
+      "step": 12736
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6159373041269318e-05,
+      "loss": 0.5672,
+      "step": 12737
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6158811460106944e-05,
+      "loss": 0.6209,
+      "step": 12738
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.615824984764984e-05,
+      "loss": 0.602,
+      "step": 12739
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6157688203900868e-05,
+      "loss": 0.5723,
+      "step": 12740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.615712652886288e-05,
+      "loss": 0.613,
+      "step": 12741
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6156564822538725e-05,
+      "loss": 0.6416,
+      "step": 12742
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6156003084931266e-05,
+      "loss": 0.6356,
+      "step": 12743
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6155441316043348e-05,
+      "loss": 0.5669,
+      "step": 12744
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6154879515877833e-05,
+      "loss": 0.5905,
+      "step": 12745
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6154317684437572e-05,
+      "loss": 0.6698,
+      "step": 12746
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6153755821725417e-05,
+      "loss": 0.6104,
+      "step": 12747
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6153193927744234e-05,
+      "loss": 0.6398,
+      "step": 12748
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6152632002496863e-05,
+      "loss": 0.6285,
+      "step": 12749
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.615207004598617e-05,
+      "loss": 0.5409,
+      "step": 12750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6151508058215007e-05,
+      "loss": 0.648,
+      "step": 12751
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.615094603918623e-05,
+      "loss": 0.6059,
+      "step": 12752
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6150383988902696e-05,
+      "loss": 0.5854,
+      "step": 12753
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614982190736726e-05,
+      "loss": 0.6688,
+      "step": 12754
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6149259794582776e-05,
+      "loss": 0.5789,
+      "step": 12755
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6148697650552104e-05,
+      "loss": 0.5566,
+      "step": 12756
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6148135475278096e-05,
+      "loss": 0.5962,
+      "step": 12757
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6147573268763612e-05,
+      "loss": 0.6292,
+      "step": 12758
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614701103101151e-05,
+      "loss": 0.6483,
+      "step": 12759
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6146448762024642e-05,
+      "loss": 0.5957,
+      "step": 12760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614588646180587e-05,
+      "loss": 0.5466,
+      "step": 12761
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6145324130358047e-05,
+      "loss": 0.6167,
+      "step": 12762
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614476176768403e-05,
+      "loss": 0.592,
+      "step": 12763
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6144199373786686e-05,
+      "loss": 0.6094,
+      "step": 12764
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614363694866886e-05,
+      "loss": 0.6236,
+      "step": 12765
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6143074492333415e-05,
+      "loss": 0.6273,
+      "step": 12766
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614251200478321e-05,
+      "loss": 0.6152,
+      "step": 12767
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6141949486021108e-05,
+      "loss": 0.6663,
+      "step": 12768
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6141386936049957e-05,
+      "loss": 0.6171,
+      "step": 12769
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614082435487262e-05,
+      "loss": 0.5947,
+      "step": 12770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.614026174249196e-05,
+      "loss": 0.6223,
+      "step": 12771
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6139699098910827e-05,
+      "loss": 0.6033,
+      "step": 12772
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6139136424132088e-05,
+      "loss": 0.638,
+      "step": 12773
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6138573718158596e-05,
+      "loss": 0.6787,
+      "step": 12774
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6138010980993218e-05,
+      "loss": 0.6094,
+      "step": 12775
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6137448212638802e-05,
+      "loss": 0.6188,
+      "step": 12776
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.613688541309822e-05,
+      "loss": 0.6406,
+      "step": 12777
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6136322582374325e-05,
+      "loss": 0.5956,
+      "step": 12778
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6135759720469976e-05,
+      "loss": 0.6314,
+      "step": 12779
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6135196827388034e-05,
+      "loss": 0.6342,
+      "step": 12780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6134633903131363e-05,
+      "loss": 0.5604,
+      "step": 12781
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6134070947702816e-05,
+      "loss": 0.6082,
+      "step": 12782
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6133507961105263e-05,
+      "loss": 0.6594,
+      "step": 12783
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.613294494334156e-05,
+      "loss": 0.598,
+      "step": 12784
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6132381894414564e-05,
+      "loss": 0.6393,
+      "step": 12785
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.613181881432714e-05,
+      "loss": 0.5998,
+      "step": 12786
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.613125570308215e-05,
+      "loss": 0.6207,
+      "step": 12787
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6130692560682454e-05,
+      "loss": 0.5782,
+      "step": 12788
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6130129387130916e-05,
+      "loss": 0.6255,
+      "step": 12789
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.612956618243039e-05,
+      "loss": 0.5949,
+      "step": 12790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.612900294658375e-05,
+      "loss": 0.6455,
+      "step": 12791
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6128439679593846e-05,
+      "loss": 0.6507,
+      "step": 12792
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6127876381463544e-05,
+      "loss": 0.64,
+      "step": 12793
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.612731305219571e-05,
+      "loss": 0.6171,
+      "step": 12794
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6126749691793207e-05,
+      "loss": 0.5855,
+      "step": 12795
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6126186300258892e-05,
+      "loss": 0.6142,
+      "step": 12796
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.612562287759563e-05,
+      "loss": 0.6763,
+      "step": 12797
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6125059423806285e-05,
+      "loss": 0.5976,
+      "step": 12798
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6124495938893715e-05,
+      "loss": 0.5914,
+      "step": 12799
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6123932422860793e-05,
+      "loss": 0.591,
+      "step": 12800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6123368875710374e-05,
+      "loss": 0.6092,
+      "step": 12801
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6122805297445324e-05,
+      "loss": 0.5787,
+      "step": 12802
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6122241688068507e-05,
+      "loss": 0.6804,
+      "step": 12803
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6121678047582786e-05,
+      "loss": 0.6495,
+      "step": 12804
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6121114375991033e-05,
+      "loss": 0.6683,
+      "step": 12805
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6120550673296098e-05,
+      "loss": 0.5855,
+      "step": 12806
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6119986939500854e-05,
+      "loss": 0.5879,
+      "step": 12807
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6119423174608168e-05,
+      "loss": 0.7191,
+      "step": 12808
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6118859378620895e-05,
+      "loss": 0.593,
+      "step": 12809
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611829555154191e-05,
+      "loss": 0.6289,
+      "step": 12810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611773169337407e-05,
+      "loss": 0.6152,
+      "step": 12811
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6117167804120244e-05,
+      "loss": 0.5815,
+      "step": 12812
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.61166038837833e-05,
+      "loss": 0.608,
+      "step": 12813
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6116039932366095e-05,
+      "loss": 0.5872,
+      "step": 12814
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6115475949871505e-05,
+      "loss": 0.6713,
+      "step": 12815
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611491193630239e-05,
+      "loss": 0.6533,
+      "step": 12816
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6114347891661613e-05,
+      "loss": 0.5695,
+      "step": 12817
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611378381595205e-05,
+      "loss": 0.6035,
+      "step": 12818
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611321970917656e-05,
+      "loss": 0.6318,
+      "step": 12819
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6112655571338007e-05,
+      "loss": 0.6008,
+      "step": 12820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611209140243926e-05,
+      "loss": 0.6134,
+      "step": 12821
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6111527202483186e-05,
+      "loss": 0.6502,
+      "step": 12822
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6110962971472653e-05,
+      "loss": 0.6175,
+      "step": 12823
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.611039870941053e-05,
+      "loss": 0.639,
+      "step": 12824
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610983441629968e-05,
+      "loss": 0.618,
+      "step": 12825
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610927009214297e-05,
+      "loss": 0.6516,
+      "step": 12826
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6108705736943276e-05,
+      "loss": 0.5727,
+      "step": 12827
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6108141350703456e-05,
+      "loss": 0.5627,
+      "step": 12828
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610757693342638e-05,
+      "loss": 0.5999,
+      "step": 12829
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6107012485114918e-05,
+      "loss": 0.6247,
+      "step": 12830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6106448005771933e-05,
+      "loss": 0.6117,
+      "step": 12831
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6105883495400302e-05,
+      "loss": 0.585,
+      "step": 12832
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6105318954002887e-05,
+      "loss": 0.6013,
+      "step": 12833
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6104754381582555e-05,
+      "loss": 0.5096,
+      "step": 12834
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6104189778142186e-05,
+      "loss": 0.6236,
+      "step": 12835
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610362514368464e-05,
+      "loss": 0.6066,
+      "step": 12836
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610306047821278e-05,
+      "loss": 0.5776,
+      "step": 12837
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.610249578172949e-05,
+      "loss": 0.5653,
+      "step": 12838
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6101931054237624e-05,
+      "loss": 0.5863,
+      "step": 12839
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6101366295740064e-05,
+      "loss": 0.6568,
+      "step": 12840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6100801506239672e-05,
+      "loss": 0.644,
+      "step": 12841
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6100236685739325e-05,
+      "loss": 0.7424,
+      "step": 12842
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6099671834241884e-05,
+      "loss": 0.6106,
+      "step": 12843
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6099106951750227e-05,
+      "loss": 0.6246,
+      "step": 12844
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6098542038267224e-05,
+      "loss": 0.6008,
+      "step": 12845
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6097977093795737e-05,
+      "loss": 0.6665,
+      "step": 12846
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.609741211833865e-05,
+      "loss": 0.6206,
+      "step": 12847
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6096847111898822e-05,
+      "loss": 0.6172,
+      "step": 12848
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6096282074479127e-05,
+      "loss": 0.6697,
+      "step": 12849
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6095717006082437e-05,
+      "loss": 0.6911,
+      "step": 12850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6095151906711627e-05,
+      "loss": 0.6612,
+      "step": 12851
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6094586776369564e-05,
+      "loss": 0.5909,
+      "step": 12852
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.609402161505912e-05,
+      "loss": 0.6521,
+      "step": 12853
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6093456422783167e-05,
+      "loss": 0.5911,
+      "step": 12854
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6092891199544576e-05,
+      "loss": 0.6328,
+      "step": 12855
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6092325945346225e-05,
+      "loss": 0.5741,
+      "step": 12856
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6091760660190982e-05,
+      "loss": 0.6104,
+      "step": 12857
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6091195344081713e-05,
+      "loss": 0.5542,
+      "step": 12858
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60906299970213e-05,
+      "loss": 0.5759,
+      "step": 12859
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.609006461901261e-05,
+      "loss": 0.5845,
+      "step": 12860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6089499210058523e-05,
+      "loss": 0.6007,
+      "step": 12861
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6088933770161903e-05,
+      "loss": 0.607,
+      "step": 12862
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6088368299325627e-05,
+      "loss": 0.6024,
+      "step": 12863
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6087802797552573e-05,
+      "loss": 0.6491,
+      "step": 12864
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6087237264845605e-05,
+      "loss": 0.6078,
+      "step": 12865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6086671701207603e-05,
+      "loss": 0.6117,
+      "step": 12866
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6086106106641443e-05,
+      "loss": 0.5855,
+      "step": 12867
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.608554048114999e-05,
+      "loss": 0.5762,
+      "step": 12868
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6084974824736124e-05,
+      "loss": 0.6028,
+      "step": 12869
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6084409137402726e-05,
+      "loss": 0.6222,
+      "step": 12870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6083843419152656e-05,
+      "loss": 0.6826,
+      "step": 12871
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60832776699888e-05,
+      "loss": 0.6339,
+      "step": 12872
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6082711889914025e-05,
+      "loss": 0.5844,
+      "step": 12873
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.608214607893121e-05,
+      "loss": 0.639,
+      "step": 12874
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6081580237043235e-05,
+      "loss": 0.5419,
+      "step": 12875
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6081014364252965e-05,
+      "loss": 0.6203,
+      "step": 12876
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.608044846056328e-05,
+      "loss": 0.6424,
+      "step": 12877
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6079882525977058e-05,
+      "loss": 0.6137,
+      "step": 12878
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6079316560497172e-05,
+      "loss": 0.5809,
+      "step": 12879
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6078750564126493e-05,
+      "loss": 0.6042,
+      "step": 12880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6078184536867908e-05,
+      "loss": 0.6047,
+      "step": 12881
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6077618478724285e-05,
+      "loss": 0.6352,
+      "step": 12882
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6077052389698505e-05,
+      "loss": 0.6037,
+      "step": 12883
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6076486269793437e-05,
+      "loss": 0.5946,
+      "step": 12884
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.607592011901197e-05,
+      "loss": 0.5563,
+      "step": 12885
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6075353937356966e-05,
+      "loss": 0.6411,
+      "step": 12886
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6074787724831313e-05,
+      "loss": 0.6345,
+      "step": 12887
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6074221481437884e-05,
+      "loss": 0.5991,
+      "step": 12888
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6073655207179558e-05,
+      "loss": 0.6343,
+      "step": 12889
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6073088902059207e-05,
+      "loss": 0.5841,
+      "step": 12890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6072522566079717e-05,
+      "loss": 0.588,
+      "step": 12891
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.607195619924396e-05,
+      "loss": 0.5894,
+      "step": 12892
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6071389801554816e-05,
+      "loss": 0.6044,
+      "step": 12893
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.607082337301516e-05,
+      "loss": 0.5892,
+      "step": 12894
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6070256913627874e-05,
+      "loss": 0.5886,
+      "step": 12895
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6069690423395838e-05,
+      "loss": 0.6361,
+      "step": 12896
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6069123902321923e-05,
+      "loss": 0.6098,
+      "step": 12897
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6068557350409016e-05,
+      "loss": 0.5946,
+      "step": 12898
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6067990767659985e-05,
+      "loss": 0.6502,
+      "step": 12899
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6067424154077723e-05,
+      "loss": 0.5718,
+      "step": 12900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6066857509665097e-05,
+      "loss": 0.61,
+      "step": 12901
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6066290834424996e-05,
+      "loss": 0.6914,
+      "step": 12902
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.606572412836029e-05,
+      "loss": 0.6305,
+      "step": 12903
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6065157391473863e-05,
+      "loss": 0.62,
+      "step": 12904
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60645906237686e-05,
+      "loss": 0.6341,
+      "step": 12905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6064023825247375e-05,
+      "loss": 0.5443,
+      "step": 12906
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6063456995913065e-05,
+      "loss": 0.5857,
+      "step": 12907
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6062890135768557e-05,
+      "loss": 0.6286,
+      "step": 12908
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6062323244816727e-05,
+      "loss": 0.6282,
+      "step": 12909
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.606175632306046e-05,
+      "loss": 0.6175,
+      "step": 12910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.606118937050263e-05,
+      "loss": 0.5966,
+      "step": 12911
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6060622387146126e-05,
+      "loss": 0.6443,
+      "step": 12912
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.606005537299382e-05,
+      "loss": 0.5526,
+      "step": 12913
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60594883280486e-05,
+      "loss": 0.6539,
+      "step": 12914
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6058921252313348e-05,
+      "loss": 0.5988,
+      "step": 12915
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605835414579094e-05,
+      "loss": 0.6037,
+      "step": 12916
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605778700848426e-05,
+      "loss": 0.7017,
+      "step": 12917
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605721984039619e-05,
+      "loss": 0.6599,
+      "step": 12918
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6056652641529615e-05,
+      "loss": 0.6618,
+      "step": 12919
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6056085411887412e-05,
+      "loss": 0.6314,
+      "step": 12920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6055518151472464e-05,
+      "loss": 0.6162,
+      "step": 12921
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6054950860287655e-05,
+      "loss": 0.5875,
+      "step": 12922
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605438353833587e-05,
+      "loss": 0.5534,
+      "step": 12923
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605381618561999e-05,
+      "loss": 0.6043,
+      "step": 12924
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6053248802142893e-05,
+      "loss": 0.6006,
+      "step": 12925
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6052681387907467e-05,
+      "loss": 0.624,
+      "step": 12926
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6052113942916596e-05,
+      "loss": 0.6862,
+      "step": 12927
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.605154646717316e-05,
+      "loss": 0.6337,
+      "step": 12928
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6050978960680045e-05,
+      "loss": 0.6152,
+      "step": 12929
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6050411423440134e-05,
+      "loss": 0.6032,
+      "step": 12930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6049843855456308e-05,
+      "loss": 0.5658,
+      "step": 12931
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6049276256731458e-05,
+      "loss": 0.5149,
+      "step": 12932
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.604870862726846e-05,
+      "loss": 0.6227,
+      "step": 12933
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6048140967070204e-05,
+      "loss": 0.6328,
+      "step": 12934
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6047573276139573e-05,
+      "loss": 0.5951,
+      "step": 12935
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.604700555447945e-05,
+      "loss": 0.6683,
+      "step": 12936
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6046437802092724e-05,
+      "loss": 0.5974,
+      "step": 12937
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.604587001898227e-05,
+      "loss": 0.6197,
+      "step": 12938
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6045302205150983e-05,
+      "loss": 0.5974,
+      "step": 12939
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.604473436060175e-05,
+      "loss": 0.6458,
+      "step": 12940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6044166485337443e-05,
+      "loss": 0.6003,
+      "step": 12941
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.604359857936096e-05,
+      "loss": 0.6681,
+      "step": 12942
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6043030642675184e-05,
+      "loss": 0.5394,
+      "step": 12943
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6042462675282997e-05,
+      "loss": 0.589,
+      "step": 12944
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6041894677187286e-05,
+      "loss": 0.668,
+      "step": 12945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6041326648390944e-05,
+      "loss": 0.575,
+      "step": 12946
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6040758588896847e-05,
+      "loss": 0.5981,
+      "step": 12947
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6040190498707892e-05,
+      "loss": 0.6468,
+      "step": 12948
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6039622377826956e-05,
+      "loss": 0.5865,
+      "step": 12949
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6039054226256928e-05,
+      "loss": 0.6568,
+      "step": 12950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6038486044000703e-05,
+      "loss": 0.6238,
+      "step": 12951
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.603791783106116e-05,
+      "loss": 0.6075,
+      "step": 12952
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6037349587441185e-05,
+      "loss": 0.5734,
+      "step": 12953
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.603678131314367e-05,
+      "loss": 0.6014,
+      "step": 12954
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6036213008171503e-05,
+      "loss": 0.64,
+      "step": 12955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6035644672527565e-05,
+      "loss": 0.6221,
+      "step": 12956
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6035076306214756e-05,
+      "loss": 0.5973,
+      "step": 12957
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6034507909235953e-05,
+      "loss": 0.6223,
+      "step": 12958
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.603393948159405e-05,
+      "loss": 0.6595,
+      "step": 12959
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6033371023291927e-05,
+      "loss": 0.6221,
+      "step": 12960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6032802534332485e-05,
+      "loss": 0.5411,
+      "step": 12961
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6032234014718605e-05,
+      "loss": 0.6799,
+      "step": 12962
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6031665464453178e-05,
+      "loss": 0.5996,
+      "step": 12963
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6031096883539094e-05,
+      "loss": 0.6208,
+      "step": 12964
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6030528271979235e-05,
+      "loss": 0.5891,
+      "step": 12965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60299596297765e-05,
+      "loss": 0.6215,
+      "step": 12966
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6029390956933773e-05,
+      "loss": 0.6598,
+      "step": 12967
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6028822253453947e-05,
+      "loss": 0.6056,
+      "step": 12968
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6028253519339903e-05,
+      "loss": 0.6532,
+      "step": 12969
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6027684754594543e-05,
+      "loss": 0.6072,
+      "step": 12970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6027115959220752e-05,
+      "loss": 0.6064,
+      "step": 12971
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6026547133221418e-05,
+      "loss": 0.6355,
+      "step": 12972
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.602597827659943e-05,
+      "loss": 0.6656,
+      "step": 12973
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6025409389357687e-05,
+      "loss": 0.6577,
+      "step": 12974
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6024840471499073e-05,
+      "loss": 0.6052,
+      "step": 12975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.602427152302648e-05,
+      "loss": 0.6638,
+      "step": 12976
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6023702543942796e-05,
+      "loss": 0.5838,
+      "step": 12977
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6023133534250916e-05,
+      "loss": 0.5541,
+      "step": 12978
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.602256449395373e-05,
+      "loss": 0.6027,
+      "step": 12979
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6021995423054132e-05,
+      "loss": 0.6562,
+      "step": 12980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6021426321555015e-05,
+      "loss": 0.5896,
+      "step": 12981
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.602085718945926e-05,
+      "loss": 0.5855,
+      "step": 12982
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6020288026769772e-05,
+      "loss": 0.5795,
+      "step": 12983
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6019718833489437e-05,
+      "loss": 0.6123,
+      "step": 12984
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6019149609621147e-05,
+      "loss": 0.5674,
+      "step": 12985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6018580355167794e-05,
+      "loss": 0.6042,
+      "step": 12986
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601801107013227e-05,
+      "loss": 0.5968,
+      "step": 12987
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601744175451747e-05,
+      "loss": 0.6159,
+      "step": 12988
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601687240832629e-05,
+      "loss": 0.62,
+      "step": 12989
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6016303031561616e-05,
+      "loss": 0.6652,
+      "step": 12990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6015733624226345e-05,
+      "loss": 0.5709,
+      "step": 12991
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6015164186323373e-05,
+      "loss": 0.6657,
+      "step": 12992
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6014594717855582e-05,
+      "loss": 0.6271,
+      "step": 12993
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601402521882588e-05,
+      "loss": 0.6192,
+      "step": 12994
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6013455689237152e-05,
+      "loss": 0.6239,
+      "step": 12995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6012886129092296e-05,
+      "loss": 0.6216,
+      "step": 12996
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6012316538394202e-05,
+      "loss": 0.5733,
+      "step": 12997
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601174691714577e-05,
+      "loss": 0.6737,
+      "step": 12998
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601117726534989e-05,
+      "loss": 0.6028,
+      "step": 12999
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6010607583009462e-05,
+      "loss": 0.586,
+      "step": 13000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.601003787012737e-05,
+      "loss": 0.5864,
+      "step": 13001
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.600946812670652e-05,
+      "loss": 0.5915,
+      "step": 13002
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.60088983527498e-05,
+      "loss": 0.6243,
+      "step": 13003
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.6008328548260107e-05,
+      "loss": 0.6926,
+      "step": 13004
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600775871324034e-05,
+      "loss": 0.6421,
+      "step": 13005
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6007188847693388e-05,
+      "loss": 0.5859,
+      "step": 13006
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6006618951622152e-05,
+      "loss": 0.697,
+      "step": 13007
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600604902502953e-05,
+      "loss": 0.6226,
+      "step": 13008
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600547906791841e-05,
+      "loss": 0.6037,
+      "step": 13009
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600490908029169e-05,
+      "loss": 0.6599,
+      "step": 13010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600433906215227e-05,
+      "loss": 0.5844,
+      "step": 13011
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6003769013503046e-05,
+      "loss": 0.596,
+      "step": 13012
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6003198934346913e-05,
+      "loss": 0.5523,
+      "step": 13013
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600262882468677e-05,
+      "loss": 0.5925,
+      "step": 13014
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6002058684525508e-05,
+      "loss": 0.5503,
+      "step": 13015
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.600148851386603e-05,
+      "loss": 0.5272,
+      "step": 13016
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6000918312711232e-05,
+      "loss": 0.6164,
+      "step": 13017
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6000348081064005e-05,
+      "loss": 0.6691,
+      "step": 13018
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599977781892726e-05,
+      "loss": 0.6145,
+      "step": 13019
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5999207526303885e-05,
+      "loss": 0.6308,
+      "step": 13020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5998637203196778e-05,
+      "loss": 0.5976,
+      "step": 13021
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5998066849608838e-05,
+      "loss": 0.6175,
+      "step": 13022
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599749646554296e-05,
+      "loss": 0.6331,
+      "step": 13023
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5996926051002053e-05,
+      "loss": 0.6214,
+      "step": 13024
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5996355605989005e-05,
+      "loss": 0.6299,
+      "step": 13025
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599578513050672e-05,
+      "loss": 0.6825,
+      "step": 13026
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5995214624558094e-05,
+      "loss": 0.4868,
+      "step": 13027
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5994644088146024e-05,
+      "loss": 0.6139,
+      "step": 13028
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599407352127342e-05,
+      "loss": 0.529,
+      "step": 13029
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5993502923943164e-05,
+      "loss": 0.6168,
+      "step": 13030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5992932296158167e-05,
+      "loss": 0.6422,
+      "step": 13031
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599236163792133e-05,
+      "loss": 0.6398,
+      "step": 13032
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5991790949235544e-05,
+      "loss": 0.6196,
+      "step": 13033
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5991220230103715e-05,
+      "loss": 0.642,
+      "step": 13034
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599064948052874e-05,
+      "loss": 0.6716,
+      "step": 13035
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.599007870051352e-05,
+      "loss": 0.6348,
+      "step": 13036
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5989507890060956e-05,
+      "loss": 0.622,
+      "step": 13037
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5988937049173952e-05,
+      "loss": 0.5878,
+      "step": 13038
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5988366177855402e-05,
+      "loss": 0.6258,
+      "step": 13039
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598779527610821e-05,
+      "loss": 0.6033,
+      "step": 13040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5987224343935276e-05,
+      "loss": 0.5811,
+      "step": 13041
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5986653381339503e-05,
+      "loss": 0.6011,
+      "step": 13042
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598608238832379e-05,
+      "loss": 0.6236,
+      "step": 13043
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598551136489104e-05,
+      "loss": 0.5415,
+      "step": 13044
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5984940311044152e-05,
+      "loss": 0.5839,
+      "step": 13045
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5984369226786035e-05,
+      "loss": 0.6861,
+      "step": 13046
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5983798112119578e-05,
+      "loss": 0.5729,
+      "step": 13047
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5983226967047695e-05,
+      "loss": 0.6421,
+      "step": 13048
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598265579157328e-05,
+      "loss": 0.5696,
+      "step": 13049
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598208458569924e-05,
+      "loss": 0.5881,
+      "step": 13050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.598151334942848e-05,
+      "loss": 0.5894,
+      "step": 13051
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5980942082763897e-05,
+      "loss": 0.623,
+      "step": 13052
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5980370785708394e-05,
+      "loss": 0.5765,
+      "step": 13053
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597979945826488e-05,
+      "loss": 0.6395,
+      "step": 13054
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597922810043625e-05,
+      "loss": 0.5983,
+      "step": 13055
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597865671222541e-05,
+      "loss": 0.7054,
+      "step": 13056
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5978085293635265e-05,
+      "loss": 0.5613,
+      "step": 13057
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5977513844668715e-05,
+      "loss": 0.6491,
+      "step": 13058
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5976942365328672e-05,
+      "loss": 0.6351,
+      "step": 13059
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5976370855618032e-05,
+      "loss": 0.5819,
+      "step": 13060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5975799315539703e-05,
+      "loss": 0.6226,
+      "step": 13061
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5975227745096584e-05,
+      "loss": 0.634,
+      "step": 13062
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5974656144291586e-05,
+      "loss": 0.7206,
+      "step": 13063
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597408451312761e-05,
+      "loss": 0.5958,
+      "step": 13064
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597351285160756e-05,
+      "loss": 0.5753,
+      "step": 13065
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597294115973434e-05,
+      "loss": 0.6291,
+      "step": 13066
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5972369437510862e-05,
+      "loss": 0.587,
+      "step": 13067
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597179768494002e-05,
+      "loss": 0.5473,
+      "step": 13068
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597122590202473e-05,
+      "loss": 0.602,
+      "step": 13069
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.597065408876789e-05,
+      "loss": 0.6351,
+      "step": 13070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5970082245172408e-05,
+      "loss": 0.6356,
+      "step": 13071
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.596951037124119e-05,
+      "loss": 0.6088,
+      "step": 13072
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.596893846697714e-05,
+      "loss": 0.615,
+      "step": 13073
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5968366532383167e-05,
+      "loss": 0.5894,
+      "step": 13074
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5967794567462178e-05,
+      "loss": 0.534,
+      "step": 13075
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5967222572217074e-05,
+      "loss": 0.5987,
+      "step": 13076
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5966650546650765e-05,
+      "loss": 0.5568,
+      "step": 13077
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.596607849076616e-05,
+      "loss": 0.5764,
+      "step": 13078
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.596550640456616e-05,
+      "loss": 0.608,
+      "step": 13079
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5964934288053672e-05,
+      "loss": 0.6485,
+      "step": 13080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5964362141231612e-05,
+      "loss": 0.6699,
+      "step": 13081
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.596378996410288e-05,
+      "loss": 0.6257,
+      "step": 13082
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5963217756670378e-05,
+      "loss": 0.6284,
+      "step": 13083
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5962645518937027e-05,
+      "loss": 0.5552,
+      "step": 13084
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5962073250905725e-05,
+      "loss": 0.5817,
+      "step": 13085
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5961500952579388e-05,
+      "loss": 0.5511,
+      "step": 13086
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5960928623960913e-05,
+      "loss": 0.6474,
+      "step": 13087
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5960356265053215e-05,
+      "loss": 0.6103,
+      "step": 13088
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.59597838758592e-05,
+      "loss": 0.6395,
+      "step": 13089
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5959211456381784e-05,
+      "loss": 0.5831,
+      "step": 13090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5958639006623862e-05,
+      "loss": 0.6178,
+      "step": 13091
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.595806652658835e-05,
+      "loss": 0.6081,
+      "step": 13092
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5957494016278162e-05,
+      "loss": 0.6122,
+      "step": 13093
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5956921475696204e-05,
+      "loss": 0.5638,
+      "step": 13094
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5956348904845378e-05,
+      "loss": 0.6114,
+      "step": 13095
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5955776303728598e-05,
+      "loss": 0.5756,
+      "step": 13096
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5955203672348778e-05,
+      "loss": 0.5872,
+      "step": 13097
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5954631010708824e-05,
+      "loss": 0.5719,
+      "step": 13098
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5954058318811647e-05,
+      "loss": 0.5855,
+      "step": 13099
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5953485596660153e-05,
+      "loss": 0.5856,
+      "step": 13100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5952912844257255e-05,
+      "loss": 0.5669,
+      "step": 13101
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5952340061605864e-05,
+      "loss": 0.5505,
+      "step": 13102
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5951767248708892e-05,
+      "loss": 0.5858,
+      "step": 13103
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5951194405569242e-05,
+      "loss": 0.6661,
+      "step": 13104
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.595062153218984e-05,
+      "loss": 0.5204,
+      "step": 13105
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5950048628573578e-05,
+      "loss": 0.6475,
+      "step": 13106
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594947569472338e-05,
+      "loss": 0.6197,
+      "step": 13107
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5948902730642155e-05,
+      "loss": 0.5635,
+      "step": 13108
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594832973633281e-05,
+      "loss": 0.6101,
+      "step": 13109
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5947756711798262e-05,
+      "loss": 0.6317,
+      "step": 13110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594718365704142e-05,
+      "loss": 0.6218,
+      "step": 13111
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5946610572065197e-05,
+      "loss": 0.6128,
+      "step": 13112
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.59460374568725e-05,
+      "loss": 0.6343,
+      "step": 13113
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594546431146625e-05,
+      "loss": 0.651,
+      "step": 13114
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594489113584935e-05,
+      "loss": 0.6209,
+      "step": 13115
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594431793002472e-05,
+      "loss": 0.6514,
+      "step": 13116
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5943744693995268e-05,
+      "loss": 0.5738,
+      "step": 13117
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594317142776391e-05,
+      "loss": 0.6288,
+      "step": 13118
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.594259813133356e-05,
+      "loss": 0.6113,
+      "step": 13119
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5942024804707124e-05,
+      "loss": 0.608,
+      "step": 13120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5941451447887525e-05,
+      "loss": 0.5559,
+      "step": 13121
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5940878060877666e-05,
+      "loss": 0.6278,
+      "step": 13122
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5940304643680464e-05,
+      "loss": 0.617,
+      "step": 13123
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593973119629884e-05,
+      "loss": 0.6556,
+      "step": 13124
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.59391577187357e-05,
+      "loss": 0.6173,
+      "step": 13125
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593858421099396e-05,
+      "loss": 0.6172,
+      "step": 13126
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5938010673076537e-05,
+      "loss": 0.6643,
+      "step": 13127
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5937437104986338e-05,
+      "loss": 0.5405,
+      "step": 13128
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5936863506726285e-05,
+      "loss": 0.5917,
+      "step": 13129
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593628987829929e-05,
+      "loss": 0.59,
+      "step": 13130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593571621970827e-05,
+      "loss": 0.614,
+      "step": 13131
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5935142530956133e-05,
+      "loss": 0.5872,
+      "step": 13132
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5934568812045798e-05,
+      "loss": 0.6229,
+      "step": 13133
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5933995062980186e-05,
+      "loss": 0.6155,
+      "step": 13134
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5933421283762203e-05,
+      "loss": 0.5693,
+      "step": 13135
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593284747439477e-05,
+      "loss": 0.5458,
+      "step": 13136
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5932273634880802e-05,
+      "loss": 0.6356,
+      "step": 13137
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5931699765223215e-05,
+      "loss": 0.6235,
+      "step": 13138
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.593112586542492e-05,
+      "loss": 0.6164,
+      "step": 13139
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5930551935488844e-05,
+      "loss": 0.658,
+      "step": 13140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5929977975417892e-05,
+      "loss": 0.5934,
+      "step": 13141
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5929403985214987e-05,
+      "loss": 0.577,
+      "step": 13142
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5928829964883045e-05,
+      "loss": 0.6158,
+      "step": 13143
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.592825591442498e-05,
+      "loss": 0.5891,
+      "step": 13144
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.592768183384371e-05,
+      "loss": 0.5705,
+      "step": 13145
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5927107723142152e-05,
+      "loss": 0.5929,
+      "step": 13146
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5926533582323226e-05,
+      "loss": 0.626,
+      "step": 13147
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.592595941138985e-05,
+      "loss": 0.6639,
+      "step": 13148
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5925385210344933e-05,
+      "loss": 0.5919,
+      "step": 13149
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.59248109791914e-05,
+      "loss": 0.553,
+      "step": 13150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5924236717932167e-05,
+      "loss": 0.5656,
+      "step": 13151
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5923662426570154e-05,
+      "loss": 0.6116,
+      "step": 13152
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5923088105108275e-05,
+      "loss": 0.5695,
+      "step": 13153
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5922513753549452e-05,
+      "loss": 0.6401,
+      "step": 13154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5921939371896603e-05,
+      "loss": 0.6242,
+      "step": 13155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5921364960152647e-05,
+      "loss": 0.6394,
+      "step": 13156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.59207905183205e-05,
+      "loss": 0.6034,
+      "step": 13157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5920216046403078e-05,
+      "loss": 0.5702,
+      "step": 13158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.591964154440331e-05,
+      "loss": 0.6539,
+      "step": 13159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5919067012324107e-05,
+      "loss": 0.5836,
+      "step": 13160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5918492450168394e-05,
+      "loss": 0.6354,
+      "step": 13161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5917917857939085e-05,
+      "loss": 0.6045,
+      "step": 13162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5917343235639103e-05,
+      "loss": 0.6134,
+      "step": 13163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5916768583271366e-05,
+      "loss": 0.553,
+      "step": 13164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5916193900838797e-05,
+      "loss": 0.6537,
+      "step": 13165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.591561918834431e-05,
+      "loss": 0.5975,
+      "step": 13166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5915044445790834e-05,
+      "loss": 0.5971,
+      "step": 13167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5914469673181285e-05,
+      "loss": 0.5933,
+      "step": 13168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.591389487051858e-05,
+      "loss": 0.5745,
+      "step": 13169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5913320037805646e-05,
+      "loss": 0.6174,
+      "step": 13170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.59127451750454e-05,
+      "loss": 0.5874,
+      "step": 13171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5912170282240764e-05,
+      "loss": 0.6019,
+      "step": 13172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.591159535939466e-05,
+      "loss": 0.5843,
+      "step": 13173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5911020406510006e-05,
+      "loss": 0.6493,
+      "step": 13174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5910445423589728e-05,
+      "loss": 0.6687,
+      "step": 13175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5909870410636748e-05,
+      "loss": 0.701,
+      "step": 13176
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5909295367653983e-05,
+      "loss": 0.6139,
+      "step": 13177
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5908720294644358e-05,
+      "loss": 0.5773,
+      "step": 13178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5908145191610797e-05,
+      "loss": 0.5393,
+      "step": 13179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.590757005855622e-05,
+      "loss": 0.5804,
+      "step": 13180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5906994895483546e-05,
+      "loss": 0.6448,
+      "step": 13181
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5906419702395705e-05,
+      "loss": 0.6305,
+      "step": 13182
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.590584447929561e-05,
+      "loss": 0.6169,
+      "step": 13183
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5905269226186196e-05,
+      "loss": 0.6141,
+      "step": 13184
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5904693943070374e-05,
+      "loss": 0.5847,
+      "step": 13185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5904118629951076e-05,
+      "loss": 0.5911,
+      "step": 13186
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5903543286831223e-05,
+      "loss": 0.6759,
+      "step": 13187
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5902967913713735e-05,
+      "loss": 0.6639,
+      "step": 13188
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5902392510601537e-05,
+      "loss": 0.6305,
+      "step": 13189
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5901817077497556e-05,
+      "loss": 0.5772,
+      "step": 13190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5901241614404718e-05,
+      "loss": 0.5795,
+      "step": 13191
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.590066612132594e-05,
+      "loss": 0.5845,
+      "step": 13192
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5900090598264148e-05,
+      "loss": 0.6642,
+      "step": 13193
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.589951504522227e-05,
+      "loss": 0.5961,
+      "step": 13194
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5898939462203224e-05,
+      "loss": 0.5804,
+      "step": 13195
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5898363849209943e-05,
+      "loss": 0.6058,
+      "step": 13196
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5897788206245346e-05,
+      "loss": 0.6296,
+      "step": 13197
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.589721253331236e-05,
+      "loss": 0.5732,
+      "step": 13198
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.589663683041391e-05,
+      "loss": 0.6163,
+      "step": 13199
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5896061097552925e-05,
+      "loss": 0.6295,
+      "step": 13200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5895485334732323e-05,
+      "loss": 0.5384,
+      "step": 13201
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.589490954195503e-05,
+      "loss": 0.6471,
+      "step": 13202
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5894333719223985e-05,
+      "loss": 0.5907,
+      "step": 13203
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5893757866542097e-05,
+      "loss": 0.6239,
+      "step": 13204
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5893181983912303e-05,
+      "loss": 0.6174,
+      "step": 13205
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5892606071337522e-05,
+      "loss": 0.5923,
+      "step": 13206
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5892030128820682e-05,
+      "loss": 0.5971,
+      "step": 13207
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5891454156364718e-05,
+      "loss": 0.6067,
+      "step": 13208
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5890878153972545e-05,
+      "loss": 0.5525,
+      "step": 13209
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5890302121647096e-05,
+      "loss": 0.6168,
+      "step": 13210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5889726059391303e-05,
+      "loss": 0.6203,
+      "step": 13211
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.588914996720808e-05,
+      "loss": 0.6123,
+      "step": 13212
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5888573845100362e-05,
+      "loss": 0.624,
+      "step": 13213
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5887997693071075e-05,
+      "loss": 0.605,
+      "step": 13214
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5887421511123147e-05,
+      "loss": 0.5931,
+      "step": 13215
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.588684529925951e-05,
+      "loss": 0.6181,
+      "step": 13216
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5886269057483087e-05,
+      "loss": 0.541,
+      "step": 13217
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5885692785796806e-05,
+      "loss": 0.6257,
+      "step": 13218
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5885116484203594e-05,
+      "loss": 0.6527,
+      "step": 13219
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5884540152706382e-05,
+      "loss": 0.6529,
+      "step": 13220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.58839637913081e-05,
+      "loss": 0.6001,
+      "step": 13221
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5883387400011677e-05,
+      "loss": 0.594,
+      "step": 13222
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.588281097882004e-05,
+      "loss": 0.6006,
+      "step": 13223
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5882234527736114e-05,
+      "loss": 0.5813,
+      "step": 13224
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.588165804676283e-05,
+      "loss": 0.6389,
+      "step": 13225
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.588108153590312e-05,
+      "loss": 0.6395,
+      "step": 13226
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5880504995159917e-05,
+      "loss": 0.5922,
+      "step": 13227
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587992842453614e-05,
+      "loss": 0.6273,
+      "step": 13228
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5879351824034727e-05,
+      "loss": 0.5844,
+      "step": 13229
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587877519365861e-05,
+      "loss": 0.5867,
+      "step": 13230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5878198533410705e-05,
+      "loss": 0.6419,
+      "step": 13231
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587762184329396e-05,
+      "loss": 0.5342,
+      "step": 13232
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5877045123311294e-05,
+      "loss": 0.6394,
+      "step": 13233
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587646837346564e-05,
+      "loss": 0.5948,
+      "step": 13234
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5875891593759932e-05,
+      "loss": 0.574,
+      "step": 13235
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5875314784197094e-05,
+      "loss": 0.6014,
+      "step": 13236
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587473794478006e-05,
+      "loss": 0.5533,
+      "step": 13237
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5874161075511765e-05,
+      "loss": 0.5886,
+      "step": 13238
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587358417639514e-05,
+      "loss": 0.5457,
+      "step": 13239
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.587300724743311e-05,
+      "loss": 0.6101,
+      "step": 13240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5872430288628614e-05,
+      "loss": 0.6031,
+      "step": 13241
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5871853299984573e-05,
+      "loss": 0.6448,
+      "step": 13242
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5871276281503932e-05,
+      "loss": 0.5862,
+      "step": 13243
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5870699233189615e-05,
+      "loss": 0.5351,
+      "step": 13244
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5870122155044556e-05,
+      "loss": 0.59,
+      "step": 13245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5869545047071685e-05,
+      "loss": 0.593,
+      "step": 13246
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5868967909273938e-05,
+      "loss": 0.6744,
+      "step": 13247
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5868390741654244e-05,
+      "loss": 0.6826,
+      "step": 13248
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586781354421554e-05,
+      "loss": 0.6535,
+      "step": 13249
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586723631696076e-05,
+      "loss": 0.6166,
+      "step": 13250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5866659059892832e-05,
+      "loss": 0.6497,
+      "step": 13251
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5866081773014687e-05,
+      "loss": 0.6092,
+      "step": 13252
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5865504456329266e-05,
+      "loss": 0.6112,
+      "step": 13253
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.58649271098395e-05,
+      "loss": 0.6406,
+      "step": 13254
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586434973354832e-05,
+      "loss": 0.597,
+      "step": 13255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5863772327458664e-05,
+      "loss": 0.5615,
+      "step": 13256
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586319489157346e-05,
+      "loss": 0.6389,
+      "step": 13257
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586261742589565e-05,
+      "loss": 0.5846,
+      "step": 13258
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586203993042816e-05,
+      "loss": 0.5951,
+      "step": 13259
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.586146240517393e-05,
+      "loss": 0.5811,
+      "step": 13260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5860884850135892e-05,
+      "loss": 0.595,
+      "step": 13261
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5860307265316984e-05,
+      "loss": 0.5603,
+      "step": 13262
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5859729650720135e-05,
+      "loss": 0.6649,
+      "step": 13263
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5859152006348282e-05,
+      "loss": 0.631,
+      "step": 13264
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5858574332204367e-05,
+      "loss": 0.6246,
+      "step": 13265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585799662829132e-05,
+      "loss": 0.6585,
+      "step": 13266
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5857418894612073e-05,
+      "loss": 0.6699,
+      "step": 13267
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5856841131169566e-05,
+      "loss": 0.5915,
+      "step": 13268
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5856263337966734e-05,
+      "loss": 0.6127,
+      "step": 13269
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585568551500652e-05,
+      "loss": 0.6062,
+      "step": 13270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5855107662291845e-05,
+      "loss": 0.5759,
+      "step": 13271
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585452977982565e-05,
+      "loss": 0.6249,
+      "step": 13272
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585395186761088e-05,
+      "loss": 0.5903,
+      "step": 13273
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585337392565047e-05,
+      "loss": 0.5929,
+      "step": 13274
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5852795953947345e-05,
+      "loss": 0.6291,
+      "step": 13275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5852217952504453e-05,
+      "loss": 0.6173,
+      "step": 13276
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585163992132473e-05,
+      "loss": 0.588,
+      "step": 13277
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.585106186041111e-05,
+      "loss": 0.6078,
+      "step": 13278
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5850483769766528e-05,
+      "loss": 0.6329,
+      "step": 13279
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5849905649393926e-05,
+      "loss": 0.6127,
+      "step": 13280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.584932749929624e-05,
+      "loss": 0.5532,
+      "step": 13281
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5848749319476406e-05,
+      "loss": 0.5962,
+      "step": 13282
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5848171109937368e-05,
+      "loss": 0.5823,
+      "step": 13283
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.584759287068206e-05,
+      "loss": 0.6221,
+      "step": 13284
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5847014601713416e-05,
+      "loss": 0.6155,
+      "step": 13285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.584643630303438e-05,
+      "loss": 0.6313,
+      "step": 13286
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5845857974647893e-05,
+      "loss": 0.6662,
+      "step": 13287
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5845279616556883e-05,
+      "loss": 0.5479,
+      "step": 13288
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5844701228764297e-05,
+      "loss": 0.608,
+      "step": 13289
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5844122811273076e-05,
+      "loss": 0.6394,
+      "step": 13290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5843544364086154e-05,
+      "loss": 0.5991,
+      "step": 13291
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.584296588720647e-05,
+      "loss": 0.6111,
+      "step": 13292
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5842387380636966e-05,
+      "loss": 0.598,
+      "step": 13293
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.584180884438058e-05,
+      "loss": 0.5531,
+      "step": 13294
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5841230278440254e-05,
+      "loss": 0.6189,
+      "step": 13295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5840651682818926e-05,
+      "loss": 0.6055,
+      "step": 13296
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5840073057519535e-05,
+      "loss": 0.6506,
+      "step": 13297
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.5839494402545022e-05,
+      "loss": 0.5786,
+      "step": 13298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5838915717898332e-05,
+      "loss": 0.5902,
+      "step": 13299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.58383370035824e-05,
+      "loss": 0.6197,
+      "step": 13300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5837758259600163e-05,
+      "loss": 0.6494,
+      "step": 13301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583717948595457e-05,
+      "loss": 0.6473,
+      "step": 13302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583660068264856e-05,
+      "loss": 0.6388,
+      "step": 13303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5836021849685068e-05,
+      "loss": 0.579,
+      "step": 13304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5835442987067045e-05,
+      "loss": 0.6409,
+      "step": 13305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5834864094797425e-05,
+      "loss": 0.6688,
+      "step": 13306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583428517287915e-05,
+      "loss": 0.609,
+      "step": 13307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5833706221315168e-05,
+      "loss": 0.6028,
+      "step": 13308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583312724010841e-05,
+      "loss": 0.6442,
+      "step": 13309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583254822926183e-05,
+      "loss": 0.5677,
+      "step": 13310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5831969188778364e-05,
+      "loss": 0.6193,
+      "step": 13311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.583139011866095e-05,
+      "loss": 0.6205,
+      "step": 13312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5830811018912536e-05,
+      "loss": 0.6164,
+      "step": 13313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5830231889536066e-05,
+      "loss": 0.6099,
+      "step": 13314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5829652730534476e-05,
+      "loss": 0.6697,
+      "step": 13315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5829073541910717e-05,
+      "loss": 0.5907,
+      "step": 13316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5828494323667728e-05,
+      "loss": 0.6176,
+      "step": 13317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5827915075808452e-05,
+      "loss": 0.5665,
+      "step": 13318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5827335798335833e-05,
+      "loss": 0.6184,
+      "step": 13319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5826756491252816e-05,
+      "loss": 0.6083,
+      "step": 13320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.582617715456234e-05,
+      "loss": 0.5678,
+      "step": 13321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.582559778826735e-05,
+      "loss": 0.5994,
+      "step": 13322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5825018392370795e-05,
+      "loss": 0.565,
+      "step": 13323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.582443896687562e-05,
+      "loss": 0.6152,
+      "step": 13324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5823859511784754e-05,
+      "loss": 0.5849,
+      "step": 13325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5823280027101158e-05,
+      "loss": 0.555,
+      "step": 13326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.582270051282777e-05,
+      "loss": 0.6732,
+      "step": 13327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5822120968967535e-05,
+      "loss": 0.5655,
+      "step": 13328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.58215413955234e-05,
+      "loss": 0.5428,
+      "step": 13329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5820961792498306e-05,
+      "loss": 0.6552,
+      "step": 13330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.58203821598952e-05,
+      "loss": 0.5838,
+      "step": 13331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581980249771703e-05,
+      "loss": 0.5869,
+      "step": 13332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581922280596674e-05,
+      "loss": 0.6247,
+      "step": 13333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581864308464727e-05,
+      "loss": 0.6615,
+      "step": 13334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581806333376158e-05,
+      "loss": 0.6059,
+      "step": 13335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5817483553312596e-05,
+      "loss": 0.5804,
+      "step": 13336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581690374330328e-05,
+      "loss": 0.5573,
+      "step": 13337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581632390373657e-05,
+      "loss": 0.6063,
+      "step": 13338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5815744034615413e-05,
+      "loss": 0.6362,
+      "step": 13339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581516413594276e-05,
+      "loss": 0.697,
+      "step": 13340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5814584207721558e-05,
+      "loss": 0.6607,
+      "step": 13341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5814004249954746e-05,
+      "loss": 0.6556,
+      "step": 13342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581342426264528e-05,
+      "loss": 0.593,
+      "step": 13343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5812844245796096e-05,
+      "loss": 0.6095,
+      "step": 13344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5812264199410156e-05,
+      "loss": 0.6276,
+      "step": 13345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5811684123490393e-05,
+      "loss": 0.5862,
+      "step": 13346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.581110401803976e-05,
+      "loss": 0.6157,
+      "step": 13347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5810523883061213e-05,
+      "loss": 0.5728,
+      "step": 13348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580994371855769e-05,
+      "loss": 0.6155,
+      "step": 13349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580936352453214e-05,
+      "loss": 0.6259,
+      "step": 13350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580878330098751e-05,
+      "loss": 0.6347,
+      "step": 13351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5808203047926754e-05,
+      "loss": 0.646,
+      "step": 13352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5807622765352816e-05,
+      "loss": 0.6132,
+      "step": 13353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5807042453268646e-05,
+      "loss": 0.6546,
+      "step": 13354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5806462111677196e-05,
+      "loss": 0.5374,
+      "step": 13355
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5805881740581406e-05,
+      "loss": 0.5714,
+      "step": 13356
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5805301339984234e-05,
+      "loss": 0.6103,
+      "step": 13357
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5804720909888626e-05,
+      "loss": 0.6201,
+      "step": 13358
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580414045029753e-05,
+      "loss": 0.5914,
+      "step": 13359
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5803559961213897e-05,
+      "loss": 0.6105,
+      "step": 13360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5802979442640672e-05,
+      "loss": 0.6016,
+      "step": 13361
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5802398894580817e-05,
+      "loss": 0.5718,
+      "step": 13362
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580181831703727e-05,
+      "loss": 0.5971,
+      "step": 13363
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5801237710012983e-05,
+      "loss": 0.621,
+      "step": 13364
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.580065707351091e-05,
+      "loss": 0.6118,
+      "step": 13365
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5800076407534e-05,
+      "loss": 0.5428,
+      "step": 13366
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5799495712085203e-05,
+      "loss": 0.6541,
+      "step": 13367
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5798914987167468e-05,
+      "loss": 0.5768,
+      "step": 13368
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579833423278375e-05,
+      "loss": 0.6941,
+      "step": 13369
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5797753448936997e-05,
+      "loss": 0.6023,
+      "step": 13370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579717263563016e-05,
+      "loss": 0.5508,
+      "step": 13371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579659179286619e-05,
+      "loss": 0.6272,
+      "step": 13372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5796010920648042e-05,
+      "loss": 0.571,
+      "step": 13373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5795430018978664e-05,
+      "loss": 0.6203,
+      "step": 13374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579484908786101e-05,
+      "loss": 0.5919,
+      "step": 13375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579426812729803e-05,
+      "loss": 0.5892,
+      "step": 13376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5793687137292676e-05,
+      "loss": 0.6105,
+      "step": 13377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.57931061178479e-05,
+      "loss": 0.5878,
+      "step": 13378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.579252506896666e-05,
+      "loss": 0.5741,
+      "step": 13379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5791943990651898e-05,
+      "loss": 0.592,
+      "step": 13380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5791362882906573e-05,
+      "loss": 0.565,
+      "step": 13381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5790781745733637e-05,
+      "loss": 0.5806,
+      "step": 13382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5790200579136047e-05,
+      "loss": 0.6589,
+      "step": 13383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.578961938311675e-05,
+      "loss": 0.638,
+      "step": 13384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.57890381576787e-05,
+      "loss": 0.591,
+      "step": 13385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5788456902824852e-05,
+      "loss": 0.5937,
+      "step": 13386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5787875618558163e-05,
+      "loss": 0.6047,
+      "step": 13387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.578729430488158e-05,
+      "loss": 0.6157,
+      "step": 13388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.578671296179806e-05,
+      "loss": 0.599,
+      "step": 13389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5786131589310555e-05,
+      "loss": 0.6407,
+      "step": 13390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.578555018742202e-05,
+      "loss": 0.5769,
+      "step": 13391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5784968756135417e-05,
+      "loss": 0.6014,
+      "step": 13392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5784387295453686e-05,
+      "loss": 0.5706,
+      "step": 13393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5783805805379794e-05,
+      "loss": 0.6492,
+      "step": 13394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.578322428591669e-05,
+      "loss": 0.6419,
+      "step": 13395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5782642737067327e-05,
+      "loss": 0.6375,
+      "step": 13396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5782061158834665e-05,
+      "loss": 0.5905,
+      "step": 13397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5781479551221653e-05,
+      "loss": 0.5963,
+      "step": 13398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5780897914231258e-05,
+      "loss": 0.5612,
+      "step": 13399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5780316247866424e-05,
+      "loss": 0.6074,
+      "step": 13400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5779734552130106e-05,
+      "loss": 0.5842,
+      "step": 13401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.577915282702527e-05,
+      "loss": 0.6023,
+      "step": 13402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5778571072554863e-05,
+      "loss": 0.578,
+      "step": 13403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5777989288721843e-05,
+      "loss": 0.596,
+      "step": 13404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.577740747552917e-05,
+      "loss": 0.6467,
+      "step": 13405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5776825632979796e-05,
+      "loss": 0.6151,
+      "step": 13406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5776243761076678e-05,
+      "loss": 0.6037,
+      "step": 13407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5775661859822778e-05,
+      "loss": 0.6349,
+      "step": 13408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5775079929221045e-05,
+      "loss": 0.6488,
+      "step": 13409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.577449796927444e-05,
+      "loss": 0.6392,
+      "step": 13410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.577391597998592e-05,
+      "loss": 0.5399,
+      "step": 13411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.577333396135844e-05,
+      "loss": 0.6341,
+      "step": 13412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5772751913394965e-05,
+      "loss": 0.6702,
+      "step": 13413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5772169836098443e-05,
+      "loss": 0.6203,
+      "step": 13414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5771587729471836e-05,
+      "loss": 0.6226,
+      "step": 13415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.57710055935181e-05,
+      "loss": 0.6459,
+      "step": 13416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5770423428240197e-05,
+      "loss": 0.5838,
+      "step": 13417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5769841233641084e-05,
+      "loss": 0.6167,
+      "step": 13418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5769259009723714e-05,
+      "loss": 0.5958,
+      "step": 13419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.576867675649105e-05,
+      "loss": 0.5867,
+      "step": 13420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.576809447394605e-05,
+      "loss": 0.602,
+      "step": 13421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5767512162091674e-05,
+      "loss": 0.5762,
+      "step": 13422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5766929820930877e-05,
+      "loss": 0.6359,
+      "step": 13423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5766347450466626e-05,
+      "loss": 0.6384,
+      "step": 13424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5765765050701873e-05,
+      "loss": 0.5671,
+      "step": 13425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5765182621639577e-05,
+      "loss": 0.5947,
+      "step": 13426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.57646001632827e-05,
+      "loss": 0.5739,
+      "step": 13427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.57640176756342e-05,
+      "loss": 0.6211,
+      "step": 13428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.576343515869704e-05,
+      "loss": 0.5773,
+      "step": 13429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.576285261247418e-05,
+      "loss": 0.6181,
+      "step": 13430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5762270036968575e-05,
+      "loss": 0.6099,
+      "step": 13431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.576168743218319e-05,
+      "loss": 0.578,
+      "step": 13432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5761104798120984e-05,
+      "loss": 0.5596,
+      "step": 13433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5760522134784917e-05,
+      "loss": 0.5603,
+      "step": 13434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.575993944217795e-05,
+      "loss": 0.6194,
+      "step": 13435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5759356720303042e-05,
+      "loss": 0.6096,
+      "step": 13436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5758773969163157e-05,
+      "loss": 0.6727,
+      "step": 13437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5758191188761255e-05,
+      "loss": 0.5597,
+      "step": 13438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5757608379100297e-05,
+      "loss": 0.6311,
+      "step": 13439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5757025540183245e-05,
+      "loss": 0.5933,
+      "step": 13440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5756442672013062e-05,
+      "loss": 0.6437,
+      "step": 13441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5755859774592705e-05,
+      "loss": 0.5606,
+      "step": 13442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.575527684792514e-05,
+      "loss": 0.5785,
+      "step": 13443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.575469389201333e-05,
+      "loss": 0.575,
+      "step": 13444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5754110906860232e-05,
+      "loss": 0.6378,
+      "step": 13445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5753527892468812e-05,
+      "loss": 0.6352,
+      "step": 13446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.575294484884203e-05,
+      "loss": 0.5948,
+      "step": 13447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5752361775982853e-05,
+      "loss": 0.5659,
+      "step": 13448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.575177867389424e-05,
+      "loss": 0.6176,
+      "step": 13449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5751195542579157e-05,
+      "loss": 0.589,
+      "step": 13450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.575061238204056e-05,
+      "loss": 0.5738,
+      "step": 13451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5750029192281425e-05,
+      "loss": 0.6058,
+      "step": 13452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.57494459733047e-05,
+      "loss": 0.6138,
+      "step": 13453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5748862725113363e-05,
+      "loss": 0.5654,
+      "step": 13454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.574827944771037e-05,
+      "loss": 0.609,
+      "step": 13455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.574769614109868e-05,
+      "loss": 0.6325,
+      "step": 13456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5747112805281267e-05,
+      "loss": 0.5813,
+      "step": 13457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.574652944026109e-05,
+      "loss": 0.6126,
+      "step": 13458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5745946046041113e-05,
+      "loss": 0.5557,
+      "step": 13459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5745362622624303e-05,
+      "loss": 0.6344,
+      "step": 13460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.574477917001362e-05,
+      "loss": 0.6402,
+      "step": 13461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5744195688212037e-05,
+      "loss": 0.6711,
+      "step": 13462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5743612177222507e-05,
+      "loss": 0.6209,
+      "step": 13463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5743028637048006e-05,
+      "loss": 0.6584,
+      "step": 13464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5742445067691492e-05,
+      "loss": 0.5799,
+      "step": 13465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5741861469155936e-05,
+      "loss": 0.6109,
+      "step": 13466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5741277841444297e-05,
+      "loss": 0.6031,
+      "step": 13467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5740694184559545e-05,
+      "loss": 0.6049,
+      "step": 13468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5740110498504645e-05,
+      "loss": 0.5961,
+      "step": 13469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5739526783282563e-05,
+      "loss": 0.6162,
+      "step": 13470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573894303889626e-05,
+      "loss": 0.6195,
+      "step": 13471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573835926534871e-05,
+      "loss": 0.6418,
+      "step": 13472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573777546264288e-05,
+      "loss": 0.6295,
+      "step": 13473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573719163078173e-05,
+      "loss": 0.6366,
+      "step": 13474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5736607769768225e-05,
+      "loss": 0.6149,
+      "step": 13475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573602387960534e-05,
+      "loss": 0.5437,
+      "step": 13476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5735439960296037e-05,
+      "loss": 0.6258,
+      "step": 13477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573485601184328e-05,
+      "loss": 0.572,
+      "step": 13478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5734272034250048e-05,
+      "loss": 0.6644,
+      "step": 13479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5733688027519293e-05,
+      "loss": 0.6497,
+      "step": 13480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5733103991653996e-05,
+      "loss": 0.5891,
+      "step": 13481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5732519926657115e-05,
+      "loss": 0.5984,
+      "step": 13482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.573193583253162e-05,
+      "loss": 0.5373,
+      "step": 13483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5731351709280483e-05,
+      "loss": 0.6,
+      "step": 13484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5730767556906668e-05,
+      "loss": 0.6198,
+      "step": 13485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5730183375413145e-05,
+      "loss": 0.5352,
+      "step": 13486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5729599164802884e-05,
+      "loss": 0.6091,
+      "step": 13487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572901492507885e-05,
+      "loss": 0.5829,
+      "step": 13488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5728430656244013e-05,
+      "loss": 0.6301,
+      "step": 13489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5727846358301345e-05,
+      "loss": 0.6008,
+      "step": 13490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572726203125381e-05,
+      "loss": 0.6037,
+      "step": 13491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572667767510438e-05,
+      "loss": 0.6566,
+      "step": 13492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572609328985602e-05,
+      "loss": 0.556,
+      "step": 13493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5725508875511705e-05,
+      "loss": 0.6134,
+      "step": 13494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5724924432074408e-05,
+      "loss": 0.6,
+      "step": 13495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5724339959547086e-05,
+      "loss": 0.6292,
+      "step": 13496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5723755457932717e-05,
+      "loss": 0.6508,
+      "step": 13497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5723170927234276e-05,
+      "loss": 0.6227,
+      "step": 13498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5722586367454723e-05,
+      "loss": 0.5916,
+      "step": 13499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572200177859703e-05,
+      "loss": 0.5908,
+      "step": 13500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5721417160664178e-05,
+      "loss": 0.5515,
+      "step": 13501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5720832513659124e-05,
+      "loss": 0.7215,
+      "step": 13502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5720247837584845e-05,
+      "loss": 0.6021,
+      "step": 13503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571966313244431e-05,
+      "loss": 0.6276,
+      "step": 13504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5719078398240495e-05,
+      "loss": 0.6308,
+      "step": 13505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5718493634976368e-05,
+      "loss": 0.6115,
+      "step": 13506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5717908842654897e-05,
+      "loss": 0.5898,
+      "step": 13507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5717324021279062e-05,
+      "loss": 0.6094,
+      "step": 13508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5716739170851826e-05,
+      "loss": 0.6961,
+      "step": 13509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5716154291376164e-05,
+      "loss": 0.6128,
+      "step": 13510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5715569382855048e-05,
+      "loss": 0.5568,
+      "step": 13511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571498444529145e-05,
+      "loss": 0.5842,
+      "step": 13512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571439947868834e-05,
+      "loss": 0.5849,
+      "step": 13513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5713814483048694e-05,
+      "loss": 0.6694,
+      "step": 13514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5713229458375484e-05,
+      "loss": 0.6289,
+      "step": 13515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5712644404671684e-05,
+      "loss": 0.643,
+      "step": 13516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.571205932194026e-05,
+      "loss": 0.6433,
+      "step": 13517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5711474210184195e-05,
+      "loss": 0.5869,
+      "step": 13518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5710889069406455e-05,
+      "loss": 0.6329,
+      "step": 13519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5710303899610013e-05,
+      "loss": 0.5444,
+      "step": 13520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5709718700797844e-05,
+      "loss": 0.6492,
+      "step": 13521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5709133472972927e-05,
+      "loss": 0.6131,
+      "step": 13522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.570854821613823e-05,
+      "loss": 0.5886,
+      "step": 13523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5707962930296722e-05,
+      "loss": 0.6463,
+      "step": 13524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.570737761545139e-05,
+      "loss": 0.6191,
+      "step": 13525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5706792271605195e-05,
+      "loss": 0.6296,
+      "step": 13526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5706206898761123e-05,
+      "loss": 0.6227,
+      "step": 13527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5705621496922136e-05,
+      "loss": 0.5282,
+      "step": 13528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5705036066091218e-05,
+      "loss": 0.6725,
+      "step": 13529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5704450606271342e-05,
+      "loss": 0.5691,
+      "step": 13530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.570386511746548e-05,
+      "loss": 0.6039,
+      "step": 13531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.570327959967661e-05,
+      "loss": 0.5705,
+      "step": 13532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5702694052907706e-05,
+      "loss": 0.5685,
+      "step": 13533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5702108477161745e-05,
+      "loss": 0.6092,
+      "step": 13534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.57015228724417e-05,
+      "loss": 0.5917,
+      "step": 13535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5700937238750546e-05,
+      "loss": 0.6101,
+      "step": 13536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.570035157609126e-05,
+      "loss": 0.6052,
+      "step": 13537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.569976588446682e-05,
+      "loss": 0.5892,
+      "step": 13538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.56991801638802e-05,
+      "loss": 0.644,
+      "step": 13539
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5698594414334375e-05,
+      "loss": 0.5802,
+      "step": 13540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5698008635832325e-05,
+      "loss": 0.566,
+      "step": 13541
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5697422828377023e-05,
+      "loss": 0.6015,
+      "step": 13542
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5696836991971445e-05,
+      "loss": 0.5825,
+      "step": 13543
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5696251126618575e-05,
+      "loss": 0.5663,
+      "step": 13544
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.569566523232138e-05,
+      "loss": 0.6466,
+      "step": 13545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5695079309082845e-05,
+      "loss": 0.5789,
+      "step": 13546
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5694493356905943e-05,
+      "loss": 0.5906,
+      "step": 13547
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.569390737579365e-05,
+      "loss": 0.635,
+      "step": 13548
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5693321365748954e-05,
+      "loss": 0.5634,
+      "step": 13549
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5692735326774817e-05,
+      "loss": 0.6219,
+      "step": 13550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5692149258874227e-05,
+      "loss": 0.629,
+      "step": 13551
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5691563162050162e-05,
+      "loss": 0.6088,
+      "step": 13552
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5690977036305594e-05,
+      "loss": 0.596,
+      "step": 13553
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5690390881643507e-05,
+      "loss": 0.5848,
+      "step": 13554
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5689804698066877e-05,
+      "loss": 0.6543,
+      "step": 13555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568921848557868e-05,
+      "loss": 0.618,
+      "step": 13556
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5688632244181904e-05,
+      "loss": 0.6124,
+      "step": 13557
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5688045973879515e-05,
+      "loss": 0.6079,
+      "step": 13558
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5687459674674502e-05,
+      "loss": 0.6101,
+      "step": 13559
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5686873346569842e-05,
+      "loss": 0.5728,
+      "step": 13560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5686286989568512e-05,
+      "loss": 0.557,
+      "step": 13561
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568570060367349e-05,
+      "loss": 0.5798,
+      "step": 13562
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5685114188887763e-05,
+      "loss": 0.6109,
+      "step": 13563
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.56845277452143e-05,
+      "loss": 0.642,
+      "step": 13564
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568394127265609e-05,
+      "loss": 0.5896,
+      "step": 13565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568335477121611e-05,
+      "loss": 0.5927,
+      "step": 13566
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568276824089734e-05,
+      "loss": 0.653,
+      "step": 13567
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568218168170276e-05,
+      "loss": 0.6152,
+      "step": 13568
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5681595093635347e-05,
+      "loss": 0.6084,
+      "step": 13569
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.568100847669809e-05,
+      "loss": 0.6849,
+      "step": 13570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5680421830893965e-05,
+      "loss": 0.5892,
+      "step": 13571
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5679835156225955e-05,
+      "loss": 0.5682,
+      "step": 13572
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5679248452697035e-05,
+      "loss": 0.5472,
+      "step": 13573
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.567866172031019e-05,
+      "loss": 0.6298,
+      "step": 13574
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5678074959068403e-05,
+      "loss": 0.6352,
+      "step": 13575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5677488168974658e-05,
+      "loss": 0.6277,
+      "step": 13576
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5676901350031925e-05,
+      "loss": 0.6509,
+      "step": 13577
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5676314502243198e-05,
+      "loss": 0.6254,
+      "step": 13578
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5675727625611457e-05,
+      "loss": 0.6782,
+      "step": 13579
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5675140720139678e-05,
+      "loss": 0.5682,
+      "step": 13580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.567455378583085e-05,
+      "loss": 0.6833,
+      "step": 13581
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5673966822687953e-05,
+      "loss": 0.6201,
+      "step": 13582
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5673379830713964e-05,
+      "loss": 0.5857,
+      "step": 13583
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5672792809911876e-05,
+      "loss": 0.5923,
+      "step": 13584
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5672205760284663e-05,
+      "loss": 0.6077,
+      "step": 13585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.567161868183531e-05,
+      "loss": 0.588,
+      "step": 13586
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5671031574566803e-05,
+      "loss": 0.5722,
+      "step": 13587
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5670444438482125e-05,
+      "loss": 0.6131,
+      "step": 13588
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5669857273584255e-05,
+      "loss": 0.6058,
+      "step": 13589
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5669270079876183e-05,
+      "loss": 0.6212,
+      "step": 13590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.566868285736089e-05,
+      "loss": 0.5854,
+      "step": 13591
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5668095606041357e-05,
+      "loss": 0.5987,
+      "step": 13592
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5667508325920572e-05,
+      "loss": 0.6613,
+      "step": 13593
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5666921017001515e-05,
+      "loss": 0.6444,
+      "step": 13594
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5666333679287172e-05,
+      "loss": 0.649,
+      "step": 13595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5665746312780533e-05,
+      "loss": 0.6014,
+      "step": 13596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5665158917484573e-05,
+      "loss": 0.586,
+      "step": 13597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5664571493402285e-05,
+      "loss": 0.601,
+      "step": 13598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.566398404053665e-05,
+      "loss": 0.6279,
+      "step": 13599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.566339655889065e-05,
+      "loss": 0.5925,
+      "step": 13600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5662809048467272e-05,
+      "loss": 0.6125,
+      "step": 13601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5662221509269508e-05,
+      "loss": 0.6081,
+      "step": 13602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5661633941300338e-05,
+      "loss": 0.6005,
+      "step": 13603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5661046344562744e-05,
+      "loss": 0.5807,
+      "step": 13604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.566045871905972e-05,
+      "loss": 0.5529,
+      "step": 13605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5659871064794246e-05,
+      "loss": 0.5766,
+      "step": 13606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5659283381769307e-05,
+      "loss": 0.6586,
+      "step": 13607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5658695669987895e-05,
+      "loss": 0.623,
+      "step": 13608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.565810792945299e-05,
+      "loss": 0.6288,
+      "step": 13609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5657520160167586e-05,
+      "loss": 0.6145,
+      "step": 13610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5656932362134662e-05,
+      "loss": 0.5956,
+      "step": 13611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5656344535357208e-05,
+      "loss": 0.6642,
+      "step": 13612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.565575667983821e-05,
+      "loss": 0.6935,
+      "step": 13613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5655168795580657e-05,
+      "loss": 0.5797,
+      "step": 13614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5654580882587534e-05,
+      "loss": 0.6562,
+      "step": 13615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5653992940861834e-05,
+      "loss": 0.6048,
+      "step": 13616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5653404970406538e-05,
+      "loss": 0.5872,
+      "step": 13617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.565281697122463e-05,
+      "loss": 0.6044,
+      "step": 13618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.565222894331911e-05,
+      "loss": 0.6042,
+      "step": 13619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5651640886692958e-05,
+      "loss": 0.5991,
+      "step": 13620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5651052801349162e-05,
+      "loss": 0.6241,
+      "step": 13621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5650464687290713e-05,
+      "loss": 0.6823,
+      "step": 13622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5649876544520597e-05,
+      "loss": 0.566,
+      "step": 13623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5649288373041804e-05,
+      "loss": 0.5704,
+      "step": 13624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5648700172857325e-05,
+      "loss": 0.6289,
+      "step": 13625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5648111943970142e-05,
+      "loss": 0.6039,
+      "step": 13626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5647523686383254e-05,
+      "loss": 0.5896,
+      "step": 13627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5646935400099638e-05,
+      "loss": 0.5928,
+      "step": 13628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.564634708512229e-05,
+      "loss": 0.5985,
+      "step": 13629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56457587414542e-05,
+      "loss": 0.5866,
+      "step": 13630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5645170369098356e-05,
+      "loss": 0.5649,
+      "step": 13631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.564458196805775e-05,
+      "loss": 0.6262,
+      "step": 13632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.564399353833537e-05,
+      "loss": 0.7568,
+      "step": 13633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5643405079934205e-05,
+      "loss": 0.5702,
+      "step": 13634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5642816592857244e-05,
+      "loss": 0.6326,
+      "step": 13635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5642228077107483e-05,
+      "loss": 0.6435,
+      "step": 13636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5641639532687904e-05,
+      "loss": 0.6317,
+      "step": 13637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5641050959601507e-05,
+      "loss": 0.5865,
+      "step": 13638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5640462357851278e-05,
+      "loss": 0.6255,
+      "step": 13639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5639873727440205e-05,
+      "loss": 0.5784,
+      "step": 13640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563928506837128e-05,
+      "loss": 0.576,
+      "step": 13641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56386963806475e-05,
+      "loss": 0.6169,
+      "step": 13642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563810766427185e-05,
+      "loss": 0.5988,
+      "step": 13643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5637518919247327e-05,
+      "loss": 0.5913,
+      "step": 13644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5636930145576918e-05,
+      "loss": 0.5612,
+      "step": 13645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5636341343263616e-05,
+      "loss": 0.5646,
+      "step": 13646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563575251231041e-05,
+      "loss": 0.5414,
+      "step": 13647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5635163652720297e-05,
+      "loss": 0.6041,
+      "step": 13648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563457476449627e-05,
+      "loss": 0.6339,
+      "step": 13649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5633985847641315e-05,
+      "loss": 0.6138,
+      "step": 13650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5633396902158428e-05,
+      "loss": 0.6109,
+      "step": 13651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5632807928050605e-05,
+      "loss": 0.6324,
+      "step": 13652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563221892532083e-05,
+      "loss": 0.598,
+      "step": 13653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5631629893972105e-05,
+      "loss": 0.5872,
+      "step": 13654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.563104083400742e-05,
+      "loss": 0.6219,
+      "step": 13655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5630451745429764e-05,
+      "loss": 0.6002,
+      "step": 13656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5629862628242135e-05,
+      "loss": 0.6336,
+      "step": 13657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5629273482447523e-05,
+      "loss": 0.6452,
+      "step": 13658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5628684308048927e-05,
+      "loss": 0.6196,
+      "step": 13659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5628095105049338e-05,
+      "loss": 0.6276,
+      "step": 13660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.562750587345175e-05,
+      "loss": 0.5384,
+      "step": 13661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5626916613259153e-05,
+      "loss": 0.6277,
+      "step": 13662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5626327324474548e-05,
+      "loss": 0.5957,
+      "step": 13663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5625738007100924e-05,
+      "loss": 0.628,
+      "step": 13664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.562514866114128e-05,
+      "loss": 0.5338,
+      "step": 13665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5624559286598606e-05,
+      "loss": 0.5281,
+      "step": 13666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56239698834759e-05,
+      "loss": 0.6267,
+      "step": 13667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5623380451776155e-05,
+      "loss": 0.6413,
+      "step": 13668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5622790991502367e-05,
+      "loss": 0.6248,
+      "step": 13669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5622201502657534e-05,
+      "loss": 0.6116,
+      "step": 13670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5621611985244648e-05,
+      "loss": 0.5784,
+      "step": 13671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56210224392667e-05,
+      "loss": 0.5857,
+      "step": 13672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5620432864726694e-05,
+      "loss": 0.6016,
+      "step": 13673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5619843261627624e-05,
+      "loss": 0.5975,
+      "step": 13674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5619253629972484e-05,
+      "loss": 0.5673,
+      "step": 13675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5618663969764273e-05,
+      "loss": 0.6026,
+      "step": 13676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561807428100598e-05,
+      "loss": 0.5811,
+      "step": 13677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5617484563700606e-05,
+      "loss": 0.616,
+      "step": 13678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561689481785115e-05,
+      "loss": 0.594,
+      "step": 13679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561630504346061e-05,
+      "loss": 0.6228,
+      "step": 13680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561571524053197e-05,
+      "loss": 0.5854,
+      "step": 13681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5615125409068247e-05,
+      "loss": 0.6825,
+      "step": 13682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561453554907242e-05,
+      "loss": 0.6256,
+      "step": 13683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56139456605475e-05,
+      "loss": 0.5596,
+      "step": 13684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5613355743496473e-05,
+      "loss": 0.662,
+      "step": 13685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561276579792234e-05,
+      "loss": 0.5732,
+      "step": 13686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5612175823828105e-05,
+      "loss": 0.6081,
+      "step": 13687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5611585821216756e-05,
+      "loss": 0.5955,
+      "step": 13688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5610995790091303e-05,
+      "loss": 0.6567,
+      "step": 13689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.561040573045473e-05,
+      "loss": 0.6352,
+      "step": 13690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5609815642310046e-05,
+      "loss": 0.5679,
+      "step": 13691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5609225525660246e-05,
+      "loss": 0.6328,
+      "step": 13692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560863538050833e-05,
+      "loss": 0.6283,
+      "step": 13693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560804520685729e-05,
+      "loss": 0.6494,
+      "step": 13694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560745500471013e-05,
+      "loss": 0.601,
+      "step": 13695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560686477406985e-05,
+      "loss": 0.6037,
+      "step": 13696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5606274514939452e-05,
+      "loss": 0.6158,
+      "step": 13697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560568422732193e-05,
+      "loss": 0.5701,
+      "step": 13698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5605093911220282e-05,
+      "loss": 0.6582,
+      "step": 13699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560450356663751e-05,
+      "loss": 0.6365,
+      "step": 13700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5603913193576617e-05,
+      "loss": 0.6285,
+      "step": 13701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5603322792040597e-05,
+      "loss": 0.5733,
+      "step": 13702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560273236203246e-05,
+      "loss": 0.6457,
+      "step": 13703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.560214190355519e-05,
+      "loss": 0.5927,
+      "step": 13704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.56015514166118e-05,
+      "loss": 0.6329,
+      "step": 13705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5600960901205287e-05,
+      "loss": 0.5767,
+      "step": 13706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5600370357338652e-05,
+      "loss": 0.5443,
+      "step": 13707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559977978501489e-05,
+      "loss": 0.6044,
+      "step": 13708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5599189184237013e-05,
+      "loss": 0.6543,
+      "step": 13709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5598598555008016e-05,
+      "loss": 0.6018,
+      "step": 13710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5598007897330897e-05,
+      "loss": 0.5684,
+      "step": 13711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5597417211208662e-05,
+      "loss": 0.6693,
+      "step": 13712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5596826496644308e-05,
+      "loss": 0.6752,
+      "step": 13713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5596235753640843e-05,
+      "loss": 0.6308,
+      "step": 13714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5595644982201266e-05,
+      "loss": 0.5982,
+      "step": 13715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5595054182328575e-05,
+      "loss": 0.5312,
+      "step": 13716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559446335402578e-05,
+      "loss": 0.6493,
+      "step": 13717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559387249729587e-05,
+      "loss": 0.5899,
+      "step": 13718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559328161214186e-05,
+      "loss": 0.6054,
+      "step": 13719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559269069856675e-05,
+      "loss": 0.6931,
+      "step": 13720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559209975657354e-05,
+      "loss": 0.5851,
+      "step": 13721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559150878616523e-05,
+      "loss": 0.6231,
+      "step": 13722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559091778734483e-05,
+      "loss": 0.5927,
+      "step": 13723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.559032676011534e-05,
+      "loss": 0.5915,
+      "step": 13724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.558973570447976e-05,
+      "loss": 0.5826,
+      "step": 13725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5589144620441093e-05,
+      "loss": 0.6191,
+      "step": 13726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.558855350800235e-05,
+      "loss": 0.5748,
+      "step": 13727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5587962367166526e-05,
+      "loss": 0.5922,
+      "step": 13728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5587371197936634e-05,
+      "loss": 0.6033,
+      "step": 13729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.558678000031567e-05,
+      "loss": 0.5845,
+      "step": 13730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5586188774306638e-05,
+      "loss": 0.5522,
+      "step": 13731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.558559751991255e-05,
+      "loss": 0.7076,
+      "step": 13732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.55850062371364e-05,
+      "loss": 0.6468,
+      "step": 13733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5584414925981203e-05,
+      "loss": 0.5785,
+      "step": 13734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5583823586449954e-05,
+      "loss": 0.6635,
+      "step": 13735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5583232218545665e-05,
+      "loss": 0.6682,
+      "step": 13736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5582640822271337e-05,
+      "loss": 0.5708,
+      "step": 13737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5582049397629976e-05,
+      "loss": 0.6054,
+      "step": 13738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.5581457944624586e-05,
+      "loss": 0.6036,
+      "step": 13739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5580866463258174e-05,
+      "loss": 0.6385,
+      "step": 13740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5580274953533747e-05,
+      "loss": 0.6557,
+      "step": 13741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5579683415454305e-05,
+      "loss": 0.6492,
+      "step": 13742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557909184902286e-05,
+      "loss": 0.6142,
+      "step": 13743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5578500254242414e-05,
+      "loss": 0.6185,
+      "step": 13744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5577908631115976e-05,
+      "loss": 0.6543,
+      "step": 13745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557731697964655e-05,
+      "loss": 0.6307,
+      "step": 13746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5576725299837142e-05,
+      "loss": 0.5556,
+      "step": 13747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5576133591690764e-05,
+      "loss": 0.6429,
+      "step": 13748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5575541855210416e-05,
+      "loss": 0.5806,
+      "step": 13749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5574950090399102e-05,
+      "loss": 0.6508,
+      "step": 13750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557435829725984e-05,
+      "loss": 0.554,
+      "step": 13751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557376647579563e-05,
+      "loss": 0.6243,
+      "step": 13752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557317462600948e-05,
+      "loss": 0.5374,
+      "step": 13753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.557258274790439e-05,
+      "loss": 0.6857,
+      "step": 13754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5571990841483386e-05,
+      "loss": 0.5848,
+      "step": 13755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5571398906749457e-05,
+      "loss": 0.5797,
+      "step": 13756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5570806943705624e-05,
+      "loss": 0.5614,
+      "step": 13757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5570214952354884e-05,
+      "loss": 0.6003,
+      "step": 13758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5569622932700258e-05,
+      "loss": 0.5928,
+      "step": 13759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5569030884744743e-05,
+      "loss": 0.5603,
+      "step": 13760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5568438808491348e-05,
+      "loss": 0.6349,
+      "step": 13761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5567846703943085e-05,
+      "loss": 0.6387,
+      "step": 13762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.556725457110297e-05,
+      "loss": 0.618,
+      "step": 13763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5566662409973993e-05,
+      "loss": 0.6505,
+      "step": 13764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.556607022055918e-05,
+      "loss": 0.5826,
+      "step": 13765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5565478002861537e-05,
+      "loss": 0.5864,
+      "step": 13766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5564885756884064e-05,
+      "loss": 0.565,
+      "step": 13767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5564293482629782e-05,
+      "loss": 0.5956,
+      "step": 13768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5563701180101692e-05,
+      "loss": 0.5706,
+      "step": 13769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5563108849302804e-05,
+      "loss": 0.5814,
+      "step": 13770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.556251649023614e-05,
+      "loss": 0.599,
+      "step": 13771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5561924102904694e-05,
+      "loss": 0.5686,
+      "step": 13772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5561331687311483e-05,
+      "loss": 0.5774,
+      "step": 13773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.556073924345952e-05,
+      "loss": 0.5654,
+      "step": 13774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5560146771351806e-05,
+      "loss": 0.5614,
+      "step": 13775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5559554270991366e-05,
+      "loss": 0.6217,
+      "step": 13776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5558961742381198e-05,
+      "loss": 0.5798,
+      "step": 13777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5558369185524317e-05,
+      "loss": 0.6333,
+      "step": 13778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5557776600423736e-05,
+      "loss": 0.5972,
+      "step": 13779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5557183987082463e-05,
+      "loss": 0.5939,
+      "step": 13780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5556591345503514e-05,
+      "loss": 0.5943,
+      "step": 13781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.555599867568989e-05,
+      "loss": 0.602,
+      "step": 13782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.555540597764462e-05,
+      "loss": 0.6485,
+      "step": 13783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5554813251370696e-05,
+      "loss": 0.6155,
+      "step": 13784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5554220496871142e-05,
+      "loss": 0.5893,
+      "step": 13785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5553627714148965e-05,
+      "loss": 0.5655,
+      "step": 13786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5553034903207184e-05,
+      "loss": 0.5596,
+      "step": 13787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5552442064048805e-05,
+      "loss": 0.6487,
+      "step": 13788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.555184919667684e-05,
+      "loss": 0.6209,
+      "step": 13789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5551256301094305e-05,
+      "loss": 0.5988,
+      "step": 13790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5550663377304208e-05,
+      "loss": 0.6157,
+      "step": 13791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5550070425309568e-05,
+      "loss": 0.6126,
+      "step": 13792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5549477445113396e-05,
+      "loss": 0.6327,
+      "step": 13793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5548884436718697e-05,
+      "loss": 0.5971,
+      "step": 13794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5548291400128498e-05,
+      "loss": 0.6873,
+      "step": 13795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5547698335345804e-05,
+      "loss": 0.6399,
+      "step": 13796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.554710524237363e-05,
+      "loss": 0.541,
+      "step": 13797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5546512121214986e-05,
+      "loss": 0.6165,
+      "step": 13798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5545918971872894e-05,
+      "loss": 0.6077,
+      "step": 13799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.554532579435036e-05,
+      "loss": 0.5782,
+      "step": 13800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5544732588650403e-05,
+      "loss": 0.5906,
+      "step": 13801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.554413935477604e-05,
+      "loss": 0.5866,
+      "step": 13802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5543546092730277e-05,
+      "loss": 0.6428,
+      "step": 13803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5542952802516133e-05,
+      "loss": 0.5875,
+      "step": 13804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5542359484136623e-05,
+      "loss": 0.6183,
+      "step": 13805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.554176613759476e-05,
+      "loss": 0.6129,
+      "step": 13806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5541172762893563e-05,
+      "loss": 0.5789,
+      "step": 13807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.554057936003604e-05,
+      "loss": 0.623,
+      "step": 13808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5539985929025214e-05,
+      "loss": 0.6393,
+      "step": 13809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5539392469864098e-05,
+      "loss": 0.6385,
+      "step": 13810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5538798982555704e-05,
+      "loss": 0.5933,
+      "step": 13811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553820546710305e-05,
+      "loss": 0.5978,
+      "step": 13812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5537611923509155e-05,
+      "loss": 0.6461,
+      "step": 13813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553701835177703e-05,
+      "loss": 0.6447,
+      "step": 13814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5536424751909692e-05,
+      "loss": 0.5969,
+      "step": 13815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5535831123910162e-05,
+      "loss": 0.6322,
+      "step": 13816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553523746778145e-05,
+      "loss": 0.5867,
+      "step": 13817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5534643783526577e-05,
+      "loss": 0.6537,
+      "step": 13818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5534050071148556e-05,
+      "loss": 0.5601,
+      "step": 13819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553345633065041e-05,
+      "loss": 0.6085,
+      "step": 13820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.553286256203515e-05,
+      "loss": 0.5839,
+      "step": 13821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5532268765305794e-05,
+      "loss": 0.6092,
+      "step": 13822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5531674940465358e-05,
+      "loss": 0.7221,
+      "step": 13823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5531081087516863e-05,
+      "loss": 0.6276,
+      "step": 13824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5530487206463326e-05,
+      "loss": 0.6936,
+      "step": 13825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5529893297307764e-05,
+      "loss": 0.556,
+      "step": 13826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5529299360053195e-05,
+      "loss": 0.5971,
+      "step": 13827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552870539470264e-05,
+      "loss": 0.5813,
+      "step": 13828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5528111401259105e-05,
+      "loss": 0.618,
+      "step": 13829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5527517379725622e-05,
+      "loss": 0.6792,
+      "step": 13830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5526923330105204e-05,
+      "loss": 0.574,
+      "step": 13831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552632925240087e-05,
+      "loss": 0.5987,
+      "step": 13832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552573514661564e-05,
+      "loss": 0.6477,
+      "step": 13833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552514101275253e-05,
+      "loss": 0.5538,
+      "step": 13834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552454685081456e-05,
+      "loss": 0.6282,
+      "step": 13835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552395266080475e-05,
+      "loss": 0.6151,
+      "step": 13836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5523358442726117e-05,
+      "loss": 0.6103,
+      "step": 13837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5522764196581685e-05,
+      "loss": 0.6417,
+      "step": 13838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5522169922374472e-05,
+      "loss": 0.5842,
+      "step": 13839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552157562010749e-05,
+      "loss": 0.5923,
+      "step": 13840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.552098128978377e-05,
+      "loss": 0.6001,
+      "step": 13841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5520386931406326e-05,
+      "loss": 0.5961,
+      "step": 13842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5519792544978177e-05,
+      "loss": 0.5731,
+      "step": 13843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5519198130502353e-05,
+      "loss": 0.6218,
+      "step": 13844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551860368798186e-05,
+      "loss": 0.6147,
+      "step": 13845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5518009217419728e-05,
+      "loss": 0.6304,
+      "step": 13846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5517414718818976e-05,
+      "loss": 0.6169,
+      "step": 13847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5516820192182623e-05,
+      "loss": 0.5998,
+      "step": 13848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551622563751369e-05,
+      "loss": 0.5745,
+      "step": 13849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5515631054815198e-05,
+      "loss": 0.6064,
+      "step": 13850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5515036444090174e-05,
+      "loss": 0.6446,
+      "step": 13851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551444180534163e-05,
+      "loss": 0.6153,
+      "step": 13852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5513847138572594e-05,
+      "loss": 0.5687,
+      "step": 13853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5513252443786085e-05,
+      "loss": 0.5226,
+      "step": 13854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5512657720985128e-05,
+      "loss": 0.6513,
+      "step": 13855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551206297017274e-05,
+      "loss": 0.6013,
+      "step": 13856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5511468191351947e-05,
+      "loss": 0.5868,
+      "step": 13857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551087338452577e-05,
+      "loss": 0.6509,
+      "step": 13858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.551027854969723e-05,
+      "loss": 0.6077,
+      "step": 13859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5509683686869353e-05,
+      "loss": 0.5922,
+      "step": 13860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5509088796045156e-05,
+      "loss": 0.5597,
+      "step": 13861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.550849387722767e-05,
+      "loss": 0.5382,
+      "step": 13862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.550789893041991e-05,
+      "loss": 0.5922,
+      "step": 13863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5507303955624907e-05,
+      "loss": 0.5768,
+      "step": 13864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5506708952845676e-05,
+      "loss": 0.5629,
+      "step": 13865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5506113922085243e-05,
+      "loss": 0.6165,
+      "step": 13866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5505518863346634e-05,
+      "loss": 0.6657,
+      "step": 13867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5504923776632875e-05,
+      "loss": 0.6634,
+      "step": 13868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.550432866194698e-05,
+      "loss": 0.6197,
+      "step": 13869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5503733519291982e-05,
+      "loss": 0.5742,
+      "step": 13870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5503138348670902e-05,
+      "loss": 0.5937,
+      "step": 13871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5502543150086764e-05,
+      "loss": 0.5897,
+      "step": 13872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5501947923542594e-05,
+      "loss": 0.5927,
+      "step": 13873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5501352669041417e-05,
+      "loss": 0.6213,
+      "step": 13874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5500757386586252e-05,
+      "loss": 0.6417,
+      "step": 13875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.550016207618013e-05,
+      "loss": 0.5911,
+      "step": 13876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.549956673782607e-05,
+      "loss": 0.6004,
+      "step": 13877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.54989713715271e-05,
+      "loss": 0.6158,
+      "step": 13878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5498375977286254e-05,
+      "loss": 0.6005,
+      "step": 13879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5497780555106547e-05,
+      "loss": 0.5595,
+      "step": 13880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5497185104991005e-05,
+      "loss": 0.6357,
+      "step": 13881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5496589626942654e-05,
+      "loss": 0.6506,
+      "step": 13882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.549599412096452e-05,
+      "loss": 0.6198,
+      "step": 13883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5495398587059635e-05,
+      "loss": 0.63,
+      "step": 13884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.549480302523102e-05,
+      "loss": 0.6611,
+      "step": 13885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.54942074354817e-05,
+      "loss": 0.5551,
+      "step": 13886
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5493611817814702e-05,
+      "loss": 0.6456,
+      "step": 13887
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5493016172233056e-05,
+      "loss": 0.5916,
+      "step": 13888
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5492420498739785e-05,
+      "loss": 0.6554,
+      "step": 13889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5491824797337916e-05,
+      "loss": 0.6634,
+      "step": 13890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5491229068030475e-05,
+      "loss": 0.6729,
+      "step": 13891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5490633310820492e-05,
+      "loss": 0.5508,
+      "step": 13892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5490037525710996e-05,
+      "loss": 0.5945,
+      "step": 13893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548944171270501e-05,
+      "loss": 0.5499,
+      "step": 13894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5488845871805564e-05,
+      "loss": 0.6112,
+      "step": 13895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5488250003015683e-05,
+      "loss": 0.6234,
+      "step": 13896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5487654106338398e-05,
+      "loss": 0.6471,
+      "step": 13897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5487058181776734e-05,
+      "loss": 0.573,
+      "step": 13898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548646222933372e-05,
+      "loss": 0.6252,
+      "step": 13899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5485866249012386e-05,
+      "loss": 0.649,
+      "step": 13900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548527024081576e-05,
+      "loss": 0.6243,
+      "step": 13901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5484674204746867e-05,
+      "loss": 0.6431,
+      "step": 13902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548407814080874e-05,
+      "loss": 0.6168,
+      "step": 13903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5483482049004403e-05,
+      "loss": 0.5493,
+      "step": 13904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5482885929336892e-05,
+      "loss": 0.6026,
+      "step": 13905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5482289781809227e-05,
+      "loss": 0.6006,
+      "step": 13906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5481693606424445e-05,
+      "loss": 0.6404,
+      "step": 13907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548109740318557e-05,
+      "loss": 0.5689,
+      "step": 13908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.548050117209564e-05,
+      "loss": 0.6439,
+      "step": 13909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5479904913157674e-05,
+      "loss": 0.6184,
+      "step": 13910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5479308626374706e-05,
+      "loss": 0.6056,
+      "step": 13911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5478712311749766e-05,
+      "loss": 0.6408,
+      "step": 13912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5478115969285885e-05,
+      "loss": 0.6578,
+      "step": 13913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5477519598986092e-05,
+      "loss": 0.6658,
+      "step": 13914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.547692320085342e-05,
+      "loss": 0.5746,
+      "step": 13915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5476326774890894e-05,
+      "loss": 0.6375,
+      "step": 13916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.547573032110155e-05,
+      "loss": 0.6203,
+      "step": 13917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5475133839488415e-05,
+      "loss": 0.5927,
+      "step": 13918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5474537330054523e-05,
+      "loss": 0.5979,
+      "step": 13919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5473940792802903e-05,
+      "loss": 0.5729,
+      "step": 13920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5473344227736585e-05,
+      "loss": 0.659,
+      "step": 13921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5472747634858608e-05,
+      "loss": 0.6107,
+      "step": 13922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.547215101417199e-05,
+      "loss": 0.6156,
+      "step": 13923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5471554365679773e-05,
+      "loss": 0.5475,
+      "step": 13924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5470957689384988e-05,
+      "loss": 0.6511,
+      "step": 13925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5470360985290663e-05,
+      "loss": 0.6086,
+      "step": 13926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5469764253399827e-05,
+      "loss": 0.6629,
+      "step": 13927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5469167493715526e-05,
+      "loss": 0.5629,
+      "step": 13928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5468570706240775e-05,
+      "loss": 0.6332,
+      "step": 13929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5467973890978624e-05,
+      "loss": 0.6321,
+      "step": 13930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5467377047932088e-05,
+      "loss": 0.5971,
+      "step": 13931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.546678017710421e-05,
+      "loss": 0.5419,
+      "step": 13932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.546618327849802e-05,
+      "loss": 0.6363,
+      "step": 13933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5465586352116555e-05,
+      "loss": 0.6222,
+      "step": 13934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5464989397962843e-05,
+      "loss": 0.5882,
+      "step": 13935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5464392416039922e-05,
+      "loss": 0.5629,
+      "step": 13936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.546379540635082e-05,
+      "loss": 0.6145,
+      "step": 13937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5463198368898572e-05,
+      "loss": 0.6757,
+      "step": 13938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5462601303686217e-05,
+      "loss": 0.5596,
+      "step": 13939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5462004210716782e-05,
+      "loss": 0.5724,
+      "step": 13940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5461407089993307e-05,
+      "loss": 0.6129,
+      "step": 13941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5460809941518823e-05,
+      "loss": 0.5868,
+      "step": 13942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5460212765296363e-05,
+      "loss": 0.6248,
+      "step": 13943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.545961556132896e-05,
+      "loss": 0.596,
+      "step": 13944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5459018329619654e-05,
+      "loss": 0.6918,
+      "step": 13945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5458421070171477e-05,
+      "loss": 0.5909,
+      "step": 13946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5457823782987467e-05,
+      "loss": 0.613,
+      "step": 13947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5457226468070653e-05,
+      "loss": 0.5868,
+      "step": 13948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.545662912542407e-05,
+      "loss": 0.5534,
+      "step": 13949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5456031755050764e-05,
+      "loss": 0.5751,
+      "step": 13950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5455434356953756e-05,
+      "loss": 0.6192,
+      "step": 13951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5454836931136094e-05,
+      "loss": 0.6123,
+      "step": 13952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5454239477600803e-05,
+      "loss": 0.5669,
+      "step": 13953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5453641996350923e-05,
+      "loss": 0.5678,
+      "step": 13954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5453044487389493e-05,
+      "loss": 0.5869,
+      "step": 13955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5452446950719552e-05,
+      "loss": 0.5959,
+      "step": 13956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5451849386344125e-05,
+      "loss": 0.6107,
+      "step": 13957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.545125179426626e-05,
+      "loss": 0.6708,
+      "step": 13958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.545065417448898e-05,
+      "loss": 0.6076,
+      "step": 13959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5450056527015337e-05,
+      "loss": 0.6008,
+      "step": 13960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5449458851848357e-05,
+      "loss": 0.6204,
+      "step": 13961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544886114899108e-05,
+      "loss": 0.5725,
+      "step": 13962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5448263418446546e-05,
+      "loss": 0.59,
+      "step": 13963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544766566021779e-05,
+      "loss": 0.5669,
+      "step": 13964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5447067874307848e-05,
+      "loss": 0.6026,
+      "step": 13965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5446470060719757e-05,
+      "loss": 0.6322,
+      "step": 13966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544587221945656e-05,
+      "loss": 0.659,
+      "step": 13967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544527435052129e-05,
+      "loss": 0.587,
+      "step": 13968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544467645391699e-05,
+      "loss": 0.6389,
+      "step": 13969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544407852964669e-05,
+      "loss": 0.5887,
+      "step": 13970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5443480577713432e-05,
+      "loss": 0.6215,
+      "step": 13971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5442882598120256e-05,
+      "loss": 0.6244,
+      "step": 13972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54422845908702e-05,
+      "loss": 0.5817,
+      "step": 13973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5441686555966304e-05,
+      "loss": 0.5929,
+      "step": 13974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5441088493411604e-05,
+      "loss": 0.6368,
+      "step": 13975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544049040320914e-05,
+      "loss": 0.6013,
+      "step": 13976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5439892285361947e-05,
+      "loss": 0.5933,
+      "step": 13977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543929413987307e-05,
+      "loss": 0.5903,
+      "step": 13978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5438695966745552e-05,
+      "loss": 0.591,
+      "step": 13979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5438097765982423e-05,
+      "loss": 0.5886,
+      "step": 13980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5437499537586726e-05,
+      "loss": 0.5744,
+      "step": 13981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54369012815615e-05,
+      "loss": 0.5942,
+      "step": 13982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543630299790979e-05,
+      "loss": 0.6143,
+      "step": 13983
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543570468663463e-05,
+      "loss": 0.5916,
+      "step": 13984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5435106347739063e-05,
+      "loss": 0.6178,
+      "step": 13985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543450798122613e-05,
+      "loss": 0.6221,
+      "step": 13986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5433909587098867e-05,
+      "loss": 0.6133,
+      "step": 13987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5433311165360323e-05,
+      "loss": 0.6246,
+      "step": 13988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543271271601353e-05,
+      "loss": 0.563,
+      "step": 13989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5432114239061533e-05,
+      "loss": 0.6163,
+      "step": 13990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5431515734507372e-05,
+      "loss": 0.5938,
+      "step": 13991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5430917202354092e-05,
+      "loss": 0.6239,
+      "step": 13992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.543031864260473e-05,
+      "loss": 0.5995,
+      "step": 13993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5429720055262327e-05,
+      "loss": 0.5418,
+      "step": 13994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5429121440329927e-05,
+      "loss": 0.6063,
+      "step": 13995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5428522797810572e-05,
+      "loss": 0.6307,
+      "step": 13996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54279241277073e-05,
+      "loss": 0.5992,
+      "step": 13997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5427325430023163e-05,
+      "loss": 0.5879,
+      "step": 13998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5426726704761187e-05,
+      "loss": 0.6059,
+      "step": 13999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5426127951924427e-05,
+      "loss": 0.6055,
+      "step": 14000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5425529171515924e-05,
+      "loss": 0.6297,
+      "step": 14001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5424930363538715e-05,
+      "loss": 0.6298,
+      "step": 14002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.542433152799585e-05,
+      "loss": 0.6319,
+      "step": 14003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5423732664890364e-05,
+      "loss": 0.5856,
+      "step": 14004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5423133774225304e-05,
+      "loss": 0.5989,
+      "step": 14005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5422534856003717e-05,
+      "loss": 0.539,
+      "step": 14006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5421935910228637e-05,
+      "loss": 0.633,
+      "step": 14007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.542133693690312e-05,
+      "loss": 0.6404,
+      "step": 14008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54207379360302e-05,
+      "loss": 0.5782,
+      "step": 14009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5420138907612915e-05,
+      "loss": 0.6212,
+      "step": 14010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5419539851654322e-05,
+      "loss": 0.6361,
+      "step": 14011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541894076815746e-05,
+      "loss": 0.6295,
+      "step": 14012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5418341657125375e-05,
+      "loss": 0.6046,
+      "step": 14013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541774251856111e-05,
+      "loss": 0.6047,
+      "step": 14014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5417143352467702e-05,
+      "loss": 0.6533,
+      "step": 14015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541654415884821e-05,
+      "loss": 0.5973,
+      "step": 14016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5415944937705666e-05,
+      "loss": 0.6049,
+      "step": 14017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5415345689043123e-05,
+      "loss": 0.5535,
+      "step": 14018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5414746412863617e-05,
+      "loss": 0.5905,
+      "step": 14019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5414147109170205e-05,
+      "loss": 0.6342,
+      "step": 14020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541354777796592e-05,
+      "loss": 0.6213,
+      "step": 14021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541294841925382e-05,
+      "loss": 0.5896,
+      "step": 14022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5412349033036937e-05,
+      "loss": 0.6276,
+      "step": 14023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.541174961931833e-05,
+      "loss": 0.577,
+      "step": 14024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5411150178101033e-05,
+      "loss": 0.6564,
+      "step": 14025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54105507093881e-05,
+      "loss": 0.6075,
+      "step": 14026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.540995121318257e-05,
+      "loss": 0.5893,
+      "step": 14027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54093516894875e-05,
+      "loss": 0.6061,
+      "step": 14028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5408752138305928e-05,
+      "loss": 0.5937,
+      "step": 14029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54081525596409e-05,
+      "loss": 0.5505,
+      "step": 14030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5407552953495465e-05,
+      "loss": 0.5623,
+      "step": 14031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.540695331987267e-05,
+      "loss": 0.5858,
+      "step": 14032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5406353658775565e-05,
+      "loss": 0.5768,
+      "step": 14033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.540575397020719e-05,
+      "loss": 0.5885,
+      "step": 14034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5405154254170596e-05,
+      "loss": 0.563,
+      "step": 14035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.540455451066883e-05,
+      "loss": 0.6249,
+      "step": 14036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5403954739704943e-05,
+      "loss": 0.58,
+      "step": 14037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5403354941281978e-05,
+      "loss": 0.694,
+      "step": 14038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5402755115402985e-05,
+      "loss": 0.6045,
+      "step": 14039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.540215526207101e-05,
+      "loss": 0.6496,
+      "step": 14040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5401555381289105e-05,
+      "loss": 0.5766,
+      "step": 14041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.540095547306031e-05,
+      "loss": 0.5524,
+      "step": 14042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.540035553738768e-05,
+      "loss": 0.5605,
+      "step": 14043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5399755574274262e-05,
+      "loss": 0.5539,
+      "step": 14044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5399155583723103e-05,
+      "loss": 0.6179,
+      "step": 14045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5398555565737255e-05,
+      "loss": 0.633,
+      "step": 14046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5397955520319767e-05,
+      "loss": 0.662,
+      "step": 14047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5397355447473682e-05,
+      "loss": 0.6031,
+      "step": 14048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5396755347202056e-05,
+      "loss": 0.5866,
+      "step": 14049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5396155219507938e-05,
+      "loss": 0.5543,
+      "step": 14050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.539555506439437e-05,
+      "loss": 0.593,
+      "step": 14051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.539495488186441e-05,
+      "loss": 0.5527,
+      "step": 14052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.53943546719211e-05,
+      "loss": 0.5736,
+      "step": 14053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5393754434567498e-05,
+      "loss": 0.6391,
+      "step": 14054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.539315416980665e-05,
+      "loss": 0.6122,
+      "step": 14055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5392553877641608e-05,
+      "loss": 0.5855,
+      "step": 14056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5391953558075415e-05,
+      "loss": 0.6238,
+      "step": 14057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5391353211111132e-05,
+      "loss": 0.5955,
+      "step": 14058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5390752836751804e-05,
+      "loss": 0.6689,
+      "step": 14059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.539015243500048e-05,
+      "loss": 0.6113,
+      "step": 14060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5389552005860212e-05,
+      "loss": 0.567,
+      "step": 14061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5388951549334054e-05,
+      "loss": 0.6213,
+      "step": 14062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5388351065425054e-05,
+      "loss": 0.6243,
+      "step": 14063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5387750554136264e-05,
+      "loss": 0.6393,
+      "step": 14064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5387150015470735e-05,
+      "loss": 0.581,
+      "step": 14065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.538654944943152e-05,
+      "loss": 0.6506,
+      "step": 14066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.538594885602167e-05,
+      "loss": 0.5602,
+      "step": 14067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.538534823524424e-05,
+      "loss": 0.6203,
+      "step": 14068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.538474758710227e-05,
+      "loss": 0.6391,
+      "step": 14069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5384146911598823e-05,
+      "loss": 0.6522,
+      "step": 14070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5383546208736953e-05,
+      "loss": 0.614,
+      "step": 14071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5382945478519708e-05,
+      "loss": 0.6037,
+      "step": 14072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5382344720950135e-05,
+      "loss": 0.6207,
+      "step": 14073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5381743936031295e-05,
+      "loss": 0.6266,
+      "step": 14074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.538114312376624e-05,
+      "loss": 0.5893,
+      "step": 14075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5380542284158017e-05,
+      "loss": 0.6476,
+      "step": 14076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5379941417209687e-05,
+      "loss": 0.5807,
+      "step": 14077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5379340522924297e-05,
+      "loss": 0.6162,
+      "step": 14078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5378739601304907e-05,
+      "loss": 0.6357,
+      "step": 14079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537813865235456e-05,
+      "loss": 0.5732,
+      "step": 14080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5377537676076316e-05,
+      "loss": 0.5552,
+      "step": 14081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537693667247323e-05,
+      "loss": 0.6246,
+      "step": 14082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537633564154835e-05,
+      "loss": 0.6424,
+      "step": 14083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537573458330474e-05,
+      "loss": 0.6107,
+      "step": 14084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5375133497745445e-05,
+      "loss": 0.5681,
+      "step": 14085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537453238487352e-05,
+      "loss": 0.6177,
+      "step": 14086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537393124469203e-05,
+      "loss": 0.5795,
+      "step": 14087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5373330077204013e-05,
+      "loss": 0.6408,
+      "step": 14088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5372728882412537e-05,
+      "loss": 0.6179,
+      "step": 14089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537212766032065e-05,
+      "loss": 0.5688,
+      "step": 14090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537152641093141e-05,
+      "loss": 0.6388,
+      "step": 14091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537092513424787e-05,
+      "loss": 0.5867,
+      "step": 14092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.537032383027309e-05,
+      "loss": 0.6145,
+      "step": 14093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5369722499010114e-05,
+      "loss": 0.6407,
+      "step": 14094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5369121140462016e-05,
+      "loss": 0.569,
+      "step": 14095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5368519754631833e-05,
+      "loss": 0.6292,
+      "step": 14096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5367918341522636e-05,
+      "loss": 0.6213,
+      "step": 14097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5367316901137466e-05,
+      "loss": 0.6364,
+      "step": 14098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5366715433479395e-05,
+      "loss": 0.582,
+      "step": 14099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.536611393855146e-05,
+      "loss": 0.6441,
+      "step": 14100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.536551241635674e-05,
+      "loss": 0.5749,
+      "step": 14101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5364910866898272e-05,
+      "loss": 0.5701,
+      "step": 14102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5364309290179125e-05,
+      "loss": 0.6119,
+      "step": 14103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.536370768620235e-05,
+      "loss": 0.5556,
+      "step": 14104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5363106054971003e-05,
+      "loss": 0.5931,
+      "step": 14105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5362504396488152e-05,
+      "loss": 0.5901,
+      "step": 14106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5361902710756837e-05,
+      "loss": 0.5411,
+      "step": 14107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5361300997780127e-05,
+      "loss": 0.5829,
+      "step": 14108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5360699257561077e-05,
+      "loss": 0.6201,
+      "step": 14109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5360097490102744e-05,
+      "loss": 0.6831,
+      "step": 14110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5359495695408185e-05,
+      "loss": 0.6143,
+      "step": 14111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5358893873480457e-05,
+      "loss": 0.5697,
+      "step": 14112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5358292024322622e-05,
+      "loss": 0.5355,
+      "step": 14113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.535769014793774e-05,
+      "loss": 0.6567,
+      "step": 14114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.535708824432886e-05,
+      "loss": 0.6053,
+      "step": 14115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5356486313499046e-05,
+      "loss": 0.6181,
+      "step": 14116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.535588435545136e-05,
+      "loss": 0.5816,
+      "step": 14117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5355282370188854e-05,
+      "loss": 0.6485,
+      "step": 14118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.535468035771459e-05,
+      "loss": 0.5674,
+      "step": 14119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5354078318031628e-05,
+      "loss": 0.6138,
+      "step": 14120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5353476251143025e-05,
+      "loss": 0.655,
+      "step": 14121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.535287415705184e-05,
+      "loss": 0.5896,
+      "step": 14122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5352272035761137e-05,
+      "loss": 0.6395,
+      "step": 14123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5351669887273967e-05,
+      "loss": 0.6053,
+      "step": 14124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.53510677115934e-05,
+      "loss": 0.6379,
+      "step": 14125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5350465508722488e-05,
+      "loss": 0.667,
+      "step": 14126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5349863278664298e-05,
+      "loss": 0.5891,
+      "step": 14127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5349261021421882e-05,
+      "loss": 0.6002,
+      "step": 14128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5348658736998305e-05,
+      "loss": 0.6176,
+      "step": 14129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5348056425396625e-05,
+      "loss": 0.5782,
+      "step": 14130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534745408661991e-05,
+      "loss": 0.6559,
+      "step": 14131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5346851720671206e-05,
+      "loss": 0.6325,
+      "step": 14132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534624932755359e-05,
+      "loss": 0.6023,
+      "step": 14133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5345646907270112e-05,
+      "loss": 0.556,
+      "step": 14134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5345044459823836e-05,
+      "loss": 0.5363,
+      "step": 14135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5344441985217825e-05,
+      "loss": 0.6365,
+      "step": 14136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534383948345514e-05,
+      "loss": 0.5172,
+      "step": 14137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534323695453884e-05,
+      "loss": 0.5501,
+      "step": 14138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534263439847199e-05,
+      "loss": 0.5764,
+      "step": 14139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.534203181525765e-05,
+      "loss": 0.6377,
+      "step": 14140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5341429204898877e-05,
+      "loss": 0.6814,
+      "step": 14141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5340826567398743e-05,
+      "loss": 0.613,
+      "step": 14142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5340223902760304e-05,
+      "loss": 0.5287,
+      "step": 14143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5339621210986622e-05,
+      "loss": 0.6441,
+      "step": 14144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.533901849208076e-05,
+      "loss": 0.5919,
+      "step": 14145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5338415746045786e-05,
+      "loss": 0.5566,
+      "step": 14146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5337812972884757e-05,
+      "loss": 0.5633,
+      "step": 14147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5337210172600734e-05,
+      "loss": 0.6133,
+      "step": 14148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5336607345196782e-05,
+      "loss": 0.6235,
+      "step": 14149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.533600449067597e-05,
+      "loss": 0.667,
+      "step": 14150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5335401609041353e-05,
+      "loss": 0.6068,
+      "step": 14151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5334798700295997e-05,
+      "loss": 0.5863,
+      "step": 14152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.533419576444297e-05,
+      "loss": 0.6518,
+      "step": 14153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.533359280148533e-05,
+      "loss": 0.5892,
+      "step": 14154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5332989811426142e-05,
+      "loss": 0.6057,
+      "step": 14155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5332386794268474e-05,
+      "loss": 0.6737,
+      "step": 14156
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5331783750015384e-05,
+      "loss": 0.6376,
+      "step": 14157
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5331180678669945e-05,
+      "loss": 0.6208,
+      "step": 14158
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.533057758023521e-05,
+      "loss": 0.5727,
+      "step": 14159
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5329974454714252e-05,
+      "loss": 0.7217,
+      "step": 14160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5329371302110127e-05,
+      "loss": 0.6515,
+      "step": 14161
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.532876812242591e-05,
+      "loss": 0.6257,
+      "step": 14162
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5328164915664664e-05,
+      "loss": 0.6297,
+      "step": 14163
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.532756168182945e-05,
+      "loss": 0.6508,
+      "step": 14164
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5326958420923332e-05,
+      "loss": 0.5927,
+      "step": 14165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5326355132949383e-05,
+      "loss": 0.6142,
+      "step": 14166
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5325751817910658e-05,
+      "loss": 0.5609,
+      "step": 14167
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.532514847581023e-05,
+      "loss": 0.6023,
+      "step": 14168
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5324545106651164e-05,
+      "loss": 0.5433,
+      "step": 14169
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5323941710436528e-05,
+      "loss": 0.6238,
+      "step": 14170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5323338287169384e-05,
+      "loss": 0.6737,
+      "step": 14171
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5322734836852795e-05,
+      "loss": 0.6312,
+      "step": 14172
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.532213135948983e-05,
+      "loss": 0.6367,
+      "step": 14173
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.532152785508356e-05,
+      "loss": 0.6339,
+      "step": 14174
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5320924323637056e-05,
+      "loss": 0.6088,
+      "step": 14175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5320320765153367e-05,
+      "loss": 0.6246,
+      "step": 14176
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5319717179635574e-05,
+      "loss": 0.5807,
+      "step": 14177
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5319113567086738e-05,
+      "loss": 0.5503,
+      "step": 14178
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.531850992750993e-05,
+      "loss": 0.6219,
+      "step": 14179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5317906260908216e-05,
+      "loss": 0.6028,
+      "step": 14180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5317302567284662e-05,
+      "loss": 0.5958,
+      "step": 14181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5316698846642336e-05,
+      "loss": 0.5663,
+      "step": 14182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5316095098984308e-05,
+      "loss": 0.6012,
+      "step": 14183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.531549132431364e-05,
+      "loss": 0.6135,
+      "step": 14184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.531488752263341e-05,
+      "loss": 0.5798,
+      "step": 14185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5314283693946674e-05,
+      "loss": 0.6058,
+      "step": 14186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5313679838256508e-05,
+      "loss": 0.5772,
+      "step": 14187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5313075955565978e-05,
+      "loss": 0.6053,
+      "step": 14188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5312472045878153e-05,
+      "loss": 0.5905,
+      "step": 14189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5311868109196107e-05,
+      "loss": 0.5832,
+      "step": 14190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5311264145522898e-05,
+      "loss": 0.6364,
+      "step": 14191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.53106601548616e-05,
+      "loss": 0.6731,
+      "step": 14192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5310056137215286e-05,
+      "loss": 0.6492,
+      "step": 14193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5309452092587022e-05,
+      "loss": 0.5852,
+      "step": 14194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.530884802097987e-05,
+      "loss": 0.6134,
+      "step": 14195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5308243922396916e-05,
+      "loss": 0.5957,
+      "step": 14196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5307639796841213e-05,
+      "loss": 0.618,
+      "step": 14197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5307035644315842e-05,
+      "loss": 0.6929,
+      "step": 14198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5306431464823863e-05,
+      "loss": 0.6241,
+      "step": 14199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5305827258368355e-05,
+      "loss": 0.6317,
+      "step": 14200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5305223024952388e-05,
+      "loss": 0.5773,
+      "step": 14201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5304618764579025e-05,
+      "loss": 0.5472,
+      "step": 14202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5304014477251342e-05,
+      "loss": 0.6466,
+      "step": 14203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5303410162972406e-05,
+      "loss": 0.5902,
+      "step": 14204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5302805821745293e-05,
+      "loss": 0.5928,
+      "step": 14205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5302201453573072e-05,
+      "loss": 0.6012,
+      "step": 14206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5301597058458807e-05,
+      "loss": 0.5943,
+      "step": 14207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5300992636405577e-05,
+      "loss": 0.5905,
+      "step": 14208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5300388187416452e-05,
+      "loss": 0.5992,
+      "step": 14209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.52997837114945e-05,
+      "loss": 0.5953,
+      "step": 14210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.52991792086428e-05,
+      "loss": 0.5263,
+      "step": 14211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5298574678864413e-05,
+      "loss": 0.5926,
+      "step": 14212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5297970122162418e-05,
+      "loss": 0.6004,
+      "step": 14213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5297365538539887e-05,
+      "loss": 0.6357,
+      "step": 14214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.529676092799989e-05,
+      "loss": 0.6414,
+      "step": 14215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5296156290545495e-05,
+      "loss": 0.6151,
+      "step": 14216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5295551626179785e-05,
+      "loss": 0.5502,
+      "step": 14217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5294946934905826e-05,
+      "loss": 0.59,
+      "step": 14218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5294342216726687e-05,
+      "loss": 0.5765,
+      "step": 14219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5293737471645448e-05,
+      "loss": 0.6506,
+      "step": 14220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5293132699665174e-05,
+      "loss": 0.6195,
+      "step": 14221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5292527900788945e-05,
+      "loss": 0.5838,
+      "step": 14222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5291923075019832e-05,
+      "loss": 0.6961,
+      "step": 14223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.529131822236091e-05,
+      "loss": 0.6313,
+      "step": 14224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.529071334281525e-05,
+      "loss": 0.5817,
+      "step": 14225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5290108436385923e-05,
+      "loss": 0.6375,
+      "step": 14226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5289503503076003e-05,
+      "loss": 0.6153,
+      "step": 14227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5288898542888576e-05,
+      "loss": 0.689,
+      "step": 14228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.52882935558267e-05,
+      "loss": 0.6785,
+      "step": 14229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5287688541893455e-05,
+      "loss": 0.591,
+      "step": 14230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.528708350109192e-05,
+      "loss": 0.5767,
+      "step": 14231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5286478433425163e-05,
+      "loss": 0.6232,
+      "step": 14232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.528587333889626e-05,
+      "loss": 0.5481,
+      "step": 14233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5285268217508293e-05,
+      "loss": 0.6427,
+      "step": 14234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5284663069264322e-05,
+      "loss": 0.6061,
+      "step": 14235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5284057894167438e-05,
+      "loss": 0.5744,
+      "step": 14236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.52834526922207e-05,
+      "loss": 0.6495,
+      "step": 14237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5282847463427196e-05,
+      "loss": 0.5558,
+      "step": 14238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5282242207789998e-05,
+      "loss": 0.5955,
+      "step": 14239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5281636925312176e-05,
+      "loss": 0.626,
+      "step": 14240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5281031615996813e-05,
+      "loss": 0.6547,
+      "step": 14241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.528042627984698e-05,
+      "loss": 0.6773,
+      "step": 14242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527982091686576e-05,
+      "loss": 0.6589,
+      "step": 14243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5279215527056218e-05,
+      "loss": 0.5754,
+      "step": 14244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5278610110421438e-05,
+      "loss": 0.5504,
+      "step": 14245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5278004666964493e-05,
+      "loss": 0.6149,
+      "step": 14246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527739919668846e-05,
+      "loss": 0.5912,
+      "step": 14247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5276793699596417e-05,
+      "loss": 0.6163,
+      "step": 14248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527618817569144e-05,
+      "loss": 0.6348,
+      "step": 14249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5275582624976607e-05,
+      "loss": 0.6341,
+      "step": 14250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5274977047454993e-05,
+      "loss": 0.6178,
+      "step": 14251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5274371443129675e-05,
+      "loss": 0.6473,
+      "step": 14252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527376581200373e-05,
+      "loss": 0.6162,
+      "step": 14253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527316015408024e-05,
+      "loss": 0.5686,
+      "step": 14254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527255446936228e-05,
+      "loss": 0.5899,
+      "step": 14255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.527194875785292e-05,
+      "loss": 0.6305,
+      "step": 14256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5271343019555248e-05,
+      "loss": 0.6106,
+      "step": 14257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5270737254472337e-05,
+      "loss": 0.6366,
+      "step": 14258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5270131462607267e-05,
+      "loss": 0.6613,
+      "step": 14259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5269525643963115e-05,
+      "loss": 0.5878,
+      "step": 14260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.526891979854296e-05,
+      "loss": 0.6666,
+      "step": 14261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.526831392634988e-05,
+      "loss": 0.5506,
+      "step": 14262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5267708027386962e-05,
+      "loss": 0.627,
+      "step": 14263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5267102101657266e-05,
+      "loss": 0.6757,
+      "step": 14264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5266496149163884e-05,
+      "loss": 0.604,
+      "step": 14265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5265890169909898e-05,
+      "loss": 0.5838,
+      "step": 14266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.526528416389838e-05,
+      "loss": 0.6722,
+      "step": 14267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5264678131132406e-05,
+      "loss": 0.6253,
+      "step": 14268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5264072071615065e-05,
+      "loss": 0.5888,
+      "step": 14269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5263465985349434e-05,
+      "loss": 0.5628,
+      "step": 14270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5262859872338585e-05,
+      "loss": 0.5861,
+      "step": 14271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.526225373258561e-05,
+      "loss": 0.619,
+      "step": 14272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5261647566093577e-05,
+      "loss": 0.5903,
+      "step": 14273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5261041372865574e-05,
+      "loss": 0.6249,
+      "step": 14274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5260435152904683e-05,
+      "loss": 0.6655,
+      "step": 14275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5259828906213973e-05,
+      "loss": 0.5519,
+      "step": 14276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5259222632796537e-05,
+      "loss": 0.5847,
+      "step": 14277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.525861633265545e-05,
+      "loss": 0.6217,
+      "step": 14278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5258010005793791e-05,
+      "loss": 0.5912,
+      "step": 14279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5257403652214647e-05,
+      "loss": 0.6295,
+      "step": 14280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5256797271921091e-05,
+      "loss": 0.5797,
+      "step": 14281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5256190864916214e-05,
+      "loss": 0.6275,
+      "step": 14282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5255584431203087e-05,
+      "loss": 0.5476,
+      "step": 14283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5254977970784797e-05,
+      "loss": 0.5752,
+      "step": 14284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5254371483664425e-05,
+      "loss": 0.6206,
+      "step": 14285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5253764969845053e-05,
+      "loss": 0.5824,
+      "step": 14286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5253158429329764e-05,
+      "loss": 0.6585,
+      "step": 14287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5252551862121638e-05,
+      "loss": 0.6281,
+      "step": 14288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5251945268223756e-05,
+      "loss": 0.5955,
+      "step": 14289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5251338647639204e-05,
+      "loss": 0.6073,
+      "step": 14290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5250732000371062e-05,
+      "loss": 0.5673,
+      "step": 14291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5250125326422413e-05,
+      "loss": 0.5825,
+      "step": 14292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5249518625796338e-05,
+      "loss": 0.558,
+      "step": 14293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5248911898495924e-05,
+      "loss": 0.5411,
+      "step": 14294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.524830514452425e-05,
+      "loss": 0.6275,
+      "step": 14295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5247698363884403e-05,
+      "loss": 0.578,
+      "step": 14296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5247091556579462e-05,
+      "loss": 0.6389,
+      "step": 14297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5246484722612511e-05,
+      "loss": 0.5605,
+      "step": 14298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5245877861986636e-05,
+      "loss": 0.6248,
+      "step": 14299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5245270974704918e-05,
+      "loss": 0.6053,
+      "step": 14300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5244664060770443e-05,
+      "loss": 0.621,
+      "step": 14301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5244057120186294e-05,
+      "loss": 0.6658,
+      "step": 14302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.524345015295555e-05,
+      "loss": 0.5515,
+      "step": 14303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5242843159081307e-05,
+      "loss": 0.6041,
+      "step": 14304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.524223613856664e-05,
+      "loss": 0.6193,
+      "step": 14305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5241629091414636e-05,
+      "loss": 0.5212,
+      "step": 14306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5241022017628379e-05,
+      "loss": 0.585,
+      "step": 14307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5240414917210956e-05,
+      "loss": 0.6416,
+      "step": 14308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5239807790165449e-05,
+      "loss": 0.5733,
+      "step": 14309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5239200636494942e-05,
+      "loss": 0.6735,
+      "step": 14310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5238593456202524e-05,
+      "loss": 0.5749,
+      "step": 14311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5237986249291279e-05,
+      "loss": 0.6238,
+      "step": 14312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5237379015764293e-05,
+      "loss": 0.6244,
+      "step": 14313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5236771755624648e-05,
+      "loss": 0.6287,
+      "step": 14314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5236164468875433e-05,
+      "loss": 0.5699,
+      "step": 14315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5235557155519733e-05,
+      "loss": 0.6454,
+      "step": 14316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5234949815560633e-05,
+      "loss": 0.6147,
+      "step": 14317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.523434244900122e-05,
+      "loss": 0.6083,
+      "step": 14318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.523373505584458e-05,
+      "loss": 0.6177,
+      "step": 14319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.52331276360938e-05,
+      "loss": 0.5978,
+      "step": 14320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5232520189751965e-05,
+      "loss": 0.5961,
+      "step": 14321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5231912716822164e-05,
+      "loss": 0.6126,
+      "step": 14322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5231305217307481e-05,
+      "loss": 0.5349,
+      "step": 14323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5230697691211004e-05,
+      "loss": 0.6652,
+      "step": 14324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5230090138535821e-05,
+      "loss": 0.6284,
+      "step": 14325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5229482559285018e-05,
+      "loss": 0.5909,
+      "step": 14326
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5228874953461681e-05,
+      "loss": 0.6482,
+      "step": 14327
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5228267321068903e-05,
+      "loss": 0.5986,
+      "step": 14328
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5227659662109764e-05,
+      "loss": 0.5949,
+      "step": 14329
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522705197658736e-05,
+      "loss": 0.5708,
+      "step": 14330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522644426450477e-05,
+      "loss": 0.5887,
+      "step": 14331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5225836525865085e-05,
+      "loss": 0.5889,
+      "step": 14332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5225228760671395e-05,
+      "loss": 0.5864,
+      "step": 14333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5224620968926788e-05,
+      "loss": 0.6645,
+      "step": 14334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5224013150634353e-05,
+      "loss": 0.5641,
+      "step": 14335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5223405305797179e-05,
+      "loss": 0.5909,
+      "step": 14336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5222797434418348e-05,
+      "loss": 0.5637,
+      "step": 14337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5222189536500955e-05,
+      "loss": 0.6402,
+      "step": 14338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.522158161204809e-05,
+      "loss": 0.632,
+      "step": 14339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5220973661062837e-05,
+      "loss": 0.5516,
+      "step": 14340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5220365683548285e-05,
+      "loss": 0.5908,
+      "step": 14341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5219757679507532e-05,
+      "loss": 0.5742,
+      "step": 14342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5219149648943661e-05,
+      "loss": 0.6436,
+      "step": 14343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.521854159185976e-05,
+      "loss": 0.6068,
+      "step": 14344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5217933508258922e-05,
+      "loss": 0.5771,
+      "step": 14345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5217325398144234e-05,
+      "loss": 0.5764,
+      "step": 14346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5216717261518788e-05,
+      "loss": 0.7038,
+      "step": 14347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5216109098385677e-05,
+      "loss": 0.6409,
+      "step": 14348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5215500908747985e-05,
+      "loss": 0.6727,
+      "step": 14349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5214892692608804e-05,
+      "loss": 0.616,
+      "step": 14350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.521428444997123e-05,
+      "loss": 0.5946,
+      "step": 14351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5213676180838347e-05,
+      "loss": 0.6256,
+      "step": 14352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.521306788521325e-05,
+      "loss": 0.6465,
+      "step": 14353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5212459563099026e-05,
+      "loss": 0.5638,
+      "step": 14354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5211851214498771e-05,
+      "loss": 0.5905,
+      "step": 14355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5211242839415574e-05,
+      "loss": 0.5818,
+      "step": 14356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5210634437852523e-05,
+      "loss": 0.6158,
+      "step": 14357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5210026009812714e-05,
+      "loss": 0.6042,
+      "step": 14358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5209417555299237e-05,
+      "loss": 0.6524,
+      "step": 14359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5208809074315183e-05,
+      "loss": 0.619,
+      "step": 14360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5208200566863647e-05,
+      "loss": 0.5746,
+      "step": 14361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5207592032947713e-05,
+      "loss": 0.566,
+      "step": 14362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5206983472570485e-05,
+      "loss": 0.608,
+      "step": 14363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5206374885735048e-05,
+      "loss": 0.5935,
+      "step": 14364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5205766272444491e-05,
+      "loss": 0.6655,
+      "step": 14365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5205157632701916e-05,
+      "loss": 0.6451,
+      "step": 14366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5204548966510408e-05,
+      "loss": 0.6301,
+      "step": 14367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5203940273873063e-05,
+      "loss": 0.6312,
+      "step": 14368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5203331554792974e-05,
+      "loss": 0.6322,
+      "step": 14369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5202722809273231e-05,
+      "loss": 0.6262,
+      "step": 14370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5202114037316936e-05,
+      "loss": 0.6573,
+      "step": 14371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5201505238927173e-05,
+      "loss": 0.6182,
+      "step": 14372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5200896414107037e-05,
+      "loss": 0.5862,
+      "step": 14373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5200287562859623e-05,
+      "loss": 0.5934,
+      "step": 14374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5199678685188027e-05,
+      "loss": 0.5547,
+      "step": 14375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5199069781095337e-05,
+      "loss": 0.6277,
+      "step": 14376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5198460850584656e-05,
+      "loss": 0.6181,
+      "step": 14377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5197851893659068e-05,
+      "loss": 0.6405,
+      "step": 14378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5197242910321674e-05,
+      "loss": 0.6109,
+      "step": 14379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5196633900575568e-05,
+      "loss": 0.6367,
+      "step": 14380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5196024864423842e-05,
+      "loss": 0.6178,
+      "step": 14381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5195415801869592e-05,
+      "loss": 0.5476,
+      "step": 14382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5194806712915916e-05,
+      "loss": 0.6408,
+      "step": 14383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5194197597565899e-05,
+      "loss": 0.5892,
+      "step": 14384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5193588455822648e-05,
+      "loss": 0.6594,
+      "step": 14385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5192979287689251e-05,
+      "loss": 0.6047,
+      "step": 14386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5192370093168806e-05,
+      "loss": 0.5806,
+      "step": 14387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.519176087226441e-05,
+      "loss": 0.6581,
+      "step": 14388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5191151624979154e-05,
+      "loss": 0.5857,
+      "step": 14389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5190542351316135e-05,
+      "loss": 0.5608,
+      "step": 14390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5189933051278453e-05,
+      "loss": 0.5874,
+      "step": 14391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.51893237248692e-05,
+      "loss": 0.5803,
+      "step": 14392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5188714372091476e-05,
+      "loss": 0.6133,
+      "step": 14393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5188104992948374e-05,
+      "loss": 0.5975,
+      "step": 14394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5187495587442986e-05,
+      "loss": 0.6068,
+      "step": 14395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5186886155578418e-05,
+      "loss": 0.628,
+      "step": 14396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5186276697357763e-05,
+      "loss": 0.6006,
+      "step": 14397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5185667212784115e-05,
+      "loss": 0.5849,
+      "step": 14398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5185057701860576e-05,
+      "loss": 0.6174,
+      "step": 14399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5184448164590236e-05,
+      "loss": 0.6102,
+      "step": 14400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.51838386009762e-05,
+      "loss": 0.6355,
+      "step": 14401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.518322901102156e-05,
+      "loss": 0.5624,
+      "step": 14402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5182619394729415e-05,
+      "loss": 0.5967,
+      "step": 14403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5182009752102866e-05,
+      "loss": 0.6081,
+      "step": 14404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5181400083145007e-05,
+      "loss": 0.6648,
+      "step": 14405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5180790387858932e-05,
+      "loss": 0.6101,
+      "step": 14406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5180180666247749e-05,
+      "loss": 0.6297,
+      "step": 14407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.517957091831455e-05,
+      "loss": 0.617,
+      "step": 14408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5178961144062433e-05,
+      "loss": 0.6634,
+      "step": 14409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5178351343494497e-05,
+      "loss": 0.7066,
+      "step": 14410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5177741516613847e-05,
+      "loss": 0.6001,
+      "step": 14411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5177131663423569e-05,
+      "loss": 0.6386,
+      "step": 14412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5176521783926772e-05,
+      "loss": 0.6191,
+      "step": 14413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5175911878126553e-05,
+      "loss": 0.5977,
+      "step": 14414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5175301946026006e-05,
+      "loss": 0.6215,
+      "step": 14415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5174691987628239e-05,
+      "loss": 0.6067,
+      "step": 14416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5174082002936341e-05,
+      "loss": 0.6565,
+      "step": 14417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.517347199195342e-05,
+      "loss": 0.5956,
+      "step": 14418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5172861954682575e-05,
+      "loss": 0.5841,
+      "step": 14419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5172251891126899e-05,
+      "loss": 0.5997,
+      "step": 14420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5171641801289498e-05,
+      "loss": 0.6667,
+      "step": 14421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5171031685173471e-05,
+      "loss": 0.5892,
+      "step": 14422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5170421542781918e-05,
+      "loss": 0.5643,
+      "step": 14423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5169811374117939e-05,
+      "loss": 0.6641,
+      "step": 14424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5169201179184634e-05,
+      "loss": 0.6002,
+      "step": 14425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5168590957985104e-05,
+      "loss": 0.5782,
+      "step": 14426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5167980710522447e-05,
+      "loss": 0.6721,
+      "step": 14427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.516737043679977e-05,
+      "loss": 0.625,
+      "step": 14428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5166760136820169e-05,
+      "loss": 0.5895,
+      "step": 14429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.516614981058675e-05,
+      "loss": 0.5785,
+      "step": 14430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.516553945810261e-05,
+      "loss": 0.5418,
+      "step": 14431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5164929079370847e-05,
+      "loss": 0.5453,
+      "step": 14432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5164318674394567e-05,
+      "loss": 0.6476,
+      "step": 14433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5163708243176873e-05,
+      "loss": 0.683,
+      "step": 14434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5163097785720866e-05,
+      "loss": 0.6046,
+      "step": 14435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5162487302029647e-05,
+      "loss": 0.6167,
+      "step": 14436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5161876792106319e-05,
+      "loss": 0.6126,
+      "step": 14437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5161266255953979e-05,
+      "loss": 0.6016,
+      "step": 14438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5160655693575738e-05,
+      "loss": 0.6591,
+      "step": 14439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5160045104974689e-05,
+      "loss": 0.6177,
+      "step": 14440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5159434490153944e-05,
+      "loss": 0.6182,
+      "step": 14441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5158823849116602e-05,
+      "loss": 0.6515,
+      "step": 14442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5158213181865762e-05,
+      "loss": 0.6457,
+      "step": 14443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5157602488404531e-05,
+      "loss": 0.626,
+      "step": 14444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5156991768736013e-05,
+      "loss": 0.5775,
+      "step": 14445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5156381022863307e-05,
+      "loss": 0.5886,
+      "step": 14446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5155770250789523e-05,
+      "loss": 0.6036,
+      "step": 14447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5155159452517758e-05,
+      "loss": 0.5847,
+      "step": 14448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5154548628051119e-05,
+      "loss": 0.6058,
+      "step": 14449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5153937777392704e-05,
+      "loss": 0.5962,
+      "step": 14450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5153326900545628e-05,
+      "loss": 0.6157,
+      "step": 14451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5152715997512986e-05,
+      "loss": 0.5872,
+      "step": 14452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5152105068297887e-05,
+      "loss": 0.6536,
+      "step": 14453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5151494112903428e-05,
+      "loss": 0.5791,
+      "step": 14454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5150883131332726e-05,
+      "loss": 0.5667,
+      "step": 14455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5150272123588874e-05,
+      "loss": 0.5872,
+      "step": 14456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.514966108967498e-05,
+      "loss": 0.6613,
+      "step": 14457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5149050029594152e-05,
+      "loss": 0.6238,
+      "step": 14458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5148438943349493e-05,
+      "loss": 0.6131,
+      "step": 14459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5147827830944107e-05,
+      "loss": 0.6002,
+      "step": 14460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5147216692381101e-05,
+      "loss": 0.5845,
+      "step": 14461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5146605527663578e-05,
+      "loss": 0.6252,
+      "step": 14462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5145994336794644e-05,
+      "loss": 0.5998,
+      "step": 14463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.514538311977741e-05,
+      "loss": 0.5688,
+      "step": 14464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5144771876614975e-05,
+      "loss": 0.5421,
+      "step": 14465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5144160607310448e-05,
+      "loss": 0.6113,
+      "step": 14466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5143549311866932e-05,
+      "loss": 0.5903,
+      "step": 14467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5142937990287538e-05,
+      "loss": 0.5921,
+      "step": 14468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5142326642575369e-05,
+      "loss": 0.6348,
+      "step": 14469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5141715268733537e-05,
+      "loss": 0.5738,
+      "step": 14470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5141103868765137e-05,
+      "loss": 0.5569,
+      "step": 14471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5140492442673287e-05,
+      "loss": 0.6212,
+      "step": 14472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5139880990461086e-05,
+      "loss": 0.6169,
+      "step": 14473
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5139269512131649e-05,
+      "loss": 0.6024,
+      "step": 14474
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5138658007688074e-05,
+      "loss": 0.6248,
+      "step": 14475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5138046477133475e-05,
+      "loss": 0.6099,
+      "step": 14476
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5137434920470954e-05,
+      "loss": 0.5737,
+      "step": 14477
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5136823337703625e-05,
+      "loss": 0.5705,
+      "step": 14478
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5136211728834591e-05,
+      "loss": 0.6453,
+      "step": 14479
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5135600093866963e-05,
+      "loss": 0.5968,
+      "step": 14480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5134988432803845e-05,
+      "loss": 0.6391,
+      "step": 14481
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5134376745648346e-05,
+      "loss": 0.6094,
+      "step": 14482
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5133765032403576e-05,
+      "loss": 0.5538,
+      "step": 14483
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5133153293072642e-05,
+      "loss": 0.6408,
+      "step": 14484
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5132541527658654e-05,
+      "loss": 0.5938,
+      "step": 14485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.513192973616472e-05,
+      "loss": 0.6092,
+      "step": 14486
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5131317918593947e-05,
+      "loss": 0.5877,
+      "step": 14487
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5130706074949443e-05,
+      "loss": 0.6369,
+      "step": 14488
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5130094205234323e-05,
+      "loss": 0.5834,
+      "step": 14489
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5129482309451689e-05,
+      "loss": 0.6183,
+      "step": 14490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5128870387604653e-05,
+      "loss": 0.5745,
+      "step": 14491
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5128258439696323e-05,
+      "loss": 0.6888,
+      "step": 14492
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5127646465729817e-05,
+      "loss": 0.6298,
+      "step": 14493
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512703446570823e-05,
+      "loss": 0.5411,
+      "step": 14494
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5126422439634682e-05,
+      "loss": 0.6262,
+      "step": 14495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512581038751228e-05,
+      "loss": 0.5152,
+      "step": 14496
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5125198309344135e-05,
+      "loss": 0.578,
+      "step": 14497
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5124586205133356e-05,
+      "loss": 0.6134,
+      "step": 14498
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5123974074883054e-05,
+      "loss": 0.5357,
+      "step": 14499
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5123361918596338e-05,
+      "loss": 0.5799,
+      "step": 14500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5122749736276321e-05,
+      "loss": 0.5784,
+      "step": 14501
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512213752792611e-05,
+      "loss": 0.6082,
+      "step": 14502
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5121525293548822e-05,
+      "loss": 0.6055,
+      "step": 14503
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5120913033147562e-05,
+      "loss": 0.6068,
+      "step": 14504
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.512030074672544e-05,
+      "loss": 0.6085,
+      "step": 14505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5119688434285573e-05,
+      "loss": 0.6062,
+      "step": 14506
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5119076095831071e-05,
+      "loss": 0.634,
+      "step": 14507
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5118463731365047e-05,
+      "loss": 0.6758,
+      "step": 14508
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5117851340890606e-05,
+      "loss": 0.5834,
+      "step": 14509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5117238924410864e-05,
+      "loss": 0.6562,
+      "step": 14510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5116626481928931e-05,
+      "loss": 0.6048,
+      "step": 14511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5116014013447921e-05,
+      "loss": 0.6064,
+      "step": 14512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5115401518970946e-05,
+      "loss": 0.6278,
+      "step": 14513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5114788998501117e-05,
+      "loss": 0.6182,
+      "step": 14514
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.511417645204155e-05,
+      "loss": 0.6436,
+      "step": 14515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.511356387959535e-05,
+      "loss": 0.6356,
+      "step": 14516
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5112951281165638e-05,
+      "loss": 0.6204,
+      "step": 14517
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.511233865675552e-05,
+      "loss": 0.6189,
+      "step": 14518
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5111726006368115e-05,
+      "loss": 0.6216,
+      "step": 14519
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5111113330006532e-05,
+      "loss": 0.5753,
+      "step": 14520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5110500627673884e-05,
+      "loss": 0.6571,
+      "step": 14521
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5109887899373288e-05,
+      "loss": 0.5538,
+      "step": 14522
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5109275145107853e-05,
+      "loss": 0.6218,
+      "step": 14523
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5108662364880697e-05,
+      "loss": 0.6326,
+      "step": 14524
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5108049558694928e-05,
+      "loss": 0.6295,
+      "step": 14525
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5107436726553668e-05,
+      "loss": 0.5359,
+      "step": 14526
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5106823868460025e-05,
+      "loss": 0.6736,
+      "step": 14527
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5106210984417114e-05,
+      "loss": 0.5936,
+      "step": 14528
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5105598074428048e-05,
+      "loss": 0.6123,
+      "step": 14529
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5104985138495942e-05,
+      "loss": 0.5928,
+      "step": 14530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5104372176623914e-05,
+      "loss": 0.563,
+      "step": 14531
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5103759188815074e-05,
+      "loss": 0.6,
+      "step": 14532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.510314617507254e-05,
+      "loss": 0.5555,
+      "step": 14533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5102533135399426e-05,
+      "loss": 0.639,
+      "step": 14534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5101920069798847e-05,
+      "loss": 0.525,
+      "step": 14535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5101306978273915e-05,
+      "loss": 0.5774,
+      "step": 14536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5100693860827753e-05,
+      "loss": 0.6301,
+      "step": 14537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5100080717463468e-05,
+      "loss": 0.5682,
+      "step": 14538
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5099467548184182e-05,
+      "loss": 0.5948,
+      "step": 14539
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5098854352993005e-05,
+      "loss": 0.6484,
+      "step": 14540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5098241131893056e-05,
+      "loss": 0.6207,
+      "step": 14541
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5097627884887454e-05,
+      "loss": 0.5713,
+      "step": 14542
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5097014611979308e-05,
+      "loss": 0.5401,
+      "step": 14543
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5096401313171741e-05,
+      "loss": 0.5926,
+      "step": 14544
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5095787988467864e-05,
+      "loss": 0.6143,
+      "step": 14545
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5095174637870798e-05,
+      "loss": 0.5593,
+      "step": 14546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5094561261383657e-05,
+      "loss": 0.584,
+      "step": 14547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.509394785900956e-05,
+      "loss": 0.6229,
+      "step": 14548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5093334430751619e-05,
+      "loss": 0.6341,
+      "step": 14549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5092720976612956e-05,
+      "loss": 0.5816,
+      "step": 14550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5092107496596684e-05,
+      "loss": 0.6268,
+      "step": 14551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5091493990705924e-05,
+      "loss": 0.5954,
+      "step": 14552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5090880458943793e-05,
+      "loss": 0.5926,
+      "step": 14553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5090266901313404e-05,
+      "loss": 0.5884,
+      "step": 14554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5089653317817879e-05,
+      "loss": 0.6494,
+      "step": 14555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5089039708460338e-05,
+      "loss": 0.612,
+      "step": 14556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5088426073243892e-05,
+      "loss": 0.6226,
+      "step": 14557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5087812412171668e-05,
+      "loss": 0.5965,
+      "step": 14558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5087198725246772e-05,
+      "loss": 0.605,
+      "step": 14559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5086585012472334e-05,
+      "loss": 0.5929,
+      "step": 14560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.508597127385147e-05,
+      "loss": 0.6325,
+      "step": 14561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5085357509387293e-05,
+      "loss": 0.5694,
+      "step": 14562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5084743719082924e-05,
+      "loss": 0.6288,
+      "step": 14563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5084129902941485e-05,
+      "loss": 0.5655,
+      "step": 14564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5083516060966092e-05,
+      "loss": 0.5724,
+      "step": 14565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.508290219315987e-05,
+      "loss": 0.5958,
+      "step": 14566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5082288299525929e-05,
+      "loss": 0.6338,
+      "step": 14567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5081674380067392e-05,
+      "loss": 0.6458,
+      "step": 14568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5081060434787383e-05,
+      "loss": 0.6159,
+      "step": 14569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5080446463689013e-05,
+      "loss": 0.5923,
+      "step": 14570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5079832466775409e-05,
+      "loss": 0.5716,
+      "step": 14571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5079218444049688e-05,
+      "loss": 0.6224,
+      "step": 14572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5078604395514971e-05,
+      "loss": 0.6161,
+      "step": 14573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507799032117438e-05,
+      "loss": 0.5884,
+      "step": 14574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.507737622103103e-05,
+      "loss": 0.6253,
+      "step": 14575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5076762095088043e-05,
+      "loss": 0.6542,
+      "step": 14576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5076147943348545e-05,
+      "loss": 0.5951,
+      "step": 14577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5075533765815649e-05,
+      "loss": 0.6932,
+      "step": 14578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5074919562492483e-05,
+      "loss": 0.5665,
+      "step": 14579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5074305333382162e-05,
+      "loss": 0.5878,
+      "step": 14580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5073691078487809e-05,
+      "loss": 0.5659,
+      "step": 14581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5073076797812546e-05,
+      "loss": 0.5932,
+      "step": 14582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5072462491359496e-05,
+      "loss": 0.611,
+      "step": 14583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5071848159131778e-05,
+      "loss": 0.5879,
+      "step": 14584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5071233801132517e-05,
+      "loss": 0.5918,
+      "step": 14585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5070619417364827e-05,
+      "loss": 0.6747,
+      "step": 14586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5070005007831836e-05,
+      "loss": 0.6191,
+      "step": 14587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5069390572536668e-05,
+      "loss": 0.6271,
+      "step": 14588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5068776111482438e-05,
+      "loss": 0.6238,
+      "step": 14589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5068161624672271e-05,
+      "loss": 0.591,
+      "step": 14590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5067547112109296e-05,
+      "loss": 0.5854,
+      "step": 14591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5066932573796624e-05,
+      "loss": 0.6412,
+      "step": 14592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5066318009737388e-05,
+      "loss": 0.6564,
+      "step": 14593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5065703419934705e-05,
+      "loss": 0.623,
+      "step": 14594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5065088804391699e-05,
+      "loss": 0.6308,
+      "step": 14595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5064474163111493e-05,
+      "loss": 0.5587,
+      "step": 14596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.506385949609721e-05,
+      "loss": 0.6391,
+      "step": 14597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5063244803351975e-05,
+      "loss": 0.6121,
+      "step": 14598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5062630084878911e-05,
+      "loss": 0.614,
+      "step": 14599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5062015340681139e-05,
+      "loss": 0.6068,
+      "step": 14600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5061400570761786e-05,
+      "loss": 0.6053,
+      "step": 14601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5060785775123976e-05,
+      "loss": 0.6404,
+      "step": 14602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5060170953770828e-05,
+      "loss": 0.6778,
+      "step": 14603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5059556106705472e-05,
+      "loss": 0.6064,
+      "step": 14604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5058941233931028e-05,
+      "loss": 0.5847,
+      "step": 14605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5058326335450623e-05,
+      "loss": 0.6314,
+      "step": 14606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5057711411267381e-05,
+      "loss": 0.6002,
+      "step": 14607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5057096461384426e-05,
+      "loss": 0.5495,
+      "step": 14608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5056481485804881e-05,
+      "loss": 0.6331,
+      "step": 14609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5055866484531873e-05,
+      "loss": 0.6458,
+      "step": 14610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5055251457568529e-05,
+      "loss": 0.6229,
+      "step": 14611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5054636404917973e-05,
+      "loss": 0.6157,
+      "step": 14612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5054021326583323e-05,
+      "loss": 0.5941,
+      "step": 14613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5053406222567714e-05,
+      "loss": 0.5674,
+      "step": 14614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505279109287427e-05,
+      "loss": 0.5846,
+      "step": 14615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.505217593750611e-05,
+      "loss": 0.6066,
+      "step": 14616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5051560756466367e-05,
+      "loss": 0.6366,
+      "step": 14617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5050945549758167e-05,
+      "loss": 0.5443,
+      "step": 14618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5050330317384629e-05,
+      "loss": 0.615,
+      "step": 14619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5049715059348887e-05,
+      "loss": 0.6133,
+      "step": 14620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5049099775654062e-05,
+      "loss": 0.5783,
+      "step": 14621
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5048484466303282e-05,
+      "loss": 0.5838,
+      "step": 14622
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5047869131299675e-05,
+      "loss": 0.6037,
+      "step": 14623
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5047253770646366e-05,
+      "loss": 0.5729,
+      "step": 14624
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5046638384346483e-05,
+      "loss": 0.6187,
+      "step": 14625
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5046022972403154e-05,
+      "loss": 0.5795,
+      "step": 14626
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5045407534819502e-05,
+      "loss": 0.5928,
+      "step": 14627
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5044792071598657e-05,
+      "loss": 0.6201,
+      "step": 14628
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5044176582743748e-05,
+      "loss": 0.6588,
+      "step": 14629
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5043561068257897e-05,
+      "loss": 0.5291,
+      "step": 14630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5042945528144239e-05,
+      "loss": 0.6292,
+      "step": 14631
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5042329962405895e-05,
+      "loss": 0.5557,
+      "step": 14632
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5041714371045996e-05,
+      "loss": 0.6072,
+      "step": 14633
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5041098754067671e-05,
+      "loss": 0.6234,
+      "step": 14634
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5040483111474045e-05,
+      "loss": 0.6586,
+      "step": 14635
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503986744326825e-05,
+      "loss": 0.5004,
+      "step": 14636
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5039251749453411e-05,
+      "loss": 0.6,
+      "step": 14637
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503863603003266e-05,
+      "loss": 0.5682,
+      "step": 14638
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5038020285009123e-05,
+      "loss": 0.5596,
+      "step": 14639
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503740451438593e-05,
+      "loss": 0.653,
+      "step": 14640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5036788718166207e-05,
+      "loss": 0.6362,
+      "step": 14641
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5036172896353089e-05,
+      "loss": 0.6331,
+      "step": 14642
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5035557048949696e-05,
+      "loss": 0.6618,
+      "step": 14643
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503494117595917e-05,
+      "loss": 0.5558,
+      "step": 14644
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503432527738463e-05,
+      "loss": 0.6365,
+      "step": 14645
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5033709353229208e-05,
+      "loss": 0.6287,
+      "step": 14646
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5033093403496034e-05,
+      "loss": 0.5777,
+      "step": 14647
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503247742818824e-05,
+      "loss": 0.6359,
+      "step": 14648
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5031861427308956e-05,
+      "loss": 0.5317,
+      "step": 14649
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5031245400861308e-05,
+      "loss": 0.5596,
+      "step": 14650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.503062934884843e-05,
+      "loss": 0.6106,
+      "step": 14651
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5030013271273447e-05,
+      "loss": 0.5875,
+      "step": 14652
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5029397168139498e-05,
+      "loss": 0.6441,
+      "step": 14653
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.502878103944971e-05,
+      "loss": 0.5643,
+      "step": 14654
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5028164885207209e-05,
+      "loss": 0.6498,
+      "step": 14655
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5027548705415131e-05,
+      "loss": 0.619,
+      "step": 14656
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5026932500076605e-05,
+      "loss": 0.5248,
+      "step": 14657
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5026316269194764e-05,
+      "loss": 0.5906,
+      "step": 14658
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5025700012772738e-05,
+      "loss": 0.6479,
+      "step": 14659
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5025083730813657e-05,
+      "loss": 0.6546,
+      "step": 14660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5024467423320656e-05,
+      "loss": 0.6202,
+      "step": 14661
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5023851090296861e-05,
+      "loss": 0.645,
+      "step": 14662
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.502323473174541e-05,
+      "loss": 0.6224,
+      "step": 14663
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5022618347669434e-05,
+      "loss": 0.5655,
+      "step": 14664
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5022001938072061e-05,
+      "loss": 0.6549,
+      "step": 14665
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5021385502956425e-05,
+      "loss": 0.6191,
+      "step": 14666
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.502076904232566e-05,
+      "loss": 0.5991,
+      "step": 14667
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5020152556182895e-05,
+      "loss": 0.5302,
+      "step": 14668
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5019536044531267e-05,
+      "loss": 0.6112,
+      "step": 14669
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5018919507373906e-05,
+      "loss": 0.5796,
+      "step": 14670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5018302944713942e-05,
+      "loss": 0.5969,
+      "step": 14671
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5017686356554515e-05,
+      "loss": 0.5868,
+      "step": 14672
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5017069742898751e-05,
+      "loss": 0.5732,
+      "step": 14673
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5016453103749788e-05,
+      "loss": 0.5797,
+      "step": 14674
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5015836439110758e-05,
+      "loss": 0.5725,
+      "step": 14675
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5015219748984793e-05,
+      "loss": 0.6499,
+      "step": 14676
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5014603033375031e-05,
+      "loss": 0.6132,
+      "step": 14677
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5013986292284598e-05,
+      "loss": 0.6092,
+      "step": 14678
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5013369525716633e-05,
+      "loss": 0.6097,
+      "step": 14679
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5012752733674272e-05,
+      "loss": 0.5816,
+      "step": 14680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5012135916160642e-05,
+      "loss": 0.5775,
+      "step": 14681
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5011519073178883e-05,
+      "loss": 0.6402,
+      "step": 14682
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5010902204732132e-05,
+      "loss": 0.5811,
+      "step": 14683
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5010285310823519e-05,
+      "loss": 0.6493,
+      "step": 14684
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5009668391456174e-05,
+      "loss": 0.5902,
+      "step": 14685
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500905144663324e-05,
+      "loss": 0.6253,
+      "step": 14686
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5008434476357846e-05,
+      "loss": 0.6511,
+      "step": 14687
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500781748063313e-05,
+      "loss": 0.5728,
+      "step": 14688
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500720045946223e-05,
+      "loss": 0.6377,
+      "step": 14689
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5006583412848276e-05,
+      "loss": 0.5664,
+      "step": 14690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005966340794405e-05,
+      "loss": 0.6372,
+      "step": 14691
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5005349243303755e-05,
+      "loss": 0.5522,
+      "step": 14692
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5004732120379455e-05,
+      "loss": 0.6582,
+      "step": 14693
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500411497202465e-05,
+      "loss": 0.6306,
+      "step": 14694
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.500349779824247e-05,
+      "loss": 0.6392,
+      "step": 14695
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002880599036051e-05,
+      "loss": 0.6544,
+      "step": 14696
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5002263374408531e-05,
+      "loss": 0.6627,
+      "step": 14697
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001646124363047e-05,
+      "loss": 0.6697,
+      "step": 14698
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5001028848902736e-05,
+      "loss": 0.6192,
+      "step": 14699
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000411548030728e-05,
+      "loss": 0.612,
+      "step": 14700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4999794221750168e-05,
+      "loss": 0.574,
+      "step": 14701
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.499917687006419e-05,
+      "loss": 0.6446,
+      "step": 14702
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4998559492975927e-05,
+      "loss": 0.6083,
+      "step": 14703
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4997942090488522e-05,
+      "loss": 0.6115,
+      "step": 14704
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4997324662605108e-05,
+      "loss": 0.6554,
+      "step": 14705
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4996707209328825e-05,
+      "loss": 0.5874,
+      "step": 14706
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4996089730662808e-05,
+      "loss": 0.6941,
+      "step": 14707
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4995472226610196e-05,
+      "loss": 0.5834,
+      "step": 14708
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4994854697174127e-05,
+      "loss": 0.581,
+      "step": 14709
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.499423714235774e-05,
+      "loss": 0.5896,
+      "step": 14710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4993619562164168e-05,
+      "loss": 0.5864,
+      "step": 14711
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4993001956596555e-05,
+      "loss": 0.6221,
+      "step": 14712
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4992384325658037e-05,
+      "loss": 0.606,
+      "step": 14713
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.499176666935175e-05,
+      "loss": 0.6417,
+      "step": 14714
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4991148987680838e-05,
+      "loss": 0.6098,
+      "step": 14715
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4990531280648432e-05,
+      "loss": 0.6069,
+      "step": 14716
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4989913548257677e-05,
+      "loss": 0.6083,
+      "step": 14717
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4989295790511713e-05,
+      "loss": 0.5839,
+      "step": 14718
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4988678007413674e-05,
+      "loss": 0.5959,
+      "step": 14719
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4988060198966698e-05,
+      "loss": 0.6358,
+      "step": 14720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4987442365173932e-05,
+      "loss": 0.6244,
+      "step": 14721
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4986824506038507e-05,
+      "loss": 0.5751,
+      "step": 14722
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4986206621563568e-05,
+      "loss": 0.6043,
+      "step": 14723
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4985588711752255e-05,
+      "loss": 0.5654,
+      "step": 14724
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4984970776607702e-05,
+      "loss": 0.6079,
+      "step": 14725
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4984352816133054e-05,
+      "loss": 0.6028,
+      "step": 14726
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4983734830331446e-05,
+      "loss": 0.5779,
+      "step": 14727
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4983116819206025e-05,
+      "loss": 0.5889,
+      "step": 14728
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4982498782759932e-05,
+      "loss": 0.6212,
+      "step": 14729
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4981880720996298e-05,
+      "loss": 0.5972,
+      "step": 14730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4981262633918271e-05,
+      "loss": 0.5656,
+      "step": 14731
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.498064452152899e-05,
+      "loss": 0.6121,
+      "step": 14732
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4980026383831593e-05,
+      "loss": 0.5801,
+      "step": 14733
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4979408220829226e-05,
+      "loss": 0.6833,
+      "step": 14734
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4978790032525026e-05,
+      "loss": 0.5952,
+      "step": 14735
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4978171818922135e-05,
+      "loss": 0.7245,
+      "step": 14736
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4977553580023698e-05,
+      "loss": 0.6243,
+      "step": 14737
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4976935315832853e-05,
+      "loss": 0.6008,
+      "step": 14738
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4976317026352742e-05,
+      "loss": 0.6294,
+      "step": 14739
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4975698711586506e-05,
+      "loss": 0.6426,
+      "step": 14740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4975080371537286e-05,
+      "loss": 0.6401,
+      "step": 14741
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4974462006208228e-05,
+      "loss": 0.6756,
+      "step": 14742
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.497384361560247e-05,
+      "loss": 0.6094,
+      "step": 14743
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4973225199723158e-05,
+      "loss": 0.5835,
+      "step": 14744
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4972606758573433e-05,
+      "loss": 0.6494,
+      "step": 14745
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4971988292156435e-05,
+      "loss": 0.6027,
+      "step": 14746
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4971369800475306e-05,
+      "loss": 0.5989,
+      "step": 14747
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4970751283533197e-05,
+      "loss": 0.5981,
+      "step": 14748
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4970132741333239e-05,
+      "loss": 0.595,
+      "step": 14749
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4969514173878586e-05,
+      "loss": 0.5877,
+      "step": 14750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4968895581172371e-05,
+      "loss": 0.5527,
+      "step": 14751
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4968276963217744e-05,
+      "loss": 0.5849,
+      "step": 14752
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.496765832001785e-05,
+      "loss": 0.6595,
+      "step": 14753
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4967039651575825e-05,
+      "loss": 0.6113,
+      "step": 14754
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4966420957894818e-05,
+      "loss": 0.6873,
+      "step": 14755
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4965802238977976e-05,
+      "loss": 0.5693,
+      "step": 14756
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4965183494828432e-05,
+      "loss": 0.6011,
+      "step": 14757
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4964564725449342e-05,
+      "loss": 0.6058,
+      "step": 14758
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.496394593084384e-05,
+      "loss": 0.6168,
+      "step": 14759
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4963327111015077e-05,
+      "loss": 0.5622,
+      "step": 14760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4962708265966197e-05,
+      "loss": 0.5695,
+      "step": 14761
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.496208939570034e-05,
+      "loss": 0.5944,
+      "step": 14762
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4961470500220655e-05,
+      "loss": 0.562,
+      "step": 14763
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4960851579530287e-05,
+      "loss": 0.5536,
+      "step": 14764
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4960232633632377e-05,
+      "loss": 0.6511,
+      "step": 14765
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4959613662530073e-05,
+      "loss": 0.6166,
+      "step": 14766
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4958994666226518e-05,
+      "loss": 0.5567,
+      "step": 14767
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4958375644724861e-05,
+      "loss": 0.6594,
+      "step": 14768
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4957756598028247e-05,
+      "loss": 0.5736,
+      "step": 14769
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4957137526139815e-05,
+      "loss": 0.6129,
+      "step": 14770
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4956518429062716e-05,
+      "loss": 0.6441,
+      "step": 14771
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.49558993068001e-05,
+      "loss": 0.554,
+      "step": 14772
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4955280159355103e-05,
+      "loss": 0.5936,
+      "step": 14773
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4954660986730877e-05,
+      "loss": 0.6151,
+      "step": 14774
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4954041788930571e-05,
+      "loss": 0.5805,
+      "step": 14775
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4953422565957325e-05,
+      "loss": 0.6058,
+      "step": 14776
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.495280331781429e-05,
+      "loss": 0.6186,
+      "step": 14777
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4952184044504609e-05,
+      "loss": 0.5967,
+      "step": 14778
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.495156474603143e-05,
+      "loss": 0.6134,
+      "step": 14779
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4950945422397905e-05,
+      "loss": 0.5763,
+      "step": 14780
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4950326073607171e-05,
+      "loss": 0.5751,
+      "step": 14781
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4949706699662382e-05,
+      "loss": 0.6019,
+      "step": 14782
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4949087300566685e-05,
+      "loss": 0.6034,
+      "step": 14783
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4948467876323225e-05,
+      "loss": 0.5938,
+      "step": 14784
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4947848426935152e-05,
+      "loss": 0.5763,
+      "step": 14785
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4947228952405613e-05,
+      "loss": 0.575,
+      "step": 14786
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.494660945273775e-05,
+      "loss": 0.6177,
+      "step": 14787
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4945989927934718e-05,
+      "loss": 0.6243,
+      "step": 14788
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4945370377999665e-05,
+      "loss": 0.5839,
+      "step": 14789
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4944750802935732e-05,
+      "loss": 0.5315,
+      "step": 14790
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4944131202746077e-05,
+      "loss": 0.6035,
+      "step": 14791
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.494351157743384e-05,
+      "loss": 0.644,
+      "step": 14792
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4942891927002174e-05,
+      "loss": 0.6346,
+      "step": 14793
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4942272251454229e-05,
+      "loss": 0.5331,
+      "step": 14794
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4941652550793147e-05,
+      "loss": 0.6754,
+      "step": 14795
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4941032825022084e-05,
+      "loss": 0.5877,
+      "step": 14796
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4940413074144184e-05,
+      "loss": 0.5967,
+      "step": 14797
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4939793298162599e-05,
+      "loss": 0.6165,
+      "step": 14798
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4939173497080479e-05,
+      "loss": 0.6268,
+      "step": 14799
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.493855367090097e-05,
+      "loss": 0.6443,
+      "step": 14800
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4937933819627227e-05,
+      "loss": 0.6478,
+      "step": 14801
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4937313943262393e-05,
+      "loss": 0.6025,
+      "step": 14802
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4936694041809623e-05,
+      "loss": 0.6117,
+      "step": 14803
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4936074115272063e-05,
+      "loss": 0.6227,
+      "step": 14804
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4935454163652864e-05,
+      "loss": 0.6138,
+      "step": 14805
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4934834186955179e-05,
+      "loss": 0.6355,
+      "step": 14806
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4934214185182156e-05,
+      "loss": 0.6392,
+      "step": 14807
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4933594158336945e-05,
+      "loss": 0.5995,
+      "step": 14808
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4932974106422698e-05,
+      "loss": 0.5913,
+      "step": 14809
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4932354029442564e-05,
+      "loss": 0.538,
+      "step": 14810
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4931733927399694e-05,
+      "loss": 0.566,
+      "step": 14811
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4931113800297241e-05,
+      "loss": 0.5788,
+      "step": 14812
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4930493648138354e-05,
+      "loss": 0.5656,
+      "step": 14813
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4929873470926184e-05,
+      "loss": 0.6473,
+      "step": 14814
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4929253268663886e-05,
+      "loss": 0.6419,
+      "step": 14815
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4928633041354605e-05,
+      "loss": 0.578,
+      "step": 14816
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4928012789001498e-05,
+      "loss": 0.6321,
+      "step": 14817
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4927392511607714e-05,
+      "loss": 0.6025,
+      "step": 14818
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4926772209176405e-05,
+      "loss": 0.6172,
+      "step": 14819
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4926151881710727e-05,
+      "loss": 0.6345,
+      "step": 14820
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4925531529213827e-05,
+      "loss": 0.5973,
+      "step": 14821
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4924911151688858e-05,
+      "loss": 0.6291,
+      "step": 14822
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4924290749138972e-05,
+      "loss": 0.6075,
+      "step": 14823
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4923670321567326e-05,
+      "loss": 0.6549,
+      "step": 14824
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4923049868977064e-05,
+      "loss": 0.5714,
+      "step": 14825
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4922429391371349e-05,
+      "loss": 0.5693,
+      "step": 14826
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4921808888753325e-05,
+      "loss": 0.6557,
+      "step": 14827
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4921188361126149e-05,
+      "loss": 0.5909,
+      "step": 14828
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4920567808492976e-05,
+      "loss": 0.5911,
+      "step": 14829
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4919947230856956e-05,
+      "loss": 0.6329,
+      "step": 14830
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4919326628221243e-05,
+      "loss": 0.5776,
+      "step": 14831
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.491870600058899e-05,
+      "loss": 0.5856,
+      "step": 14832
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4918085347963352e-05,
+      "loss": 0.5612,
+      "step": 14833
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.491746467034748e-05,
+      "loss": 0.5768,
+      "step": 14834
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4916843967744533e-05,
+      "loss": 0.6261,
+      "step": 14835
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4916223240157658e-05,
+      "loss": 0.5989,
+      "step": 14836
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4915602487590017e-05,
+      "loss": 0.5938,
+      "step": 14837
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4914981710044758e-05,
+      "loss": 0.5809,
+      "step": 14838
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4914360907525036e-05,
+      "loss": 0.6165,
+      "step": 14839
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4913740080034012e-05,
+      "loss": 0.6363,
+      "step": 14840
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4913119227574835e-05,
+      "loss": 0.6232,
+      "step": 14841
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4912498350150655e-05,
+      "loss": 0.592,
+      "step": 14842
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4911877447764637e-05,
+      "loss": 0.5922,
+      "step": 14843
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4911256520419928e-05,
+      "loss": 0.5825,
+      "step": 14844
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4910635568119689e-05,
+      "loss": 0.625,
+      "step": 14845
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4910014590867073e-05,
+      "loss": 0.587,
+      "step": 14846
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.490939358866523e-05,
+      "loss": 0.5794,
+      "step": 14847
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4908772561517326e-05,
+      "loss": 0.5544,
+      "step": 14848
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4908151509426507e-05,
+      "loss": 0.5978,
+      "step": 14849
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4907530432395937e-05,
+      "loss": 0.5699,
+      "step": 14850
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4906909330428763e-05,
+      "loss": 0.5763,
+      "step": 14851
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4906288203528148e-05,
+      "loss": 0.6642,
+      "step": 14852
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4905667051697243e-05,
+      "loss": 0.6756,
+      "step": 14853
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.490504587493921e-05,
+      "loss": 0.6826,
+      "step": 14854
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.49044246732572e-05,
+      "loss": 0.6342,
+      "step": 14855
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4903803446654377e-05,
+      "loss": 0.5623,
+      "step": 14856
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4903182195133888e-05,
+      "loss": 0.6209,
+      "step": 14857
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4902560918698895e-05,
+      "loss": 0.5933,
+      "step": 14858
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4901939617352554e-05,
+      "loss": 0.6607,
+      "step": 14859
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4901318291098026e-05,
+      "loss": 0.6447,
+      "step": 14860
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4900696939938459e-05,
+      "loss": 0.6582,
+      "step": 14861
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4900075563877019e-05,
+      "loss": 0.5754,
+      "step": 14862
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4899454162916858e-05,
+      "loss": 0.5651,
+      "step": 14863
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4898832737061138e-05,
+      "loss": 0.5721,
+      "step": 14864
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.489821128631301e-05,
+      "loss": 0.6124,
+      "step": 14865
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.489758981067564e-05,
+      "loss": 0.545,
+      "step": 14866
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.489696831015218e-05,
+      "loss": 0.5983,
+      "step": 14867
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4896346784745789e-05,
+      "loss": 0.61,
+      "step": 14868
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.489572523445963e-05,
+      "loss": 0.6141,
+      "step": 14869
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4895103659296855e-05,
+      "loss": 0.625,
+      "step": 14870
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4894482059260622e-05,
+      "loss": 0.5645,
+      "step": 14871
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4893860434354096e-05,
+      "loss": 0.6242,
+      "step": 14872
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4893238784580431e-05,
+      "loss": 0.5843,
+      "step": 14873
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4892617109942785e-05,
+      "loss": 0.5661,
+      "step": 14874
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.489199541044432e-05,
+      "loss": 0.5994,
+      "step": 14875
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4891373686088194e-05,
+      "loss": 0.6376,
+      "step": 14876
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4890751936877563e-05,
+      "loss": 0.5803,
+      "step": 14877
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4890130162815593e-05,
+      "loss": 0.6086,
+      "step": 14878
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4889508363905436e-05,
+      "loss": 0.5665,
+      "step": 14879
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4888886540150258e-05,
+      "loss": 0.686,
+      "step": 14880
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4888264691553215e-05,
+      "loss": 0.5262,
+      "step": 14881
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4887642818117466e-05,
+      "loss": 0.6277,
+      "step": 14882
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4887020919846176e-05,
+      "loss": 0.5845,
+      "step": 14883
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4886398996742497e-05,
+      "loss": 0.6026,
+      "step": 14884
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4885777048809595e-05,
+      "loss": 0.6368,
+      "step": 14885
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4885155076050631e-05,
+      "loss": 0.5423,
+      "step": 14886
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4884533078468761e-05,
+      "loss": 0.555,
+      "step": 14887
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.488391105606715e-05,
+      "loss": 0.5434,
+      "step": 14888
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4883289008848953e-05,
+      "loss": 0.6056,
+      "step": 14889
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4882666936817338e-05,
+      "loss": 0.5443,
+      "step": 14890
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4882044839975463e-05,
+      "loss": 0.6234,
+      "step": 14891
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4881422718326484e-05,
+      "loss": 0.5906,
+      "step": 14892
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4880800571873567e-05,
+      "loss": 0.6167,
+      "step": 14893
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4880178400619877e-05,
+      "loss": 0.5985,
+      "step": 14894
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4879556204568569e-05,
+      "loss": 0.5493,
+      "step": 14895
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4878933983722807e-05,
+      "loss": 0.5641,
+      "step": 14896
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4878311738085753e-05,
+      "loss": 0.6663,
+      "step": 14897
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4877689467660568e-05,
+      "loss": 0.642,
+      "step": 14898
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4877067172450414e-05,
+      "loss": 0.6101,
+      "step": 14899
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4876444852458457e-05,
+      "loss": 0.6201,
+      "step": 14900
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4875822507687851e-05,
+      "loss": 0.6457,
+      "step": 14901
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4875200138141766e-05,
+      "loss": 0.6011,
+      "step": 14902
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4874577743823357e-05,
+      "loss": 0.5872,
+      "step": 14903
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4873955324735793e-05,
+      "loss": 0.5626,
+      "step": 14904
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4873332880882237e-05,
+      "loss": 0.6569,
+      "step": 14905
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4872710412265848e-05,
+      "loss": 0.5916,
+      "step": 14906
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.487208791888979e-05,
+      "loss": 0.6237,
+      "step": 14907
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4871465400757226e-05,
+      "loss": 0.628,
+      "step": 14908
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4870842857871316e-05,
+      "loss": 0.6052,
+      "step": 14909
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4870220290235233e-05,
+      "loss": 0.6001,
+      "step": 14910
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4869597697852132e-05,
+      "loss": 0.6472,
+      "step": 14911
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4868975080725178e-05,
+      "loss": 0.5952,
+      "step": 14912
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.486835243885754e-05,
+      "loss": 0.5723,
+      "step": 14913
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.4867729772252373e-05,
+      "loss": 0.5996,
+      "step": 14914
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4867107080912849e-05,
+      "loss": 0.5429,
+      "step": 14915
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4866484364842124e-05,
+      "loss": 0.6046,
+      "step": 14916
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.486586162404337e-05,
+      "loss": 0.5805,
+      "step": 14917
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4865238858519746e-05,
+      "loss": 0.6447,
+      "step": 14918
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.486461606827442e-05,
+      "loss": 0.6227,
+      "step": 14919
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4863993253310556e-05,
+      "loss": 0.5598,
+      "step": 14920
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4863370413631317e-05,
+      "loss": 0.5816,
+      "step": 14921
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.486274754923987e-05,
+      "loss": 0.5914,
+      "step": 14922
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4862124660139376e-05,
+      "loss": 0.5753,
+      "step": 14923
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4861501746333003e-05,
+      "loss": 0.6316,
+      "step": 14924
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4860878807823916e-05,
+      "loss": 0.6393,
+      "step": 14925
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.486025584461528e-05,
+      "loss": 0.5988,
+      "step": 14926
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.485963285671026e-05,
+      "loss": 0.5927,
+      "step": 14927
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4859009844112024e-05,
+      "loss": 0.6138,
+      "step": 14928
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4858386806823734e-05,
+      "loss": 0.6108,
+      "step": 14929
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4857763744848562e-05,
+      "loss": 0.6727,
+      "step": 14930
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4857140658189664e-05,
+      "loss": 0.6184,
+      "step": 14931
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4856517546850216e-05,
+      "loss": 0.5771,
+      "step": 14932
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4855894410833378e-05,
+      "loss": 0.5877,
+      "step": 14933
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.485527125014232e-05,
+      "loss": 0.6158,
+      "step": 14934
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4854648064780205e-05,
+      "loss": 0.585,
+      "step": 14935
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4854024854750201e-05,
+      "loss": 0.6206,
+      "step": 14936
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4853401620055478e-05,
+      "loss": 0.6038,
+      "step": 14937
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4852778360699198e-05,
+      "loss": 0.5378,
+      "step": 14938
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4852155076684534e-05,
+      "loss": 0.5886,
+      "step": 14939
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4851531768014643e-05,
+      "loss": 0.6023,
+      "step": 14940
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4850908434692701e-05,
+      "loss": 0.5638,
+      "step": 14941
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4850285076721876e-05,
+      "loss": 0.6231,
+      "step": 14942
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4849661694105326e-05,
+      "loss": 0.6234,
+      "step": 14943
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4849038286846227e-05,
+      "loss": 0.6265,
+      "step": 14944
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4848414854947745e-05,
+      "loss": 0.6072,
+      "step": 14945
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4847791398413045e-05,
+      "loss": 0.6105,
+      "step": 14946
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.48471679172453e-05,
+      "loss": 0.6286,
+      "step": 14947
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4846544411447675e-05,
+      "loss": 0.588,
+      "step": 14948
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4845920881023337e-05,
+      "loss": 0.6451,
+      "step": 14949
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4845297325975454e-05,
+      "loss": 0.6582,
+      "step": 14950
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.48446737463072e-05,
+      "loss": 0.6558,
+      "step": 14951
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4844050142021737e-05,
+      "loss": 0.6291,
+      "step": 14952
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4843426513122241e-05,
+      "loss": 0.5801,
+      "step": 14953
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4842802859611871e-05,
+      "loss": 0.6093,
+      "step": 14954
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4842179181493803e-05,
+      "loss": 0.6546,
+      "step": 14955
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4841555478771208e-05,
+      "loss": 0.5474,
+      "step": 14956
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4840931751447248e-05,
+      "loss": 0.5739,
+      "step": 14957
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4840307999525099e-05,
+      "loss": 0.5822,
+      "step": 14958
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4839684223007924e-05,
+      "loss": 0.6219,
+      "step": 14959
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4839060421898899e-05,
+      "loss": 0.5609,
+      "step": 14960
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.483843659620119e-05,
+      "loss": 0.5467,
+      "step": 14961
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4837812745917966e-05,
+      "loss": 0.6289,
+      "step": 14962
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4837188871052399e-05,
+      "loss": 0.6279,
+      "step": 14963
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4836564971607661e-05,
+      "loss": 0.6462,
+      "step": 14964
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4835941047586918e-05,
+      "loss": 0.6533,
+      "step": 14965
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4835317098993342e-05,
+      "loss": 0.571,
+      "step": 14966
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4834693125830107e-05,
+      "loss": 0.6198,
+      "step": 14967
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4834069128100375e-05,
+      "loss": 0.6113,
+      "step": 14968
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4833445105807326e-05,
+      "loss": 0.605,
+      "step": 14969
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4832821058954125e-05,
+      "loss": 0.6465,
+      "step": 14970
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4832196987543945e-05,
+      "loss": 0.6083,
+      "step": 14971
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4831572891579957e-05,
+      "loss": 0.605,
+      "step": 14972
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4830948771065332e-05,
+      "loss": 0.6099,
+      "step": 14973
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.483032462600324e-05,
+      "loss": 0.6115,
+      "step": 14974
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4829700456396859e-05,
+      "loss": 0.6037,
+      "step": 14975
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4829076262249352e-05,
+      "loss": 0.6106,
+      "step": 14976
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4828452043563895e-05,
+      "loss": 0.5953,
+      "step": 14977
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4827827800343659e-05,
+      "loss": 0.548,
+      "step": 14978
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4827203532591816e-05,
+      "loss": 0.5581,
+      "step": 14979
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4826579240311538e-05,
+      "loss": 0.6856,
+      "step": 14980
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4825954923505998e-05,
+      "loss": 0.609,
+      "step": 14981
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4825330582178368e-05,
+      "loss": 0.6382,
+      "step": 14982
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4824706216331819e-05,
+      "loss": 0.5619,
+      "step": 14983
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4824081825969525e-05,
+      "loss": 0.5383,
+      "step": 14984
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4823457411094658e-05,
+      "loss": 0.6074,
+      "step": 14985
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4822832971710392e-05,
+      "loss": 0.6304,
+      "step": 14986
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4822208507819898e-05,
+      "loss": 0.5611,
+      "step": 14987
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4821584019426353e-05,
+      "loss": 0.6013,
+      "step": 14988
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4820959506532923e-05,
+      "loss": 0.641,
+      "step": 14989
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4820334969142788e-05,
+      "loss": 0.5922,
+      "step": 14990
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4819710407259122e-05,
+      "loss": 0.6151,
+      "step": 14991
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4819085820885093e-05,
+      "loss": 0.6159,
+      "step": 14992
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4818461210023878e-05,
+      "loss": 0.6033,
+      "step": 14993
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4817836574678652e-05,
+      "loss": 0.5732,
+      "step": 14994
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4817211914852588e-05,
+      "loss": 0.5634,
+      "step": 14995
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4816587230548857e-05,
+      "loss": 0.5862,
+      "step": 14996
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4815962521770635e-05,
+      "loss": 0.5556,
+      "step": 14997
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.48153377885211e-05,
+      "loss": 0.6025,
+      "step": 14998
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4814713030803423e-05,
+      "loss": 0.6106,
+      "step": 14999
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4814088248620777e-05,
+      "loss": 0.5911,
+      "step": 15000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4813463441976339e-05,
+      "loss": 0.5594,
+      "step": 15001
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4812838610873286e-05,
+      "loss": 0.5972,
+      "step": 15002
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4812213755314789e-05,
+      "loss": 0.6017,
+      "step": 15003
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4811588875304023e-05,
+      "loss": 0.625,
+      "step": 15004
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4810963970844166e-05,
+      "loss": 0.5644,
+      "step": 15005
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4810339041938393e-05,
+      "loss": 0.6316,
+      "step": 15006
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.480971408858988e-05,
+      "loss": 0.6011,
+      "step": 15007
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4809089110801798e-05,
+      "loss": 0.6017,
+      "step": 15008
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4808464108577325e-05,
+      "loss": 0.6084,
+      "step": 15009
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4807839081919639e-05,
+      "loss": 0.6901,
+      "step": 15010
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4807214030831916e-05,
+      "loss": 0.5814,
+      "step": 15011
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.480658895531733e-05,
+      "loss": 0.5838,
+      "step": 15012
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4805963855379057e-05,
+      "loss": 0.5856,
+      "step": 15013
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4805338731020273e-05,
+      "loss": 0.5839,
+      "step": 15014
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.480471358224416e-05,
+      "loss": 0.629,
+      "step": 15015
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4804088409053886e-05,
+      "loss": 0.6432,
+      "step": 15016
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4803463211452633e-05,
+      "loss": 0.5917,
+      "step": 15017
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4802837989443576e-05,
+      "loss": 0.5952,
+      "step": 15018
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4802212743029895e-05,
+      "loss": 0.6123,
+      "step": 15019
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4801587472214763e-05,
+      "loss": 0.5837,
+      "step": 15020
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4800962177001362e-05,
+      "loss": 0.5543,
+      "step": 15021
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4800336857392861e-05,
+      "loss": 0.6284,
+      "step": 15022
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4799711513392443e-05,
+      "loss": 0.6342,
+      "step": 15023
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.479908614500329e-05,
+      "loss": 0.6204,
+      "step": 15024
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4798460752228569e-05,
+      "loss": 0.5126,
+      "step": 15025
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4797835335071468e-05,
+      "loss": 0.5566,
+      "step": 15026
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4797209893535159e-05,
+      "loss": 0.5922,
+      "step": 15027
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.479658442762282e-05,
+      "loss": 0.6072,
+      "step": 15028
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4795958937337634e-05,
+      "loss": 0.5749,
+      "step": 15029
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4795333422682774e-05,
+      "loss": 0.5539,
+      "step": 15030
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4794707883661418e-05,
+      "loss": 0.5705,
+      "step": 15031
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4794082320276753e-05,
+      "loss": 0.5482,
+      "step": 15032
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4793456732531948e-05,
+      "loss": 0.6781,
+      "step": 15033
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4792831120430187e-05,
+      "loss": 0.5582,
+      "step": 15034
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4792205483974644e-05,
+      "loss": 0.5541,
+      "step": 15035
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4791579823168505e-05,
+      "loss": 0.5834,
+      "step": 15036
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4790954138014946e-05,
+      "loss": 0.6188,
+      "step": 15037
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4790328428517144e-05,
+      "loss": 0.5964,
+      "step": 15038
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.478970269467828e-05,
+      "loss": 0.5986,
+      "step": 15039
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4789076936501536e-05,
+      "loss": 0.6434,
+      "step": 15040
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4788451153990088e-05,
+      "loss": 0.6398,
+      "step": 15041
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4787825347147118e-05,
+      "loss": 0.641,
+      "step": 15042
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4787199515975805e-05,
+      "loss": 0.6652,
+      "step": 15043
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4786573660479329e-05,
+      "loss": 0.6064,
+      "step": 15044
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4785947780660872e-05,
+      "loss": 0.5617,
+      "step": 15045
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.478532187652361e-05,
+      "loss": 0.6112,
+      "step": 15046
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4784695948070726e-05,
+      "loss": 0.5731,
+      "step": 15047
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4784069995305404e-05,
+      "loss": 0.6096,
+      "step": 15048
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4783444018230819e-05,
+      "loss": 0.6133,
+      "step": 15049
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4782818016850153e-05,
+      "loss": 0.6135,
+      "step": 15050
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.478219199116659e-05,
+      "loss": 0.5881,
+      "step": 15051
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4781565941183309e-05,
+      "loss": 0.6142,
+      "step": 15052
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.478093986690349e-05,
+      "loss": 0.608,
+      "step": 15053
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4780313768330317e-05,
+      "loss": 0.5811,
+      "step": 15054
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4779687645466968e-05,
+      "loss": 0.6435,
+      "step": 15055
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4779061498316627e-05,
+      "loss": 0.5807,
+      "step": 15056
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4778435326882474e-05,
+      "loss": 0.5696,
+      "step": 15057
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4777809131167693e-05,
+      "loss": 0.6257,
+      "step": 15058
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4777182911175465e-05,
+      "loss": 0.6175,
+      "step": 15059
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.477655666690897e-05,
+      "loss": 0.5933,
+      "step": 15060
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.4775930398371392e-05,
+      "loss": 0.6357,
+      "step": 15061
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4775304105565913e-05,
+      "loss": 0.5639,
+      "step": 15062
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4774677788495715e-05,
+      "loss": 0.5661,
+      "step": 15063
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4774051447163982e-05,
+      "loss": 0.6733,
+      "step": 15064
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4773425081573894e-05,
+      "loss": 0.6158,
+      "step": 15065
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4772798691728636e-05,
+      "loss": 0.5757,
+      "step": 15066
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4772172277631392e-05,
+      "loss": 0.5903,
+      "step": 15067
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4771545839285337e-05,
+      "loss": 0.6041,
+      "step": 15068
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4770919376693664e-05,
+      "loss": 0.6945,
+      "step": 15069
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4770292889859554e-05,
+      "loss": 0.5635,
+      "step": 15070
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4769666378786187e-05,
+      "loss": 0.5666,
+      "step": 15071
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4769039843476751e-05,
+      "loss": 0.5837,
+      "step": 15072
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4768413283934422e-05,
+      "loss": 0.6152,
+      "step": 15073
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4767786700162392e-05,
+      "loss": 0.619,
+      "step": 15074
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4767160092163841e-05,
+      "loss": 0.6117,
+      "step": 15075
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4766533459941952e-05,
+      "loss": 0.6155,
+      "step": 15076
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4765906803499914e-05,
+      "loss": 0.6191,
+      "step": 15077
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4765280122840907e-05,
+      "loss": 0.5851,
+      "step": 15078
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4764653417968114e-05,
+      "loss": 0.6162,
+      "step": 15079
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4764026688884722e-05,
+      "loss": 0.6199,
+      "step": 15080
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4763399935593914e-05,
+      "loss": 0.6467,
+      "step": 15081
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4762773158098877e-05,
+      "loss": 0.5829,
+      "step": 15082
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4762146356402795e-05,
+      "loss": 0.6473,
+      "step": 15083
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4761519530508852e-05,
+      "loss": 0.5703,
+      "step": 15084
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4760892680420233e-05,
+      "loss": 0.5391,
+      "step": 15085
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4760265806140129e-05,
+      "loss": 0.6373,
+      "step": 15086
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4759638907671713e-05,
+      "loss": 0.6078,
+      "step": 15087
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4759011985018183e-05,
+      "loss": 0.5763,
+      "step": 15088
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4758385038182719e-05,
+      "loss": 0.5729,
+      "step": 15089
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4757758067168506e-05,
+      "loss": 0.5086,
+      "step": 15090
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4757131071978732e-05,
+      "loss": 0.6041,
+      "step": 15091
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.475650405261658e-05,
+      "loss": 0.6371,
+      "step": 15092
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4755877009085237e-05,
+      "loss": 0.5912,
+      "step": 15093
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4755249941387892e-05,
+      "loss": 0.6394,
+      "step": 15094
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4754622849527728e-05,
+      "loss": 0.6223,
+      "step": 15095
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4753995733507935e-05,
+      "loss": 0.5643,
+      "step": 15096
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4753368593331698e-05,
+      "loss": 0.6299,
+      "step": 15097
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4752741429002202e-05,
+      "loss": 0.6097,
+      "step": 15098
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4752114240522634e-05,
+      "loss": 0.6593,
+      "step": 15099
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4751487027896183e-05,
+      "loss": 0.6129,
+      "step": 15100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4750859791126033e-05,
+      "loss": 0.6398,
+      "step": 15101
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4750232530215373e-05,
+      "loss": 0.6264,
+      "step": 15102
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4749605245167394e-05,
+      "loss": 0.598,
+      "step": 15103
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4748977935985275e-05,
+      "loss": 0.6418,
+      "step": 15104
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4748350602672212e-05,
+      "loss": 0.5986,
+      "step": 15105
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4747723245231387e-05,
+      "loss": 0.6632,
+      "step": 15106
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.474709586366599e-05,
+      "loss": 0.6588,
+      "step": 15107
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4746468457979208e-05,
+      "loss": 0.6543,
+      "step": 15108
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4745841028174231e-05,
+      "loss": 0.5638,
+      "step": 15109
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4745213574254246e-05,
+      "loss": 0.5917,
+      "step": 15110
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4744586096222439e-05,
+      "loss": 0.5567,
+      "step": 15111
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4743958594082002e-05,
+      "loss": 0.6108,
+      "step": 15112
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4743331067836124e-05,
+      "loss": 0.5143,
+      "step": 15113
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4742703517487988e-05,
+      "loss": 0.6548,
+      "step": 15114
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.474207594304079e-05,
+      "loss": 0.6577,
+      "step": 15115
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4741448344497712e-05,
+      "loss": 0.6175,
+      "step": 15116
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.474082072186195e-05,
+      "loss": 0.6124,
+      "step": 15117
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4740193075136689e-05,
+      "loss": 0.6024,
+      "step": 15118
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4739565404325116e-05,
+      "loss": 0.6003,
+      "step": 15119
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4738937709430424e-05,
+      "loss": 0.537,
+      "step": 15120
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4738309990455804e-05,
+      "loss": 0.6153,
+      "step": 15121
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.473768224740444e-05,
+      "loss": 0.6315,
+      "step": 15122
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4737054480279525e-05,
+      "loss": 0.6106,
+      "step": 15123
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4736426689084252e-05,
+      "loss": 0.5847,
+      "step": 15124
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4735798873821804e-05,
+      "loss": 0.6049,
+      "step": 15125
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.473517103449538e-05,
+      "loss": 0.6493,
+      "step": 15126
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.473454317110816e-05,
+      "loss": 0.6137,
+      "step": 15127
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4733915283663342e-05,
+      "loss": 0.603,
+      "step": 15128
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4733287372164116e-05,
+      "loss": 0.5673,
+      "step": 15129
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4732659436613666e-05,
+      "loss": 0.5453,
+      "step": 15130
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.473203147701519e-05,
+      "loss": 0.6198,
+      "step": 15131
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4731403493371876e-05,
+      "loss": 0.6396,
+      "step": 15132
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4730775485686914e-05,
+      "loss": 0.5778,
+      "step": 15133
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.47301474539635e-05,
+      "loss": 0.6291,
+      "step": 15134
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4729519398204819e-05,
+      "loss": 0.6445,
+      "step": 15135
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4728891318414065e-05,
+      "loss": 0.6443,
+      "step": 15136
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4728263214594431e-05,
+      "loss": 0.6375,
+      "step": 15137
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4727635086749103e-05,
+      "loss": 0.5972,
+      "step": 15138
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4727006934881278e-05,
+      "loss": 0.5439,
+      "step": 15139
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4726378758994149e-05,
+      "loss": 0.6696,
+      "step": 15140
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4725750559090904e-05,
+      "loss": 0.6132,
+      "step": 15141
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4725122335174735e-05,
+      "loss": 0.6375,
+      "step": 15142
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4724494087248835e-05,
+      "loss": 0.6187,
+      "step": 15143
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4723865815316399e-05,
+      "loss": 0.6184,
+      "step": 15144
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4723237519380618e-05,
+      "loss": 0.6555,
+      "step": 15145
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4722609199444682e-05,
+      "loss": 0.6193,
+      "step": 15146
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4721980855511784e-05,
+      "loss": 0.6076,
+      "step": 15147
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4721352487585124e-05,
+      "loss": 0.5837,
+      "step": 15148
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4720724095667884e-05,
+      "loss": 0.5891,
+      "step": 15149
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4720095679763264e-05,
+      "loss": 0.601,
+      "step": 15150
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4719467239874457e-05,
+      "loss": 0.6033,
+      "step": 15151
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4718838776004652e-05,
+      "loss": 0.5803,
+      "step": 15152
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.471821028815705e-05,
+      "loss": 0.6179,
+      "step": 15153
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4717581776334833e-05,
+      "loss": 0.6556,
+      "step": 15154
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4716953240541205e-05,
+      "loss": 0.5899,
+      "step": 15155
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4716324680779357e-05,
+      "loss": 0.5914,
+      "step": 15156
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.471569609705248e-05,
+      "loss": 0.5747,
+      "step": 15157
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4715067489363772e-05,
+      "loss": 0.5721,
+      "step": 15158
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4714438857716425e-05,
+      "loss": 0.6627,
+      "step": 15159
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4713810202113632e-05,
+      "loss": 0.5806,
+      "step": 15160
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4713181522558589e-05,
+      "loss": 0.558,
+      "step": 15161
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4712552819054491e-05,
+      "loss": 0.6282,
+      "step": 15162
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4711924091604532e-05,
+      "loss": 0.6656,
+      "step": 15163
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4711295340211909e-05,
+      "loss": 0.6438,
+      "step": 15164
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.471066656487981e-05,
+      "loss": 0.5565,
+      "step": 15165
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4710037765611435e-05,
+      "loss": 0.5864,
+      "step": 15166
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4709408942409984e-05,
+      "loss": 0.5499,
+      "step": 15167
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4708780095278642e-05,
+      "loss": 0.5756,
+      "step": 15168
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.470815122422061e-05,
+      "loss": 0.6129,
+      "step": 15169
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4707522329239085e-05,
+      "loss": 0.6493,
+      "step": 15170
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.470689341033726e-05,
+      "loss": 0.6421,
+      "step": 15171
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.470626446751833e-05,
+      "loss": 0.6144,
+      "step": 15172
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4705635500785492e-05,
+      "loss": 0.5962,
+      "step": 15173
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.470500651014194e-05,
+      "loss": 0.6059,
+      "step": 15174
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4704377495590876e-05,
+      "loss": 0.5961,
+      "step": 15175
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4703748457135492e-05,
+      "loss": 0.6088,
+      "step": 15176
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4703119394778982e-05,
+      "loss": 0.5541,
+      "step": 15177
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4702490308524546e-05,
+      "loss": 0.6187,
+      "step": 15178
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4701861198375378e-05,
+      "loss": 0.6156,
+      "step": 15179
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4701232064334678e-05,
+      "loss": 0.6199,
+      "step": 15180
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.470060290640564e-05,
+      "loss": 0.5921,
+      "step": 15181
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4699973724591463e-05,
+      "loss": 0.5573,
+      "step": 15182
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4699344518895345e-05,
+      "loss": 0.5635,
+      "step": 15183
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4698715289320478e-05,
+      "loss": 0.6299,
+      "step": 15184
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4698086035870062e-05,
+      "loss": 0.6345,
+      "step": 15185
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4697456758547297e-05,
+      "loss": 0.6716,
+      "step": 15186
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4696827457355377e-05,
+      "loss": 0.5785,
+      "step": 15187
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4696198132297501e-05,
+      "loss": 0.6251,
+      "step": 15188
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.469556878337687e-05,
+      "loss": 0.6301,
+      "step": 15189
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4694939410596676e-05,
+      "loss": 0.6683,
+      "step": 15190
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4694310013960123e-05,
+      "loss": 0.6283,
+      "step": 15191
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.46936805934704e-05,
+      "loss": 0.5962,
+      "step": 15192
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4693051149130717e-05,
+      "loss": 0.5739,
+      "step": 15193
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4692421680944265e-05,
+      "loss": 0.5518,
+      "step": 15194
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4691792188914245e-05,
+      "loss": 0.6237,
+      "step": 15195
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4691162673043852e-05,
+      "loss": 0.6133,
+      "step": 15196
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.469053313333629e-05,
+      "loss": 0.6322,
+      "step": 15197
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4689903569794757e-05,
+      "loss": 0.6071,
+      "step": 15198
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4689273982422447e-05,
+      "loss": 0.6138,
+      "step": 15199
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4688644371222564e-05,
+      "loss": 0.5905,
+      "step": 15200
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4688014736198306e-05,
+      "loss": 0.5745,
+      "step": 15201
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4687385077352872e-05,
+      "loss": 0.6459,
+      "step": 15202
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.468675539468946e-05,
+      "loss": 0.6287,
+      "step": 15203
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4686125688211273e-05,
+      "loss": 0.6056,
+      "step": 15204
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.468549595792151e-05,
+      "loss": 0.5838,
+      "step": 15205
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4684866203823367e-05,
+      "loss": 0.5573,
+      "step": 15206
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4684236425920046e-05,
+      "loss": 0.577,
+      "step": 15207
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.4683606624214753e-05,
+      "loss": 0.6094,
+      "step": 15208
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4682976798710679e-05,
+      "loss": 0.657,
+      "step": 15209
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4682346949411032e-05,
+      "loss": 0.6266,
+      "step": 15210
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4681717076319007e-05,
+      "loss": 0.5516,
+      "step": 15211
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4681087179437803e-05,
+      "loss": 0.5981,
+      "step": 15212
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4680457258770629e-05,
+      "loss": 0.6134,
+      "step": 15213
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4679827314320681e-05,
+      "loss": 0.5998,
+      "step": 15214
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4679197346091157e-05,
+      "loss": 0.5632,
+      "step": 15215
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4678567354085263e-05,
+      "loss": 0.6281,
+      "step": 15216
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4677937338306196e-05,
+      "loss": 0.6337,
+      "step": 15217
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.467730729875716e-05,
+      "loss": 0.5951,
+      "step": 15218
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4676677235441356e-05,
+      "loss": 0.6016,
+      "step": 15219
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4676047148361985e-05,
+      "loss": 0.6531,
+      "step": 15220
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.467541703752225e-05,
+      "loss": 0.6197,
+      "step": 15221
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.467478690292535e-05,
+      "loss": 0.6122,
+      "step": 15222
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.467415674457449e-05,
+      "loss": 0.5864,
+      "step": 15223
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4673526562472873e-05,
+      "loss": 0.5705,
+      "step": 15224
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4672896356623695e-05,
+      "loss": 0.592,
+      "step": 15225
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4672266127030162e-05,
+      "loss": 0.5845,
+      "step": 15226
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.467163587369548e-05,
+      "loss": 0.6064,
+      "step": 15227
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4671005596622845e-05,
+      "loss": 0.5913,
+      "step": 15228
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4670375295815465e-05,
+      "loss": 0.5981,
+      "step": 15229
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4669744971276538e-05,
+      "loss": 0.6089,
+      "step": 15230
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4669114623009268e-05,
+      "loss": 0.6366,
+      "step": 15231
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4668484251016864e-05,
+      "loss": 0.5338,
+      "step": 15232
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4667853855302524e-05,
+      "loss": 0.599,
+      "step": 15233
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4667223435869447e-05,
+      "loss": 0.6097,
+      "step": 15234
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4666592992720845e-05,
+      "loss": 0.6401,
+      "step": 15235
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4665962525859913e-05,
+      "loss": 0.5811,
+      "step": 15236
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4665332035289863e-05,
+      "loss": 0.5974,
+      "step": 15237
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.466470152101389e-05,
+      "loss": 0.6194,
+      "step": 15238
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4664070983035205e-05,
+      "loss": 0.5451,
+      "step": 15239
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4663440421357011e-05,
+      "loss": 0.6075,
+      "step": 15240
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.466280983598251e-05,
+      "loss": 0.6011,
+      "step": 15241
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4662179226914903e-05,
+      "loss": 0.6321,
+      "step": 15242
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4661548594157405e-05,
+      "loss": 0.5337,
+      "step": 15243
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4660917937713207e-05,
+      "loss": 0.6091,
+      "step": 15244
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4660287257585523e-05,
+      "loss": 0.5812,
+      "step": 15245
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4659656553777553e-05,
+      "loss": 0.5986,
+      "step": 15246
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4659025826292503e-05,
+      "loss": 0.5459,
+      "step": 15247
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.465839507513358e-05,
+      "loss": 0.5704,
+      "step": 15248
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4657764300303987e-05,
+      "loss": 0.5782,
+      "step": 15249
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4657133501806929e-05,
+      "loss": 0.6135,
+      "step": 15250
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4656502679645613e-05,
+      "loss": 0.6436,
+      "step": 15251
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.465587183382324e-05,
+      "loss": 0.6695,
+      "step": 15252
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4655240964343022e-05,
+      "loss": 0.5778,
+      "step": 15253
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4654610071208163e-05,
+      "loss": 0.616,
+      "step": 15254
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4653979154421862e-05,
+      "loss": 0.6269,
+      "step": 15255
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4653348213987335e-05,
+      "loss": 0.5919,
+      "step": 15256
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.465271724990778e-05,
+      "loss": 0.6183,
+      "step": 15257
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4652086262186404e-05,
+      "loss": 0.568,
+      "step": 15258
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4651455250826419e-05,
+      "loss": 0.5803,
+      "step": 15259
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4650824215831025e-05,
+      "loss": 0.6244,
+      "step": 15260
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4650193157203431e-05,
+      "loss": 0.6015,
+      "step": 15261
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4649562074946846e-05,
+      "loss": 0.6238,
+      "step": 15262
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4648930969064474e-05,
+      "loss": 0.5957,
+      "step": 15263
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4648299839559524e-05,
+      "loss": 0.6533,
+      "step": 15264
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4647668686435195e-05,
+      "loss": 0.5754,
+      "step": 15265
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4647037509694704e-05,
+      "loss": 0.568,
+      "step": 15266
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4646406309341256e-05,
+      "loss": 0.5651,
+      "step": 15267
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4645775085378056e-05,
+      "loss": 0.6497,
+      "step": 15268
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4645143837808309e-05,
+      "loss": 0.5992,
+      "step": 15269
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.464451256663523e-05,
+      "loss": 0.5709,
+      "step": 15270
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.464388127186202e-05,
+      "loss": 0.5799,
+      "step": 15271
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.464324995349189e-05,
+      "loss": 0.6399,
+      "step": 15272
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4642618611528047e-05,
+      "loss": 0.5564,
+      "step": 15273
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4641987245973698e-05,
+      "loss": 0.6321,
+      "step": 15274
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4641355856832058e-05,
+      "loss": 0.6038,
+      "step": 15275
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4640724444106323e-05,
+      "loss": 0.588,
+      "step": 15276
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.464009300779971e-05,
+      "loss": 0.5912,
+      "step": 15277
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4639461547915428e-05,
+      "loss": 0.6404,
+      "step": 15278
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.463883006445668e-05,
+      "loss": 0.5732,
+      "step": 15279
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4638198557426676e-05,
+      "loss": 0.5883,
+      "step": 15280
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4637567026828632e-05,
+      "loss": 0.5688,
+      "step": 15281
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4636935472665748e-05,
+      "loss": 0.6204,
+      "step": 15282
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.463630389494124e-05,
+      "loss": 0.5741,
+      "step": 15283
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4635672293658311e-05,
+      "loss": 0.6028,
+      "step": 15284
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4635040668820174e-05,
+      "loss": 0.5165,
+      "step": 15285
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4634409020430038e-05,
+      "loss": 0.6388,
+      "step": 15286
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4633777348491114e-05,
+      "loss": 0.5696,
+      "step": 15287
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4633145653006609e-05,
+      "loss": 0.5322,
+      "step": 15288
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4632513933979736e-05,
+      "loss": 0.6142,
+      "step": 15289
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.46318821914137e-05,
+      "loss": 0.5698,
+      "step": 15290
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4631250425311716e-05,
+      "loss": 0.567,
+      "step": 15291
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4630618635676991e-05,
+      "loss": 0.5355,
+      "step": 15292
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4629986822512738e-05,
+      "loss": 0.563,
+      "step": 15293
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4629354985822169e-05,
+      "loss": 0.6745,
+      "step": 15294
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4628723125608487e-05,
+      "loss": 0.6554,
+      "step": 15295
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4628091241874907e-05,
+      "loss": 0.6095,
+      "step": 15296
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4627459334624641e-05,
+      "loss": 0.6124,
+      "step": 15297
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4626827403860899e-05,
+      "loss": 0.6256,
+      "step": 15298
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4626195449586891e-05,
+      "loss": 0.6092,
+      "step": 15299
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.462556347180583e-05,
+      "loss": 0.5717,
+      "step": 15300
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4624931470520926e-05,
+      "loss": 0.5859,
+      "step": 15301
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4624299445735392e-05,
+      "loss": 0.6297,
+      "step": 15302
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4623667397452436e-05,
+      "loss": 0.5557,
+      "step": 15303
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4623035325675274e-05,
+      "loss": 0.5764,
+      "step": 15304
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4622403230407116e-05,
+      "loss": 0.6079,
+      "step": 15305
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.462177111165117e-05,
+      "loss": 0.6918,
+      "step": 15306
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4621138969410653e-05,
+      "loss": 0.569,
+      "step": 15307
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4620506803688779e-05,
+      "loss": 0.5641,
+      "step": 15308
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4619874614488752e-05,
+      "loss": 0.542,
+      "step": 15309
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.461924240181379e-05,
+      "loss": 0.6253,
+      "step": 15310
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4618610165667107e-05,
+      "loss": 0.607,
+      "step": 15311
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4617977906051914e-05,
+      "loss": 0.5531,
+      "step": 15312
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.461734562297142e-05,
+      "loss": 0.6464,
+      "step": 15313
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4616713316428842e-05,
+      "loss": 0.608,
+      "step": 15314
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.461608098642739e-05,
+      "loss": 0.5845,
+      "step": 15315
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.461544863297028e-05,
+      "loss": 0.6241,
+      "step": 15316
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4614816256060723e-05,
+      "loss": 0.5687,
+      "step": 15317
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4614183855701931e-05,
+      "loss": 0.5716,
+      "step": 15318
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4613551431897124e-05,
+      "loss": 0.6408,
+      "step": 15319
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4612918984649507e-05,
+      "loss": 0.5606,
+      "step": 15320
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4612286513962302e-05,
+      "loss": 0.6348,
+      "step": 15321
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4611654019838715e-05,
+      "loss": 0.6068,
+      "step": 15322
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4611021502281963e-05,
+      "loss": 0.6206,
+      "step": 15323
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4610388961295263e-05,
+      "loss": 0.5666,
+      "step": 15324
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4609756396881824e-05,
+      "loss": 0.5877,
+      "step": 15325
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4609123809044863e-05,
+      "loss": 0.6248,
+      "step": 15326
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4608491197787595e-05,
+      "loss": 0.6067,
+      "step": 15327
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4607858563113232e-05,
+      "loss": 0.6704,
+      "step": 15328
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4607225905024995e-05,
+      "loss": 0.6048,
+      "step": 15329
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4606593223526088e-05,
+      "loss": 0.632,
+      "step": 15330
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4605960518619736e-05,
+      "loss": 0.645,
+      "step": 15331
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4605327790309148e-05,
+      "loss": 0.5931,
+      "step": 15332
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4604695038597542e-05,
+      "loss": 0.6609,
+      "step": 15333
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.460406226348813e-05,
+      "loss": 0.5841,
+      "step": 15334
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4603429464984131e-05,
+      "loss": 0.6449,
+      "step": 15335
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4602796643088756e-05,
+      "loss": 0.6163,
+      "step": 15336
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4602163797805226e-05,
+      "loss": 0.5992,
+      "step": 15337
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4601530929136752e-05,
+      "loss": 0.6162,
+      "step": 15338
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4600898037086552e-05,
+      "loss": 0.6358,
+      "step": 15339
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4600265121657844e-05,
+      "loss": 0.5681,
+      "step": 15340
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4599632182853838e-05,
+      "loss": 0.5898,
+      "step": 15341
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4598999220677757e-05,
+      "loss": 0.6059,
+      "step": 15342
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4598366235132813e-05,
+      "loss": 0.5625,
+      "step": 15343
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4597733226222222e-05,
+      "loss": 0.6319,
+      "step": 15344
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4597100193949203e-05,
+      "loss": 0.6114,
+      "step": 15345
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4596467138316972e-05,
+      "loss": 0.5977,
+      "step": 15346
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4595834059328744e-05,
+      "loss": 0.5653,
+      "step": 15347
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4595200956987741e-05,
+      "loss": 0.642,
+      "step": 15348
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4594567831297171e-05,
+      "loss": 0.6221,
+      "step": 15349
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.459393468226026e-05,
+      "loss": 0.5519,
+      "step": 15350
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.459330150988022e-05,
+      "loss": 0.6179,
+      "step": 15351
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4592668314160272e-05,
+      "loss": 0.6321,
+      "step": 15352
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4592035095103629e-05,
+      "loss": 0.6248,
+      "step": 15353
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4591401852713511e-05,
+      "loss": 0.6403,
+      "step": 15354
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.4590768586993136e-05,
+      "loss": 0.6363,
+      "step": 15355
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4590135297945722e-05,
+      "loss": 0.5487,
+      "step": 15356
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4589501985574484e-05,
+      "loss": 0.5596,
+      "step": 15357
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4588868649882643e-05,
+      "loss": 0.6349,
+      "step": 15358
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4588235290873417e-05,
+      "loss": 0.6124,
+      "step": 15359
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4587601908550025e-05,
+      "loss": 0.5895,
+      "step": 15360
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4586968502915682e-05,
+      "loss": 0.6359,
+      "step": 15361
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.458633507397361e-05,
+      "loss": 0.5599,
+      "step": 15362
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4585701621727024e-05,
+      "loss": 0.5781,
+      "step": 15363
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4585068146179148e-05,
+      "loss": 0.6098,
+      "step": 15364
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4584434647333199e-05,
+      "loss": 0.5986,
+      "step": 15365
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4583801125192391e-05,
+      "loss": 0.6036,
+      "step": 15366
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4583167579759948e-05,
+      "loss": 0.5938,
+      "step": 15367
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4582534011039092e-05,
+      "loss": 0.5977,
+      "step": 15368
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4581900419033035e-05,
+      "loss": 0.5909,
+      "step": 15369
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4581266803745e-05,
+      "loss": 0.5909,
+      "step": 15370
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.458063316517821e-05,
+      "loss": 0.5522,
+      "step": 15371
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.457999950333588e-05,
+      "loss": 0.6136,
+      "step": 15372
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4579365818221229e-05,
+      "loss": 0.5966,
+      "step": 15373
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4578732109837482e-05,
+      "loss": 0.6138,
+      "step": 15374
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4578098378187854e-05,
+      "loss": 0.6282,
+      "step": 15375
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4577464623275568e-05,
+      "loss": 0.579,
+      "step": 15376
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4576830845103844e-05,
+      "loss": 0.612,
+      "step": 15377
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4576197043675902e-05,
+      "loss": 0.6184,
+      "step": 15378
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4575563218994961e-05,
+      "loss": 0.5952,
+      "step": 15379
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4574929371064244e-05,
+      "loss": 0.6008,
+      "step": 15380
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4574295499886972e-05,
+      "loss": 0.6287,
+      "step": 15381
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4573661605466365e-05,
+      "loss": 0.6202,
+      "step": 15382
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4573027687805645e-05,
+      "loss": 0.6205,
+      "step": 15383
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4572393746908029e-05,
+      "loss": 0.5792,
+      "step": 15384
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4571759782776744e-05,
+      "loss": 0.61,
+      "step": 15385
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4571125795415009e-05,
+      "loss": 0.5907,
+      "step": 15386
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4570491784826042e-05,
+      "loss": 0.5594,
+      "step": 15387
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4569857751013069e-05,
+      "loss": 0.5445,
+      "step": 15388
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.456922369397931e-05,
+      "loss": 0.6459,
+      "step": 15389
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4568589613727991e-05,
+      "loss": 0.6378,
+      "step": 15390
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4567955510262326e-05,
+      "loss": 0.5948,
+      "step": 15391
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4567321383585544e-05,
+      "loss": 0.5468,
+      "step": 15392
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4566687233700863e-05,
+      "loss": 0.5964,
+      "step": 15393
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4566053060611507e-05,
+      "loss": 0.6185,
+      "step": 15394
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4565418864320697e-05,
+      "loss": 0.6012,
+      "step": 15395
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4564784644831657e-05,
+      "loss": 0.6336,
+      "step": 15396
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4564150402147611e-05,
+      "loss": 0.6718,
+      "step": 15397
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4563516136271778e-05,
+      "loss": 0.6363,
+      "step": 15398
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4562881847207384e-05,
+      "loss": 0.5526,
+      "step": 15399
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4562247534957652e-05,
+      "loss": 0.5941,
+      "step": 15400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4561613199525804e-05,
+      "loss": 0.6109,
+      "step": 15401
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4560978840915064e-05,
+      "loss": 0.5678,
+      "step": 15402
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4560344459128653e-05,
+      "loss": 0.5839,
+      "step": 15403
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4559710054169793e-05,
+      "loss": 0.6304,
+      "step": 15404
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4559075626041717e-05,
+      "loss": 0.5779,
+      "step": 15405
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.455844117474764e-05,
+      "loss": 0.6057,
+      "step": 15406
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4557806700290787e-05,
+      "loss": 0.6306,
+      "step": 15407
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4557172202674386e-05,
+      "loss": 0.6254,
+      "step": 15408
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4556537681901657e-05,
+      "loss": 0.6603,
+      "step": 15409
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4555903137975827e-05,
+      "loss": 0.551,
+      "step": 15410
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4555268570900119e-05,
+      "loss": 0.6158,
+      "step": 15411
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4554633980677755e-05,
+      "loss": 0.5521,
+      "step": 15412
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4553999367311964e-05,
+      "loss": 0.6223,
+      "step": 15413
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4553364730805966e-05,
+      "loss": 0.591,
+      "step": 15414
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4552730071162988e-05,
+      "loss": 0.5802,
+      "step": 15415
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.455209538838626e-05,
+      "loss": 0.5515,
+      "step": 15416
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4551460682478998e-05,
+      "loss": 0.595,
+      "step": 15417
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4550825953444432e-05,
+      "loss": 0.6301,
+      "step": 15418
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4550191201285788e-05,
+      "loss": 0.6309,
+      "step": 15419
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4549556426006288e-05,
+      "loss": 0.5192,
+      "step": 15420
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.454892162760916e-05,
+      "loss": 0.6028,
+      "step": 15421
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.454828680609763e-05,
+      "loss": 0.5933,
+      "step": 15422
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4547651961474922e-05,
+      "loss": 0.5838,
+      "step": 15423
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4547017093744261e-05,
+      "loss": 0.645,
+      "step": 15424
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4546382202908878e-05,
+      "loss": 0.5463,
+      "step": 15425
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4545747288971992e-05,
+      "loss": 0.6128,
+      "step": 15426
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4545112351936835e-05,
+      "loss": 0.5916,
+      "step": 15427
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4544477391806631e-05,
+      "loss": 0.5387,
+      "step": 15428
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4543842408584605e-05,
+      "loss": 0.5801,
+      "step": 15429
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4543207402273986e-05,
+      "loss": 0.5892,
+      "step": 15430
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4542572372878002e-05,
+      "loss": 0.6091,
+      "step": 15431
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4541937320399875e-05,
+      "loss": 0.5912,
+      "step": 15432
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4541302244842833e-05,
+      "loss": 0.577,
+      "step": 15433
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4540667146210105e-05,
+      "loss": 0.599,
+      "step": 15434
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.454003202450492e-05,
+      "loss": 0.6419,
+      "step": 15435
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4539396879730498e-05,
+      "loss": 0.6171,
+      "step": 15436
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4538761711890073e-05,
+      "loss": 0.5666,
+      "step": 15437
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4538126520986872e-05,
+      "loss": 0.627,
+      "step": 15438
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.453749130702412e-05,
+      "loss": 0.6556,
+      "step": 15439
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4536856070005048e-05,
+      "loss": 0.5697,
+      "step": 15440
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.453622080993288e-05,
+      "loss": 0.6039,
+      "step": 15441
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4535585526810843e-05,
+      "loss": 0.5685,
+      "step": 15442
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.453495022064217e-05,
+      "loss": 0.5717,
+      "step": 15443
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4534314891430085e-05,
+      "loss": 0.5691,
+      "step": 15444
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.453367953917782e-05,
+      "loss": 0.5785,
+      "step": 15445
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4533044163888603e-05,
+      "loss": 0.65,
+      "step": 15446
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4532408765565659e-05,
+      "loss": 0.592,
+      "step": 15447
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4531773344212221e-05,
+      "loss": 0.6009,
+      "step": 15448
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4531137899831511e-05,
+      "loss": 0.5696,
+      "step": 15449
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4530502432426767e-05,
+      "loss": 0.6026,
+      "step": 15450
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4529866942001212e-05,
+      "loss": 0.5657,
+      "step": 15451
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4529231428558074e-05,
+      "loss": 0.6111,
+      "step": 15452
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4528595892100587e-05,
+      "loss": 0.5775,
+      "step": 15453
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4527960332631979e-05,
+      "loss": 0.5751,
+      "step": 15454
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4527324750155476e-05,
+      "loss": 0.5737,
+      "step": 15455
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4526689144674309e-05,
+      "loss": 0.5752,
+      "step": 15456
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4526053516191712e-05,
+      "loss": 0.5745,
+      "step": 15457
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4525417864710909e-05,
+      "loss": 0.5941,
+      "step": 15458
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4524782190235134e-05,
+      "loss": 0.5987,
+      "step": 15459
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4524146492767615e-05,
+      "loss": 0.631,
+      "step": 15460
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4523510772311582e-05,
+      "loss": 0.6311,
+      "step": 15461
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4522875028870269e-05,
+      "loss": 0.5803,
+      "step": 15462
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.45222392624469e-05,
+      "loss": 0.6575,
+      "step": 15463
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.452160347304471e-05,
+      "loss": 0.5631,
+      "step": 15464
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4520967660666929e-05,
+      "loss": 0.6125,
+      "step": 15465
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4520331825316786e-05,
+      "loss": 0.5716,
+      "step": 15466
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4519695966997516e-05,
+      "loss": 0.6581,
+      "step": 15467
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4519060085712345e-05,
+      "loss": 0.6468,
+      "step": 15468
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4518424181464505e-05,
+      "loss": 0.6225,
+      "step": 15469
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4517788254257233e-05,
+      "loss": 0.5891,
+      "step": 15470
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4517152304093752e-05,
+      "loss": 0.5658,
+      "step": 15471
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4516516330977297e-05,
+      "loss": 0.6155,
+      "step": 15472
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4515880334911101e-05,
+      "loss": 0.5771,
+      "step": 15473
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4515244315898393e-05,
+      "loss": 0.647,
+      "step": 15474
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.451460827394241e-05,
+      "loss": 0.6051,
+      "step": 15475
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4513972209046376e-05,
+      "loss": 0.5817,
+      "step": 15476
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4513336121213529e-05,
+      "loss": 0.5629,
+      "step": 15477
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.45127000104471e-05,
+      "loss": 0.5506,
+      "step": 15478
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4512063876750319e-05,
+      "loss": 0.6179,
+      "step": 15479
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.451142772012642e-05,
+      "loss": 0.5548,
+      "step": 15480
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.451079154057864e-05,
+      "loss": 0.5916,
+      "step": 15481
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4510155338110202e-05,
+      "loss": 0.6167,
+      "step": 15482
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4509519112724343e-05,
+      "loss": 0.6048,
+      "step": 15483
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4508882864424302e-05,
+      "loss": 0.6158,
+      "step": 15484
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4508246593213303e-05,
+      "loss": 0.6133,
+      "step": 15485
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4507610299094585e-05,
+      "loss": 0.5763,
+      "step": 15486
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4506973982071378e-05,
+      "loss": 0.6064,
+      "step": 15487
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4506337642146916e-05,
+      "loss": 0.5564,
+      "step": 15488
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.450570127932443e-05,
+      "loss": 0.6433,
+      "step": 15489
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4505064893607162e-05,
+      "loss": 0.5777,
+      "step": 15490
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4504428484998334e-05,
+      "loss": 0.5958,
+      "step": 15491
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4503792053501191e-05,
+      "loss": 0.6561,
+      "step": 15492
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4503155599118959e-05,
+      "loss": 0.5588,
+      "step": 15493
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4502519121854874e-05,
+      "loss": 0.6176,
+      "step": 15494
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4501882621712171e-05,
+      "loss": 0.5602,
+      "step": 15495
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4501246098694084e-05,
+      "loss": 0.5988,
+      "step": 15496
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4500609552803848e-05,
+      "loss": 0.6192,
+      "step": 15497
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4499972984044697e-05,
+      "loss": 0.6135,
+      "step": 15498
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4499336392419863e-05,
+      "loss": 0.6066,
+      "step": 15499
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4498699777932585e-05,
+      "loss": 0.6545,
+      "step": 15500
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4498063140586097e-05,
+      "loss": 0.5553,
+      "step": 15501
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.4497426480383632e-05,
+      "loss": 0.6645,
+      "step": 15502
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4496789797328425e-05,
+      "loss": 0.6347,
+      "step": 15503
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4496153091423712e-05,
+      "loss": 0.5948,
+      "step": 15504
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.449551636267273e-05,
+      "loss": 0.5282,
+      "step": 15505
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4494879611078711e-05,
+      "loss": 0.6306,
+      "step": 15506
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4494242836644894e-05,
+      "loss": 0.6521,
+      "step": 15507
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4493606039374512e-05,
+      "loss": 0.5678,
+      "step": 15508
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4492969219270803e-05,
+      "loss": 0.6396,
+      "step": 15509
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4492332376336999e-05,
+      "loss": 0.6456,
+      "step": 15510
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4491695510576343e-05,
+      "loss": 0.6107,
+      "step": 15511
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4491058621992063e-05,
+      "loss": 0.5666,
+      "step": 15512
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4490421710587398e-05,
+      "loss": 0.5569,
+      "step": 15513
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4489784776365587e-05,
+      "loss": 0.6261,
+      "step": 15514
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4489147819329862e-05,
+      "loss": 0.6272,
+      "step": 15515
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4488510839483467e-05,
+      "loss": 0.5864,
+      "step": 15516
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4487873836829631e-05,
+      "loss": 0.6199,
+      "step": 15517
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4487236811371594e-05,
+      "loss": 0.545,
+      "step": 15518
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4486599763112594e-05,
+      "loss": 0.6044,
+      "step": 15519
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4485962692055863e-05,
+      "loss": 0.6074,
+      "step": 15520
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4485325598204647e-05,
+      "loss": 0.6046,
+      "step": 15521
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4484688481562175e-05,
+      "loss": 0.592,
+      "step": 15522
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4484051342131686e-05,
+      "loss": 0.571,
+      "step": 15523
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4483414179916421e-05,
+      "loss": 0.5872,
+      "step": 15524
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4482776994919613e-05,
+      "loss": 0.6137,
+      "step": 15525
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4482139787144507e-05,
+      "loss": 0.6509,
+      "step": 15526
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4481502556594334e-05,
+      "loss": 0.5992,
+      "step": 15527
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4480865303272334e-05,
+      "loss": 0.6119,
+      "step": 15528
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4480228027181744e-05,
+      "loss": 0.61,
+      "step": 15529
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4479590728325805e-05,
+      "loss": 0.619,
+      "step": 15530
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4478953406707751e-05,
+      "loss": 0.6266,
+      "step": 15531
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4478316062330826e-05,
+      "loss": 0.6441,
+      "step": 15532
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4477678695198263e-05,
+      "loss": 0.6106,
+      "step": 15533
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4477041305313305e-05,
+      "loss": 0.5697,
+      "step": 15534
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4476403892679188e-05,
+      "loss": 0.5919,
+      "step": 15535
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4475766457299153e-05,
+      "loss": 0.591,
+      "step": 15536
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4475128999176436e-05,
+      "loss": 0.6256,
+      "step": 15537
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4474491518314277e-05,
+      "loss": 0.6338,
+      "step": 15538
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4473854014715918e-05,
+      "loss": 0.6236,
+      "step": 15539
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4473216488384599e-05,
+      "loss": 0.6038,
+      "step": 15540
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4472578939323554e-05,
+      "loss": 0.5716,
+      "step": 15541
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4471941367536024e-05,
+      "loss": 0.6358,
+      "step": 15542
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4471303773025254e-05,
+      "loss": 0.6574,
+      "step": 15543
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.447066615579448e-05,
+      "loss": 0.6127,
+      "step": 15544
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.447002851584694e-05,
+      "loss": 0.6597,
+      "step": 15545
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4469390853185874e-05,
+      "loss": 0.5954,
+      "step": 15546
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.446875316781453e-05,
+      "loss": 0.6151,
+      "step": 15547
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4468115459736138e-05,
+      "loss": 0.6679,
+      "step": 15548
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4467477728953944e-05,
+      "loss": 0.5698,
+      "step": 15549
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4466839975471186e-05,
+      "loss": 0.5839,
+      "step": 15550
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4466202199291109e-05,
+      "loss": 0.5854,
+      "step": 15551
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4465564400416947e-05,
+      "loss": 0.6143,
+      "step": 15552
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4464926578851946e-05,
+      "loss": 0.6311,
+      "step": 15553
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4464288734599346e-05,
+      "loss": 0.6091,
+      "step": 15554
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4463650867662387e-05,
+      "loss": 0.6252,
+      "step": 15555
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.446301297804431e-05,
+      "loss": 0.6198,
+      "step": 15556
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4462375065748358e-05,
+      "loss": 0.5398,
+      "step": 15557
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.446173713077777e-05,
+      "loss": 0.5644,
+      "step": 15558
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4461099173135793e-05,
+      "loss": 0.5504,
+      "step": 15559
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4460461192825661e-05,
+      "loss": 0.6394,
+      "step": 15560
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4459823189850619e-05,
+      "loss": 0.6682,
+      "step": 15561
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4459185164213911e-05,
+      "loss": 0.5305,
+      "step": 15562
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4458547115918775e-05,
+      "loss": 0.6271,
+      "step": 15563
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4457909044968456e-05,
+      "loss": 0.5699,
+      "step": 15564
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4457270951366198e-05,
+      "loss": 0.6162,
+      "step": 15565
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.445663283511524e-05,
+      "loss": 0.6515,
+      "step": 15566
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4455994696218825e-05,
+      "loss": 0.6077,
+      "step": 15567
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4455356534680199e-05,
+      "loss": 0.6064,
+      "step": 15568
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4454718350502596e-05,
+      "loss": 0.589,
+      "step": 15569
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4454080143689266e-05,
+      "loss": 0.6074,
+      "step": 15570
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4453441914243453e-05,
+      "loss": 0.6115,
+      "step": 15571
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4452803662168393e-05,
+      "loss": 0.6202,
+      "step": 15572
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4452165387467339e-05,
+      "loss": 0.6222,
+      "step": 15573
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4451527090143524e-05,
+      "loss": 0.6351,
+      "step": 15574
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4450888770200198e-05,
+      "loss": 0.5973,
+      "step": 15575
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4450250427640604e-05,
+      "loss": 0.5641,
+      "step": 15576
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4449612062467981e-05,
+      "loss": 0.5243,
+      "step": 15577
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4448973674685581e-05,
+      "loss": 0.5598,
+      "step": 15578
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4448335264296637e-05,
+      "loss": 0.5757,
+      "step": 15579
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.44476968313044e-05,
+      "loss": 0.6006,
+      "step": 15580
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4447058375712117e-05,
+      "loss": 0.6134,
+      "step": 15581
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4446419897523026e-05,
+      "loss": 0.6204,
+      "step": 15582
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4445781396740371e-05,
+      "loss": 0.5838,
+      "step": 15583
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4445142873367401e-05,
+      "loss": 0.5969,
+      "step": 15584
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4444504327407356e-05,
+      "loss": 0.5547,
+      "step": 15585
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4443865758863487e-05,
+      "loss": 0.6013,
+      "step": 15586
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4443227167739034e-05,
+      "loss": 0.5958,
+      "step": 15587
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4442588554037241e-05,
+      "loss": 0.6485,
+      "step": 15588
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4441949917761352e-05,
+      "loss": 0.6102,
+      "step": 15589
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.444131125891462e-05,
+      "loss": 0.5209,
+      "step": 15590
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.444067257750028e-05,
+      "loss": 0.6591,
+      "step": 15591
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4440033873521584e-05,
+      "loss": 0.6156,
+      "step": 15592
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4439395146981774e-05,
+      "loss": 0.5815,
+      "step": 15593
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4438756397884097e-05,
+      "loss": 0.6466,
+      "step": 15594
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.44381176262318e-05,
+      "loss": 0.5341,
+      "step": 15595
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.443747883202813e-05,
+      "loss": 0.5922,
+      "step": 15596
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4436840015276327e-05,
+      "loss": 0.5612,
+      "step": 15597
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.443620117597964e-05,
+      "loss": 0.5565,
+      "step": 15598
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4435562314141319e-05,
+      "loss": 0.5689,
+      "step": 15599
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4434923429764605e-05,
+      "loss": 0.6163,
+      "step": 15600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4434284522852747e-05,
+      "loss": 0.6471,
+      "step": 15601
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4433645593408989e-05,
+      "loss": 0.6373,
+      "step": 15602
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.443300664143658e-05,
+      "loss": 0.5507,
+      "step": 15603
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4432367666938766e-05,
+      "loss": 0.5471,
+      "step": 15604
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4431728669918796e-05,
+      "loss": 0.6129,
+      "step": 15605
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4431089650379913e-05,
+      "loss": 0.6489,
+      "step": 15606
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4430450608325366e-05,
+      "loss": 0.546,
+      "step": 15607
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4429811543758403e-05,
+      "loss": 0.6376,
+      "step": 15608
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4429172456682268e-05,
+      "loss": 0.5468,
+      "step": 15609
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4428533347100212e-05,
+      "loss": 0.5943,
+      "step": 15610
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4427894215015478e-05,
+      "loss": 0.5961,
+      "step": 15611
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.442725506043132e-05,
+      "loss": 0.5426,
+      "step": 15612
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4426615883350984e-05,
+      "loss": 0.6026,
+      "step": 15613
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.442597668377771e-05,
+      "loss": 0.5981,
+      "step": 15614
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.442533746171476e-05,
+      "loss": 0.6073,
+      "step": 15615
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4424698217165369e-05,
+      "loss": 0.621,
+      "step": 15616
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4424058950132791e-05,
+      "loss": 0.5669,
+      "step": 15617
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4423419660620274e-05,
+      "loss": 0.6375,
+      "step": 15618
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.442278034863107e-05,
+      "loss": 0.5768,
+      "step": 15619
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4422141014168421e-05,
+      "loss": 0.5967,
+      "step": 15620
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4421501657235576e-05,
+      "loss": 0.6311,
+      "step": 15621
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4420862277835789e-05,
+      "loss": 0.6388,
+      "step": 15622
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4420222875972307e-05,
+      "loss": 0.6093,
+      "step": 15623
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4419583451648378e-05,
+      "loss": 0.6106,
+      "step": 15624
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4418944004867248e-05,
+      "loss": 0.5206,
+      "step": 15625
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4418304535632173e-05,
+      "loss": 0.5774,
+      "step": 15626
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4417665043946399e-05,
+      "loss": 0.5872,
+      "step": 15627
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4417025529813172e-05,
+      "loss": 0.619,
+      "step": 15628
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4416385993235744e-05,
+      "loss": 0.584,
+      "step": 15629
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4415746434217369e-05,
+      "loss": 0.6233,
+      "step": 15630
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4415106852761289e-05,
+      "loss": 0.5873,
+      "step": 15631
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4414467248870762e-05,
+      "loss": 0.5968,
+      "step": 15632
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4413827622549033e-05,
+      "loss": 0.5959,
+      "step": 15633
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4413187973799352e-05,
+      "loss": 0.5789,
+      "step": 15634
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4412548302624971e-05,
+      "loss": 0.5979,
+      "step": 15635
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4411908609029139e-05,
+      "loss": 0.5647,
+      "step": 15636
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.441126889301511e-05,
+      "loss": 0.5769,
+      "step": 15637
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4410629154586128e-05,
+      "loss": 0.5222,
+      "step": 15638
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.440998939374545e-05,
+      "loss": 0.6005,
+      "step": 15639
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4409349610496324e-05,
+      "loss": 0.6288,
+      "step": 15640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4408709804842e-05,
+      "loss": 0.5791,
+      "step": 15641
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4408069976785732e-05,
+      "loss": 0.6022,
+      "step": 15642
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.440743012633077e-05,
+      "loss": 0.5512,
+      "step": 15643
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4406790253480363e-05,
+      "loss": 0.5852,
+      "step": 15644
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4406150358237766e-05,
+      "loss": 0.5825,
+      "step": 15645
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4405510440606228e-05,
+      "loss": 0.611,
+      "step": 15646
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4404870500589e-05,
+      "loss": 0.606,
+      "step": 15647
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4404230538189336e-05,
+      "loss": 0.6301,
+      "step": 15648
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4403590553410487e-05,
+      "loss": 0.618,
+      "step": 15649
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4402950546255702e-05,
+      "loss": 0.6395,
+      "step": 15650
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4402310516728241e-05,
+      "loss": 0.5598,
+      "step": 15651
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4401670464831347e-05,
+      "loss": 0.5972,
+      "step": 15652
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4401030390568278e-05,
+      "loss": 0.606,
+      "step": 15653
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4400390293942286e-05,
+      "loss": 0.6097,
+      "step": 15654
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.439975017495662e-05,
+      "loss": 0.5467,
+      "step": 15655
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4399110033614531e-05,
+      "loss": 0.6189,
+      "step": 15656
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4398469869919282e-05,
+      "loss": 0.6134,
+      "step": 15657
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4397829683874116e-05,
+      "loss": 0.5837,
+      "step": 15658
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4397189475482292e-05,
+      "loss": 0.6379,
+      "step": 15659
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4396549244747057e-05,
+      "loss": 0.6287,
+      "step": 15660
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.439590899167167e-05,
+      "loss": 0.6178,
+      "step": 15661
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4395268716259383e-05,
+      "loss": 0.6287,
+      "step": 15662
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4394628418513446e-05,
+      "loss": 0.6228,
+      "step": 15663
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4393988098437118e-05,
+      "loss": 0.6269,
+      "step": 15664
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4393347756033647e-05,
+      "loss": 0.6475,
+      "step": 15665
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.439270739130629e-05,
+      "loss": 0.5967,
+      "step": 15666
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4392067004258301e-05,
+      "loss": 0.5773,
+      "step": 15667
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4391426594892935e-05,
+      "loss": 0.6052,
+      "step": 15668
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.439078616321344e-05,
+      "loss": 0.5788,
+      "step": 15669
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4390145709223078e-05,
+      "loss": 0.6054,
+      "step": 15670
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4389505232925098e-05,
+      "loss": 0.613,
+      "step": 15671
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4388864734322756e-05,
+      "loss": 0.5978,
+      "step": 15672
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4388224213419309e-05,
+      "loss": 0.5736,
+      "step": 15673
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4387583670218009e-05,
+      "loss": 0.6565,
+      "step": 15674
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4386943104722109e-05,
+      "loss": 0.6208,
+      "step": 15675
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4386302516934869e-05,
+      "loss": 0.5546,
+      "step": 15676
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.438566190685954e-05,
+      "loss": 0.6175,
+      "step": 15677
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.438502127449938e-05,
+      "loss": 0.5809,
+      "step": 15678
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4384380619857643e-05,
+      "loss": 0.614,
+      "step": 15679
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.438373994293758e-05,
+      "loss": 0.5867,
+      "step": 15680
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4383099243742455e-05,
+      "loss": 0.5985,
+      "step": 15681
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4382458522275518e-05,
+      "loss": 0.6265,
+      "step": 15682
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4381817778540027e-05,
+      "loss": 0.5977,
+      "step": 15683
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4381177012539231e-05,
+      "loss": 0.5999,
+      "step": 15684
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4380536224276399e-05,
+      "loss": 0.5717,
+      "step": 15685
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4379895413754774e-05,
+      "loss": 0.6034,
+      "step": 15686
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4379254580977619e-05,
+      "loss": 0.5735,
+      "step": 15687
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.437861372594819e-05,
+      "loss": 0.5751,
+      "step": 15688
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4377972848669744e-05,
+      "loss": 0.6132,
+      "step": 15689
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4377331949145532e-05,
+      "loss": 0.6232,
+      "step": 15690
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4376691027378815e-05,
+      "loss": 0.6411,
+      "step": 15691
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4376050083372853e-05,
+      "loss": 0.6173,
+      "step": 15692
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4375409117130893e-05,
+      "loss": 0.5934,
+      "step": 15693
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.43747681286562e-05,
+      "loss": 0.59,
+      "step": 15694
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4374127117952032e-05,
+      "loss": 0.6726,
+      "step": 15695
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4373486085021639e-05,
+      "loss": 0.6644,
+      "step": 15696
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4372845029868288e-05,
+      "loss": 0.5833,
+      "step": 15697
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4372203952495225e-05,
+      "loss": 0.5929,
+      "step": 15698
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4371562852905715e-05,
+      "loss": 0.596,
+      "step": 15699
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4370921731103016e-05,
+      "loss": 0.5492,
+      "step": 15700
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.437028058709038e-05,
+      "loss": 0.6245,
+      "step": 15701
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4369639420871072e-05,
+      "loss": 0.6578,
+      "step": 15702
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4368998232448343e-05,
+      "loss": 0.5882,
+      "step": 15703
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.436835702182546e-05,
+      "loss": 0.5607,
+      "step": 15704
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.436771578900567e-05,
+      "loss": 0.6234,
+      "step": 15705
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4367074533992237e-05,
+      "loss": 0.6017,
+      "step": 15706
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4366433256788423e-05,
+      "loss": 0.6342,
+      "step": 15707
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4365791957397482e-05,
+      "loss": 0.5836,
+      "step": 15708
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.436515063582267e-05,
+      "loss": 0.63,
+      "step": 15709
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4364509292067252e-05,
+      "loss": 0.6021,
+      "step": 15710
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4363867926134486e-05,
+      "loss": 0.6281,
+      "step": 15711
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4363226538027623e-05,
+      "loss": 0.634,
+      "step": 15712
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4362585127749934e-05,
+      "loss": 0.6175,
+      "step": 15713
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.436194369530467e-05,
+      "loss": 0.5777,
+      "step": 15714
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4361302240695093e-05,
+      "loss": 0.5663,
+      "step": 15715
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4360660763924462e-05,
+      "loss": 0.5725,
+      "step": 15716
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4360019264996035e-05,
+      "loss": 0.6047,
+      "step": 15717
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4359377743913075e-05,
+      "loss": 0.5945,
+      "step": 15718
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.435873620067884e-05,
+      "loss": 0.6323,
+      "step": 15719
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4358094635296589e-05,
+      "loss": 0.6031,
+      "step": 15720
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4357453047769583e-05,
+      "loss": 0.5921,
+      "step": 15721
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4356811438101083e-05,
+      "loss": 0.6639,
+      "step": 15722
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4356169806294349e-05,
+      "loss": 0.6376,
+      "step": 15723
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4355528152352638e-05,
+      "loss": 0.5749,
+      "step": 15724
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4354886476279215e-05,
+      "loss": 0.6714,
+      "step": 15725
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4354244778077336e-05,
+      "loss": 0.589,
+      "step": 15726
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4353603057750267e-05,
+      "loss": 0.5922,
+      "step": 15727
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4352961315301262e-05,
+      "loss": 0.6204,
+      "step": 15728
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4352319550733587e-05,
+      "loss": 0.6837,
+      "step": 15729
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4351677764050504e-05,
+      "loss": 0.5999,
+      "step": 15730
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.435103595525527e-05,
+      "loss": 0.5721,
+      "step": 15731
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4350394124351148e-05,
+      "loss": 0.5856,
+      "step": 15732
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4349752271341402e-05,
+      "loss": 0.6096,
+      "step": 15733
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4349110396229287e-05,
+      "loss": 0.6004,
+      "step": 15734
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.434846849901807e-05,
+      "loss": 0.61,
+      "step": 15735
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.434782657971101e-05,
+      "loss": 0.5678,
+      "step": 15736
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4347184638311371e-05,
+      "loss": 0.6277,
+      "step": 15737
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4346542674822415e-05,
+      "loss": 0.613,
+      "step": 15738
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.43459006892474e-05,
+      "loss": 0.6238,
+      "step": 15739
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4345258681589591e-05,
+      "loss": 0.5757,
+      "step": 15740
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4344616651852255e-05,
+      "loss": 0.5914,
+      "step": 15741
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4343974600038644e-05,
+      "loss": 0.6336,
+      "step": 15742
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4343332526152028e-05,
+      "loss": 0.6311,
+      "step": 15743
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4342690430195667e-05,
+      "loss": 0.6223,
+      "step": 15744
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4342048312172824e-05,
+      "loss": 0.5699,
+      "step": 15745
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4341406172086765e-05,
+      "loss": 0.611,
+      "step": 15746
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4340764009940745e-05,
+      "loss": 0.6202,
+      "step": 15747
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4340121825738032e-05,
+      "loss": 0.5857,
+      "step": 15748
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4339479619481891e-05,
+      "loss": 0.6535,
+      "step": 15749
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4338837391175582e-05,
+      "loss": 0.5797,
+      "step": 15750
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4338195140822372e-05,
+      "loss": 0.5669,
+      "step": 15751
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.433755286842552e-05,
+      "loss": 0.5625,
+      "step": 15752
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.433691057398829e-05,
+      "loss": 0.56,
+      "step": 15753
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.433626825751395e-05,
+      "loss": 0.6351,
+      "step": 15754
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.433562591900576e-05,
+      "loss": 0.6385,
+      "step": 15755
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4334983558466983e-05,
+      "loss": 0.5531,
+      "step": 15756
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4334341175900889e-05,
+      "loss": 0.6101,
+      "step": 15757
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4333698771310735e-05,
+      "loss": 0.5583,
+      "step": 15758
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.433305634469979e-05,
+      "loss": 0.571,
+      "step": 15759
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4332413896071317e-05,
+      "loss": 0.6139,
+      "step": 15760
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4331771425428577e-05,
+      "loss": 0.6339,
+      "step": 15761
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4331128932774842e-05,
+      "loss": 0.6214,
+      "step": 15762
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4330486418113374e-05,
+      "loss": 0.5858,
+      "step": 15763
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4329843881447432e-05,
+      "loss": 0.6312,
+      "step": 15764
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4329201322780286e-05,
+      "loss": 0.5486,
+      "step": 15765
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.43285587421152e-05,
+      "loss": 0.638,
+      "step": 15766
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4327916139455438e-05,
+      "loss": 0.6174,
+      "step": 15767
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4327273514804269e-05,
+      "loss": 0.6,
+      "step": 15768
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4326630868164955e-05,
+      "loss": 0.6307,
+      "step": 15769
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4325988199540764e-05,
+      "loss": 0.6521,
+      "step": 15770
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4325345508934959e-05,
+      "loss": 0.633,
+      "step": 15771
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4324702796350804e-05,
+      "loss": 0.5395,
+      "step": 15772
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4324060061791571e-05,
+      "loss": 0.619,
+      "step": 15773
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.432341730526052e-05,
+      "loss": 0.6988,
+      "step": 15774
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4322774526760923e-05,
+      "loss": 0.6391,
+      "step": 15775
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.432213172629604e-05,
+      "loss": 0.6202,
+      "step": 15776
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4321488903869141e-05,
+      "loss": 0.6012,
+      "step": 15777
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.432084605948349e-05,
+      "loss": 0.5964,
+      "step": 15778
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4320203193142357e-05,
+      "loss": 0.5709,
+      "step": 15779
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4319560304849002e-05,
+      "loss": 0.6132,
+      "step": 15780
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4318917394606701e-05,
+      "loss": 0.546,
+      "step": 15781
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4318274462418713e-05,
+      "loss": 0.5951,
+      "step": 15782
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.431763150828831e-05,
+      "loss": 0.6043,
+      "step": 15783
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4316988532218757e-05,
+      "loss": 0.6062,
+      "step": 15784
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.431634553421332e-05,
+      "loss": 0.5885,
+      "step": 15785
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4315702514275262e-05,
+      "loss": 0.5937,
+      "step": 15786
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4315059472407863e-05,
+      "loss": 0.5949,
+      "step": 15787
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.431441640861438e-05,
+      "loss": 0.5928,
+      "step": 15788
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4313773322898085e-05,
+      "loss": 0.5894,
+      "step": 15789
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4313130215262243e-05,
+      "loss": 0.5918,
+      "step": 15790
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4312487085710122e-05,
+      "loss": 0.662,
+      "step": 15791
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4311843934244993e-05,
+      "loss": 0.6631,
+      "step": 15792
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4311200760870123e-05,
+      "loss": 0.6365,
+      "step": 15793
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4310557565588777e-05,
+      "loss": 0.6139,
+      "step": 15794
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4309914348404228e-05,
+      "loss": 0.6168,
+      "step": 15795
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.430927110931974e-05,
+      "loss": 0.6458,
+      "step": 15796
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4308627848338585e-05,
+      "loss": 0.6047,
+      "step": 15797
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.430798456546403e-05,
+      "loss": 0.622,
+      "step": 15798
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4307341260699343e-05,
+      "loss": 0.5485,
+      "step": 15799
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4306697934047793e-05,
+      "loss": 0.5634,
+      "step": 15800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.430605458551265e-05,
+      "loss": 0.591,
+      "step": 15801
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4305411215097185e-05,
+      "loss": 0.5502,
+      "step": 15802
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4304767822804665e-05,
+      "loss": 0.6097,
+      "step": 15803
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4304124408638353e-05,
+      "loss": 0.5925,
+      "step": 15804
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.430348097260153e-05,
+      "loss": 0.6103,
+      "step": 15805
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4302837514697459e-05,
+      "loss": 0.6308,
+      "step": 15806
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4302194034929406e-05,
+      "loss": 0.6248,
+      "step": 15807
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.430155053330065e-05,
+      "loss": 0.6126,
+      "step": 15808
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4300907009814454e-05,
+      "loss": 0.5837,
+      "step": 15809
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4300263464474089e-05,
+      "loss": 0.5987,
+      "step": 15810
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4299619897282828e-05,
+      "loss": 0.5395,
+      "step": 15811
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4298976308243937e-05,
+      "loss": 0.6218,
+      "step": 15812
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.429833269736069e-05,
+      "loss": 0.5824,
+      "step": 15813
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4297689064636356e-05,
+      "loss": 0.6128,
+      "step": 15814
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4297045410074203e-05,
+      "loss": 0.6439,
+      "step": 15815
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4296401733677508e-05,
+      "loss": 0.6329,
+      "step": 15816
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4295758035449535e-05,
+      "loss": 0.6469,
+      "step": 15817
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4295114315393556e-05,
+      "loss": 0.5806,
+      "step": 15818
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4294470573512847e-05,
+      "loss": 0.6405,
+      "step": 15819
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4293826809810672e-05,
+      "loss": 0.5138,
+      "step": 15820
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4293183024290307e-05,
+      "loss": 0.673,
+      "step": 15821
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4292539216955019e-05,
+      "loss": 0.6417,
+      "step": 15822
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4291895387808086e-05,
+      "loss": 0.5594,
+      "step": 15823
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4291251536852773e-05,
+      "loss": 0.6223,
+      "step": 15824
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4290607664092354e-05,
+      "loss": 0.5721,
+      "step": 15825
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4289963769530104e-05,
+      "loss": 0.5745,
+      "step": 15826
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.428931985316929e-05,
+      "loss": 0.5892,
+      "step": 15827
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4288675915013187e-05,
+      "loss": 0.5875,
+      "step": 15828
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.428803195506506e-05,
+      "loss": 0.57,
+      "step": 15829
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4287387973328193e-05,
+      "loss": 0.599,
+      "step": 15830
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4286743969805849e-05,
+      "loss": 0.5799,
+      "step": 15831
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4286099944501304e-05,
+      "loss": 0.6481,
+      "step": 15832
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4285455897417832e-05,
+      "loss": 0.5999,
+      "step": 15833
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42848118285587e-05,
+      "loss": 0.5749,
+      "step": 15834
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4284167737927188e-05,
+      "loss": 0.6126,
+      "step": 15835
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4283523625526563e-05,
+      "loss": 0.6038,
+      "step": 15836
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4282879491360097e-05,
+      "loss": 0.6494,
+      "step": 15837
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4282235335431072e-05,
+      "loss": 0.5823,
+      "step": 15838
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4281591157742752e-05,
+      "loss": 0.6257,
+      "step": 15839
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4280946958298414e-05,
+      "loss": 0.6162,
+      "step": 15840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.428030273710133e-05,
+      "loss": 0.5972,
+      "step": 15841
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4279658494154776e-05,
+      "loss": 0.6146,
+      "step": 15842
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.427901422946202e-05,
+      "loss": 0.5234,
+      "step": 15843
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4278369943026345e-05,
+      "loss": 0.6098,
+      "step": 15844
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4277725634851013e-05,
+      "loss": 0.6115,
+      "step": 15845
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.427708130493931e-05,
+      "loss": 0.609,
+      "step": 15846
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42764369532945e-05,
+      "loss": 0.6326,
+      "step": 15847
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4275792579919863e-05,
+      "loss": 0.6207,
+      "step": 15848
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4275148184818675e-05,
+      "loss": 0.5766,
+      "step": 15849
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42745037679942e-05,
+      "loss": 0.6125,
+      "step": 15850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4273859329449726e-05,
+      "loss": 0.5388,
+      "step": 15851
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4273214869188518e-05,
+      "loss": 0.6434,
+      "step": 15852
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4272570387213854e-05,
+      "loss": 0.5929,
+      "step": 15853
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.427192588352901e-05,
+      "loss": 0.5634,
+      "step": 15854
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.427128135813726e-05,
+      "loss": 0.6302,
+      "step": 15855
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4270636811041874e-05,
+      "loss": 0.6526,
+      "step": 15856
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4269992242246136e-05,
+      "loss": 0.6145,
+      "step": 15857
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4269347651753316e-05,
+      "loss": 0.6137,
+      "step": 15858
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4268703039566689e-05,
+      "loss": 0.5956,
+      "step": 15859
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4268058405689534e-05,
+      "loss": 0.5552,
+      "step": 15860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4267413750125123e-05,
+      "loss": 0.5552,
+      "step": 15861
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4266769072876732e-05,
+      "loss": 0.578,
+      "step": 15862
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4266124373947641e-05,
+      "loss": 0.5619,
+      "step": 15863
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4265479653341119e-05,
+      "loss": 0.5866,
+      "step": 15864
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4264834911060448e-05,
+      "loss": 0.5778,
+      "step": 15865
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4264190147108898e-05,
+      "loss": 0.5593,
+      "step": 15866
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4263545361489751e-05,
+      "loss": 0.6113,
+      "step": 15867
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4262900554206285e-05,
+      "loss": 0.6544,
+      "step": 15868
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4262255725261769e-05,
+      "loss": 0.5893,
+      "step": 15869
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4261610874659484e-05,
+      "loss": 0.637,
+      "step": 15870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4260966002402707e-05,
+      "loss": 0.6006,
+      "step": 15871
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4260321108494713e-05,
+      "loss": 0.6207,
+      "step": 15872
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.425967619293878e-05,
+      "loss": 0.6322,
+      "step": 15873
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4259031255738185e-05,
+      "loss": 0.5846,
+      "step": 15874
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4258386296896202e-05,
+      "loss": 0.5725,
+      "step": 15875
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4257741316416116e-05,
+      "loss": 0.6056,
+      "step": 15876
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4257096314301198e-05,
+      "loss": 0.6229,
+      "step": 15877
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4256451290554724e-05,
+      "loss": 0.634,
+      "step": 15878
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4255806245179978e-05,
+      "loss": 0.6372,
+      "step": 15879
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4255161178180232e-05,
+      "loss": 0.5728,
+      "step": 15880
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4254516089558768e-05,
+      "loss": 0.6199,
+      "step": 15881
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.425387097931886e-05,
+      "loss": 0.6223,
+      "step": 15882
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4253225847463788e-05,
+      "loss": 0.5574,
+      "step": 15883
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.425258069399683e-05,
+      "loss": 0.6125,
+      "step": 15884
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4251935518921263e-05,
+      "loss": 0.6413,
+      "step": 15885
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4251290322240366e-05,
+      "loss": 0.5756,
+      "step": 15886
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4250645103957421e-05,
+      "loss": 0.5821,
+      "step": 15887
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42499998640757e-05,
+      "loss": 0.5723,
+      "step": 15888
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4249354602598489e-05,
+      "loss": 0.5887,
+      "step": 15889
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4248709319529059e-05,
+      "loss": 0.6087,
+      "step": 15890
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4248064014870693e-05,
+      "loss": 0.5573,
+      "step": 15891
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4247418688626672e-05,
+      "loss": 0.556,
+      "step": 15892
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.424677334080027e-05,
+      "loss": 0.5609,
+      "step": 15893
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.424612797139477e-05,
+      "loss": 0.6698,
+      "step": 15894
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.424548258041345e-05,
+      "loss": 0.6087,
+      "step": 15895
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4244837167859591e-05,
+      "loss": 0.5302,
+      "step": 15896
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.424419173373647e-05,
+      "loss": 0.6055,
+      "step": 15897
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4243546278047368e-05,
+      "loss": 0.6578,
+      "step": 15898
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4242900800795567e-05,
+      "loss": 0.6694,
+      "step": 15899
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.424225530198434e-05,
+      "loss": 0.5699,
+      "step": 15900
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4241609781616976e-05,
+      "loss": 0.5895,
+      "step": 15901
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4240964239696746e-05,
+      "loss": 0.5754,
+      "step": 15902
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4240318676226938e-05,
+      "loss": 0.5732,
+      "step": 15903
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4239673091210828e-05,
+      "loss": 0.588,
+      "step": 15904
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4239027484651696e-05,
+      "loss": 0.5909,
+      "step": 15905
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4238381856552826e-05,
+      "loss": 0.6034,
+      "step": 15906
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4237736206917494e-05,
+      "loss": 0.6245,
+      "step": 15907
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4237090535748987e-05,
+      "loss": 0.6146,
+      "step": 15908
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4236444843050578e-05,
+      "loss": 0.5634,
+      "step": 15909
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4235799128825553e-05,
+      "loss": 0.5614,
+      "step": 15910
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4235153393077195e-05,
+      "loss": 0.6165,
+      "step": 15911
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4234507635808781e-05,
+      "loss": 0.523,
+      "step": 15912
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.423386185702359e-05,
+      "loss": 0.5714,
+      "step": 15913
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4233216056724912e-05,
+      "loss": 0.6182,
+      "step": 15914
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.423257023491602e-05,
+      "loss": 0.5841,
+      "step": 15915
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42319243916002e-05,
+      "loss": 0.5883,
+      "step": 15916
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4231278526780734e-05,
+      "loss": 0.5866,
+      "step": 15917
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4230632640460902e-05,
+      "loss": 0.5821,
+      "step": 15918
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4229986732643988e-05,
+      "loss": 0.5983,
+      "step": 15919
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4229340803333271e-05,
+      "loss": 0.5695,
+      "step": 15920
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4228694852532034e-05,
+      "loss": 0.5793,
+      "step": 15921
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4228048880243562e-05,
+      "loss": 0.6631,
+      "step": 15922
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4227402886471134e-05,
+      "loss": 0.6103,
+      "step": 15923
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4226756871218032e-05,
+      "loss": 0.6198,
+      "step": 15924
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4226110834487544e-05,
+      "loss": 0.5938,
+      "step": 15925
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4225464776282946e-05,
+      "loss": 0.668,
+      "step": 15926
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4224818696607528e-05,
+      "loss": 0.62,
+      "step": 15927
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4224172595464565e-05,
+      "loss": 0.6058,
+      "step": 15928
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4223526472857345e-05,
+      "loss": 0.5501,
+      "step": 15929
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4222880328789153e-05,
+      "loss": 0.5915,
+      "step": 15930
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4222234163263265e-05,
+      "loss": 0.5875,
+      "step": 15931
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4221587976282969e-05,
+      "loss": 0.6305,
+      "step": 15932
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.422094176785155e-05,
+      "loss": 0.5877,
+      "step": 15933
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4220295537972288e-05,
+      "loss": 0.6044,
+      "step": 15934
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4219649286648474e-05,
+      "loss": 0.6492,
+      "step": 15935
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4219003013883379e-05,
+      "loss": 0.6191,
+      "step": 15936
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4218356719680297e-05,
+      "loss": 0.5663,
+      "step": 15937
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.421771040404251e-05,
+      "loss": 0.5542,
+      "step": 15938
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.42170640669733e-05,
+      "loss": 0.6251,
+      "step": 15939
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4216417708475953e-05,
+      "loss": 0.6219,
+      "step": 15940
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4215771328553756e-05,
+      "loss": 0.5967,
+      "step": 15941
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4215124927209987e-05,
+      "loss": 0.6205,
+      "step": 15942
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4214478504447931e-05,
+      "loss": 0.6078,
+      "step": 15943
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.421383206027088e-05,
+      "loss": 0.6531,
+      "step": 15944
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.421318559468211e-05,
+      "loss": 0.6085,
+      "step": 15945
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4212539107684915e-05,
+      "loss": 0.6063,
+      "step": 15946
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4211892599282573e-05,
+      "loss": 0.634,
+      "step": 15947
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4211246069478368e-05,
+      "loss": 0.5718,
+      "step": 15948
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4210599518275592e-05,
+      "loss": 0.6319,
+      "step": 15949
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4209952945677525e-05,
+      "loss": 0.5143,
+      "step": 15950
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4209306351687456e-05,
+      "loss": 0.6406,
+      "step": 15951
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4208659736308668e-05,
+      "loss": 0.6047,
+      "step": 15952
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4208013099544446e-05,
+      "loss": 0.6062,
+      "step": 15953
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.420736644139808e-05,
+      "loss": 0.635,
+      "step": 15954
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4206719761872849e-05,
+      "loss": 0.6093,
+      "step": 15955
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4206073060972046e-05,
+      "loss": 0.6344,
+      "step": 15956
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4205426338698954e-05,
+      "loss": 0.6247,
+      "step": 15957
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4204779595056857e-05,
+      "loss": 0.6424,
+      "step": 15958
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4204132830049045e-05,
+      "loss": 0.5555,
+      "step": 15959
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4203486043678803e-05,
+      "loss": 0.664,
+      "step": 15960
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4202839235949415e-05,
+      "loss": 0.6282,
+      "step": 15961
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4202192406864173e-05,
+      "loss": 0.5874,
+      "step": 15962
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4201545556426359e-05,
+      "loss": 0.6043,
+      "step": 15963
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4200898684639261e-05,
+      "loss": 0.5541,
+      "step": 15964
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.420025179150617e-05,
+      "loss": 0.629,
+      "step": 15965
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4199604877030366e-05,
+      "loss": 0.5904,
+      "step": 15966
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4198957941215138e-05,
+      "loss": 0.6802,
+      "step": 15967
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.419831098406378e-05,
+      "loss": 0.6244,
+      "step": 15968
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4197664005579572e-05,
+      "loss": 0.6177,
+      "step": 15969
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4197017005765804e-05,
+      "loss": 0.5918,
+      "step": 15970
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4196369984625764e-05,
+      "loss": 0.5906,
+      "step": 15971
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4195722942162737e-05,
+      "loss": 0.6283,
+      "step": 15972
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4195075878380016e-05,
+      "loss": 0.6019,
+      "step": 15973
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4194428793280883e-05,
+      "loss": 0.5931,
+      "step": 15974
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4193781686868628e-05,
+      "loss": 0.6596,
+      "step": 15975
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4193134559146545e-05,
+      "loss": 0.6694,
+      "step": 15976
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4192487410117913e-05,
+      "loss": 0.5532,
+      "step": 15977
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4191840239786028e-05,
+      "loss": 0.616,
+      "step": 15978
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4191193048154172e-05,
+      "loss": 0.6267,
+      "step": 15979
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.419054583522564e-05,
+      "loss": 0.6402,
+      "step": 15980
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4189898601003716e-05,
+      "loss": 0.5724,
+      "step": 15981
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4189251345491687e-05,
+      "loss": 0.5588,
+      "step": 15982
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4188604068692848e-05,
+      "loss": 0.6137,
+      "step": 15983
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4187956770610485e-05,
+      "loss": 0.6052,
+      "step": 15984
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4187309451247885e-05,
+      "loss": 0.5138,
+      "step": 15985
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4186662110608342e-05,
+      "loss": 0.5234,
+      "step": 15986
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4186014748695142e-05,
+      "loss": 0.5414,
+      "step": 15987
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4185367365511576e-05,
+      "loss": 0.548,
+      "step": 15988
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.418471996106093e-05,
+      "loss": 0.6299,
+      "step": 15989
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4184072535346499e-05,
+      "loss": 0.5777,
+      "step": 15990
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4183425088371568e-05,
+      "loss": 0.6106,
+      "step": 15991
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4182777620139429e-05,
+      "loss": 0.6191,
+      "step": 15992
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4182130130653373e-05,
+      "loss": 0.5153,
+      "step": 15993
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4181482619916689e-05,
+      "loss": 0.5936,
+      "step": 15994
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4180835087932668e-05,
+      "loss": 0.6645,
+      "step": 15995
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4180187534704598e-05,
+      "loss": 0.6367,
+      "step": 15996
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4179539960235774e-05,
+      "loss": 0.6319,
+      "step": 15997
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4178892364529477e-05,
+      "loss": 0.6276,
+      "step": 15998
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4178244747589012e-05,
+      "loss": 0.6077,
+      "step": 15999
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4177597109417658e-05,
+      "loss": 0.552,
+      "step": 16000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4176949450018707e-05,
+      "loss": 0.6137,
+      "step": 16001
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4176301769395455e-05,
+      "loss": 0.6315,
+      "step": 16002
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4175654067551192e-05,
+      "loss": 0.6201,
+      "step": 16003
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4175006344489206e-05,
+      "loss": 0.6017,
+      "step": 16004
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4174358600212789e-05,
+      "loss": 0.6297,
+      "step": 16005
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4173710834725237e-05,
+      "loss": 0.5731,
+      "step": 16006
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4173063048029833e-05,
+      "loss": 0.5967,
+      "step": 16007
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4172415240129876e-05,
+      "loss": 0.6358,
+      "step": 16008
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4171767411028657e-05,
+      "loss": 0.6036,
+      "step": 16009
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4171119560729463e-05,
+      "loss": 0.6343,
+      "step": 16010
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4170471689235591e-05,
+      "loss": 0.593,
+      "step": 16011
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.416982379655033e-05,
+      "loss": 0.6867,
+      "step": 16012
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4169175882676973e-05,
+      "loss": 0.5956,
+      "step": 16013
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4168527947618813e-05,
+      "loss": 0.6113,
+      "step": 16014
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4167879991379142e-05,
+      "loss": 0.6022,
+      "step": 16015
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4167232013961254e-05,
+      "loss": 0.5651,
+      "step": 16016
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4166584015368434e-05,
+      "loss": 0.5886,
+      "step": 16017
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4165935995603985e-05,
+      "loss": 0.6006,
+      "step": 16018
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4165287954671195e-05,
+      "loss": 0.5994,
+      "step": 16019
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4164639892573355e-05,
+      "loss": 0.5994,
+      "step": 16020
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4163991809313761e-05,
+      "loss": 0.6284,
+      "step": 16021
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4163343704895707e-05,
+      "loss": 0.6815,
+      "step": 16022
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.416269557932248e-05,
+      "loss": 0.6114,
+      "step": 16023
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4162047432597381e-05,
+      "loss": 0.5532,
+      "step": 16024
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.41613992647237e-05,
+      "loss": 0.5521,
+      "step": 16025
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4160751075704728e-05,
+      "loss": 0.6047,
+      "step": 16026
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4160102865543765e-05,
+      "loss": 0.6022,
+      "step": 16027
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4159454634244098e-05,
+      "loss": 0.5781,
+      "step": 16028
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4158806381809024e-05,
+      "loss": 0.5786,
+      "step": 16029
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.415815810824184e-05,
+      "loss": 0.6014,
+      "step": 16030
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4157509813545833e-05,
+      "loss": 0.6445,
+      "step": 16031
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4156861497724302e-05,
+      "loss": 0.6135,
+      "step": 16032
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4156213160780544e-05,
+      "loss": 0.6312,
+      "step": 16033
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4155564802717846e-05,
+      "loss": 0.6044,
+      "step": 16034
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4154916423539504e-05,
+      "loss": 0.5893,
+      "step": 16035
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4154268023248818e-05,
+      "loss": 0.6712,
+      "step": 16036
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4153619601849083e-05,
+      "loss": 0.6056,
+      "step": 16037
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4152971159343584e-05,
+      "loss": 0.5667,
+      "step": 16038
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4152322695735628e-05,
+      "loss": 0.5496,
+      "step": 16039
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4151674211028499e-05,
+      "loss": 0.5946,
+      "step": 16040
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.41510257052255e-05,
+      "loss": 0.6364,
+      "step": 16041
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4150377178329924e-05,
+      "loss": 0.594,
+      "step": 16042
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4149728630345066e-05,
+      "loss": 0.5814,
+      "step": 16043
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4149080061274221e-05,
+      "loss": 0.619,
+      "step": 16044
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4148431471120686e-05,
+      "loss": 0.6162,
+      "step": 16045
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4147782859887756e-05,
+      "loss": 0.554,
+      "step": 16046
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4147134227578722e-05,
+      "loss": 0.6563,
+      "step": 16047
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4146485574196889e-05,
+      "loss": 0.6076,
+      "step": 16048
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4145836899745549e-05,
+      "loss": 0.5943,
+      "step": 16049
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4145188204227994e-05,
+      "loss": 0.6112,
+      "step": 16050
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4144539487647523e-05,
+      "loss": 0.5749,
+      "step": 16051
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4143890750007439e-05,
+      "loss": 0.5445,
+      "step": 16052
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4143241991311028e-05,
+      "loss": 0.6544,
+      "step": 16053
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4142593211561591e-05,
+      "loss": 0.588,
+      "step": 16054
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4141944410762427e-05,
+      "loss": 0.6191,
+      "step": 16055
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.414129558891683e-05,
+      "loss": 0.5833,
+      "step": 16056
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4140646746028097e-05,
+      "loss": 0.6131,
+      "step": 16057
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4139997882099525e-05,
+      "loss": 0.5952,
+      "step": 16058
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.413934899713441e-05,
+      "loss": 0.6079,
+      "step": 16059
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4138700091136053e-05,
+      "loss": 0.5868,
+      "step": 16060
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4138051164107748e-05,
+      "loss": 0.6686,
+      "step": 16061
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4137402216052792e-05,
+      "loss": 0.601,
+      "step": 16062
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4136753246974483e-05,
+      "loss": 0.5293,
+      "step": 16063
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4136104256876122e-05,
+      "loss": 0.6222,
+      "step": 16064
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4135455245761003e-05,
+      "loss": 0.5434,
+      "step": 16065
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4134806213632425e-05,
+      "loss": 0.5922,
+      "step": 16066
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4134157160493682e-05,
+      "loss": 0.6121,
+      "step": 16067
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.413350808634808e-05,
+      "loss": 0.6234,
+      "step": 16068
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4132858991198912e-05,
+      "loss": 0.6636,
+      "step": 16069
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4132209875049475e-05,
+      "loss": 0.5844,
+      "step": 16070
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4131560737903072e-05,
+      "loss": 0.5599,
+      "step": 16071
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4130911579762997e-05,
+      "loss": 0.5959,
+      "step": 16072
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4130262400632552e-05,
+      "loss": 0.6012,
+      "step": 16073
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4129613200515033e-05,
+      "loss": 0.6134,
+      "step": 16074
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4128963979413741e-05,
+      "loss": 0.5759,
+      "step": 16075
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4128314737331974e-05,
+      "loss": 0.5611,
+      "step": 16076
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4127665474273032e-05,
+      "loss": 0.5907,
+      "step": 16077
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4127016190240208e-05,
+      "loss": 0.5945,
+      "step": 16078
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4126366885236811e-05,
+      "loss": 0.5975,
+      "step": 16079
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4125717559266132e-05,
+      "loss": 0.6195,
+      "step": 16080
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4125068212331474e-05,
+      "loss": 0.5935,
+      "step": 16081
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4124418844436138e-05,
+      "loss": 0.5984,
+      "step": 16082
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4123769455583422e-05,
+      "loss": 0.577,
+      "step": 16083
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4123120045776629e-05,
+      "loss": 0.6306,
+      "step": 16084
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4122470615019048e-05,
+      "loss": 0.6745,
+      "step": 16085
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4121821163313988e-05,
+      "loss": 0.5659,
+      "step": 16086
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4121171690664755e-05,
+      "loss": 0.6069,
+      "step": 16087
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4120522197074633e-05,
+      "loss": 0.5501,
+      "step": 16088
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4119872682546935e-05,
+      "loss": 0.6165,
+      "step": 16089
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4119223147084957e-05,
+      "loss": 0.5873,
+      "step": 16090
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4118573590692e-05,
+      "loss": 0.6264,
+      "step": 16091
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4117924013371365e-05,
+      "loss": 0.607,
+      "step": 16092
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.411727441512635e-05,
+      "loss": 0.5739,
+      "step": 16093
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4116624795960259e-05,
+      "loss": 0.5929,
+      "step": 16094
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4115975155876392e-05,
+      "loss": 0.6043,
+      "step": 16095
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.411532549487805e-05,
+      "loss": 0.6219,
+      "step": 16096
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4114675812968532e-05,
+      "loss": 0.583,
+      "step": 16097
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4114026110151145e-05,
+      "loss": 0.6025,
+      "step": 16098
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4113376386429182e-05,
+      "loss": 0.6035,
+      "step": 16099
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4112726641805952e-05,
+      "loss": 0.5883,
+      "step": 16100
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4112076876284753e-05,
+      "loss": 0.6173,
+      "step": 16101
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4111427089868887e-05,
+      "loss": 0.6322,
+      "step": 16102
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4110777282561656e-05,
+      "loss": 0.609,
+      "step": 16103
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.411012745436636e-05,
+      "loss": 0.6428,
+      "step": 16104
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4109477605286303e-05,
+      "loss": 0.6107,
+      "step": 16105
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.410882773532479e-05,
+      "loss": 0.6391,
+      "step": 16106
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4108177844485116e-05,
+      "loss": 0.5378,
+      "step": 16107
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4107527932770588e-05,
+      "loss": 0.633,
+      "step": 16108
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.410687800018451e-05,
+      "loss": 0.6315,
+      "step": 16109
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4106228046730181e-05,
+      "loss": 0.663,
+      "step": 16110
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4105578072410904e-05,
+      "loss": 0.5698,
+      "step": 16111
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4104928077229982e-05,
+      "loss": 0.6184,
+      "step": 16112
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4104278061190719e-05,
+      "loss": 0.6217,
+      "step": 16113
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4103628024296418e-05,
+      "loss": 0.5985,
+      "step": 16114
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.410297796655038e-05,
+      "loss": 0.5758,
+      "step": 16115
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4102327887955912e-05,
+      "loss": 0.6225,
+      "step": 16116
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4101677788516313e-05,
+      "loss": 0.5586,
+      "step": 16117
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4101027668234885e-05,
+      "loss": 0.6141,
+      "step": 16118
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4100377527114939e-05,
+      "loss": 0.6001,
+      "step": 16119
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.409972736515977e-05,
+      "loss": 0.6182,
+      "step": 16120
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4099077182372688e-05,
+      "loss": 0.6242,
+      "step": 16121
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4098426978756994e-05,
+      "loss": 0.6613,
+      "step": 16122
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4097776754315993e-05,
+      "loss": 0.6013,
+      "step": 16123
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4097126509052985e-05,
+      "loss": 0.5867,
+      "step": 16124
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.409647624297128e-05,
+      "loss": 0.607,
+      "step": 16125
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4095825956074178e-05,
+      "loss": 0.5966,
+      "step": 16126
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4095175648364985e-05,
+      "loss": 0.6485,
+      "step": 16127
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4094525319847007e-05,
+      "loss": 0.6157,
+      "step": 16128
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4093874970523544e-05,
+      "loss": 0.6534,
+      "step": 16129
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4093224600397908e-05,
+      "loss": 0.5671,
+      "step": 16130
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4092574209473396e-05,
+      "loss": 0.5827,
+      "step": 16131
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4091923797753313e-05,
+      "loss": 0.5767,
+      "step": 16132
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4091273365240972e-05,
+      "loss": 0.6243,
+      "step": 16133
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4090622911939668e-05,
+      "loss": 0.5865,
+      "step": 16134
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4089972437852715e-05,
+      "loss": 0.5747,
+      "step": 16135
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4089321942983413e-05,
+      "loss": 0.5244,
+      "step": 16136
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4088671427335068e-05,
+      "loss": 0.6741,
+      "step": 16137
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4088020890910988e-05,
+      "loss": 0.6409,
+      "step": 16138
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4087370333714474e-05,
+      "loss": 0.6277,
+      "step": 16139
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4086719755748834e-05,
+      "loss": 0.6312,
+      "step": 16140
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4086069157017376e-05,
+      "loss": 0.5568,
+      "step": 16141
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4085418537523405e-05,
+      "loss": 0.6633,
+      "step": 16142
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4084767897270221e-05,
+      "loss": 0.6341,
+      "step": 16143
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4084117236261141e-05,
+      "loss": 0.5921,
+      "step": 16144
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4083466554499463e-05,
+      "loss": 0.5916,
+      "step": 16145
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4082815851988491e-05,
+      "loss": 0.5934,
+      "step": 16146
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4082165128731542e-05,
+      "loss": 0.6874,
+      "step": 16147
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4081514384731913e-05,
+      "loss": 0.5662,
+      "step": 16148
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.408086361999292e-05,
+      "loss": 0.6523,
+      "step": 16149
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4080212834517857e-05,
+      "loss": 0.6249,
+      "step": 16150
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.407956202831004e-05,
+      "loss": 0.5958,
+      "step": 16151
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4078911201372773e-05,
+      "loss": 0.5864,
+      "step": 16152
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4078260353709366e-05,
+      "loss": 0.5134,
+      "step": 16153
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4077609485323121e-05,
+      "loss": 0.5964,
+      "step": 16154
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.407695859621735e-05,
+      "loss": 0.5773,
+      "step": 16155
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4076307686395358e-05,
+      "loss": 0.6228,
+      "step": 16156
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4075656755860455e-05,
+      "loss": 0.521,
+      "step": 16157
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.407500580461594e-05,
+      "loss": 0.5522,
+      "step": 16158
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4074354832665132e-05,
+      "loss": 0.6261,
+      "step": 16159
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4073703840011332e-05,
+      "loss": 0.5861,
+      "step": 16160
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4073052826657849e-05,
+      "loss": 0.6356,
+      "step": 16161
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4072401792607992e-05,
+      "loss": 0.6175,
+      "step": 16162
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4071750737865072e-05,
+      "loss": 0.5921,
+      "step": 16163
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.407109966243239e-05,
+      "loss": 0.5885,
+      "step": 16164
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4070448566313263e-05,
+      "loss": 0.6046,
+      "step": 16165
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4069797449510992e-05,
+      "loss": 0.6663,
+      "step": 16166
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4069146312028887e-05,
+      "loss": 0.61,
+      "step": 16167
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.406849515387026e-05,
+      "loss": 0.6228,
+      "step": 16168
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4067843975038415e-05,
+      "loss": 0.6116,
+      "step": 16169
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4067192775536664e-05,
+      "loss": 0.5377,
+      "step": 16170
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4066541555368318e-05,
+      "loss": 0.6165,
+      "step": 16171
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.406589031453668e-05,
+      "loss": 0.5524,
+      "step": 16172
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4065239053045068e-05,
+      "loss": 0.6392,
+      "step": 16173
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4064587770896783e-05,
+      "loss": 0.6231,
+      "step": 16174
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4063936468095136e-05,
+      "loss": 0.5923,
+      "step": 16175
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4063285144643438e-05,
+      "loss": 0.6025,
+      "step": 16176
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4062633800545001e-05,
+      "loss": 0.5698,
+      "step": 16177
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4061982435803128e-05,
+      "loss": 0.6243,
+      "step": 16178
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4061331050421134e-05,
+      "loss": 0.6182,
+      "step": 16179
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4060679644402326e-05,
+      "loss": 0.5528,
+      "step": 16180
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4060028217750017e-05,
+      "loss": 0.5499,
+      "step": 16181
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4059376770467517e-05,
+      "loss": 0.5789,
+      "step": 16182
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4058725302558132e-05,
+      "loss": 0.5489,
+      "step": 16183
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4058073814025177e-05,
+      "loss": 0.6168,
+      "step": 16184
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.405742230487196e-05,
+      "loss": 0.6388,
+      "step": 16185
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4056770775101792e-05,
+      "loss": 0.5861,
+      "step": 16186
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4056119224717981e-05,
+      "loss": 0.575,
+      "step": 16187
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4055467653723845e-05,
+      "loss": 0.6196,
+      "step": 16188
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4054816062122684e-05,
+      "loss": 0.6602,
+      "step": 16189
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4054164449917821e-05,
+      "loss": 0.658,
+      "step": 16190
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4053512817112558e-05,
+      "loss": 0.5784,
+      "step": 16191
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4052861163710212e-05,
+      "loss": 0.5352,
+      "step": 16192
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4052209489714091e-05,
+      "loss": 0.5365,
+      "step": 16193
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4051557795127507e-05,
+      "loss": 0.6262,
+      "step": 16194
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4050906079953769e-05,
+      "loss": 0.621,
+      "step": 16195
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4050254344196193e-05,
+      "loss": 0.6171,
+      "step": 16196
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4049602587858084e-05,
+      "loss": 0.6366,
+      "step": 16197
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4048950810942766e-05,
+      "loss": 0.6214,
+      "step": 16198
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4048299013453534e-05,
+      "loss": 0.574,
+      "step": 16199
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4047647195393714e-05,
+      "loss": 0.601,
+      "step": 16200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4046995356766616e-05,
+      "loss": 0.6458,
+      "step": 16201
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4046343497575545e-05,
+      "loss": 0.6449,
+      "step": 16202
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.404569161782382e-05,
+      "loss": 0.633,
+      "step": 16203
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.404503971751475e-05,
+      "loss": 0.6391,
+      "step": 16204
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4044387796651648e-05,
+      "loss": 0.5874,
+      "step": 16205
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4043735855237828e-05,
+      "loss": 0.5879,
+      "step": 16206
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.40430838932766e-05,
+      "loss": 0.6285,
+      "step": 16207
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4042431910771281e-05,
+      "loss": 0.6179,
+      "step": 16208
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4041779907725181e-05,
+      "loss": 0.6528,
+      "step": 16209
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4041127884141614e-05,
+      "loss": 0.5342,
+      "step": 16210
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4040475840023892e-05,
+      "loss": 0.587,
+      "step": 16211
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4039823775375329e-05,
+      "loss": 0.5757,
+      "step": 16212
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.403917169019924e-05,
+      "loss": 0.6541,
+      "step": 16213
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4038519584498936e-05,
+      "loss": 0.6165,
+      "step": 16214
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.403786745827773e-05,
+      "loss": 0.5856,
+      "step": 16215
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4037215311538936e-05,
+      "loss": 0.63,
+      "step": 16216
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.403656314428587e-05,
+      "loss": 0.5885,
+      "step": 16217
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4035910956521845e-05,
+      "loss": 0.6359,
+      "step": 16218
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4035258748250174e-05,
+      "loss": 0.6122,
+      "step": 16219
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4034606519474172e-05,
+      "loss": 0.5739,
+      "step": 16220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4033954270197151e-05,
+      "loss": 0.6253,
+      "step": 16221
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4033302000422431e-05,
+      "loss": 0.5833,
+      "step": 16222
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4032649710153318e-05,
+      "loss": 0.5443,
+      "step": 16223
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4031997399393132e-05,
+      "loss": 0.5583,
+      "step": 16224
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4031345068145188e-05,
+      "loss": 0.6105,
+      "step": 16225
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4030692716412797e-05,
+      "loss": 0.5582,
+      "step": 16226
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4030040344199276e-05,
+      "loss": 0.6481,
+      "step": 16227
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.402938795150794e-05,
+      "loss": 0.5454,
+      "step": 16228
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4028735538342104e-05,
+      "loss": 0.6131,
+      "step": 16229
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4028083104705084e-05,
+      "loss": 0.6248,
+      "step": 16230
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4027430650600194e-05,
+      "loss": 0.5928,
+      "step": 16231
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4026778176030749e-05,
+      "loss": 0.5635,
+      "step": 16232
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4026125681000062e-05,
+      "loss": 0.6391,
+      "step": 16233
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4025473165511457e-05,
+      "loss": 0.556,
+      "step": 16234
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4024820629568239e-05,
+      "loss": 0.6002,
+      "step": 16235
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4024168073173729e-05,
+      "loss": 0.5781,
+      "step": 16236
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.4023515496331244e-05,
+      "loss": 0.6106,
+      "step": 16237
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4022862899044096e-05,
+      "loss": 0.5947,
+      "step": 16238
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4022210281315607e-05,
+      "loss": 0.5962,
+      "step": 16239
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4021557643149088e-05,
+      "loss": 0.5469,
+      "step": 16240
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4020904984547858e-05,
+      "loss": 0.5585,
+      "step": 16241
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.402025230551523e-05,
+      "loss": 0.6261,
+      "step": 16242
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4019599606054524e-05,
+      "loss": 0.5811,
+      "step": 16243
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4018946886169056e-05,
+      "loss": 0.617,
+      "step": 16244
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.401829414586214e-05,
+      "loss": 0.5669,
+      "step": 16245
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4017641385137096e-05,
+      "loss": 0.6196,
+      "step": 16246
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4016988603997237e-05,
+      "loss": 0.6076,
+      "step": 16247
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4016335802445884e-05,
+      "loss": 0.6085,
+      "step": 16248
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4015682980486354e-05,
+      "loss": 0.5941,
+      "step": 16249
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4015030138121964e-05,
+      "loss": 0.6509,
+      "step": 16250
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4014377275356026e-05,
+      "loss": 0.5626,
+      "step": 16251
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4013724392191864e-05,
+      "loss": 0.5701,
+      "step": 16252
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4013071488632793e-05,
+      "loss": 0.6022,
+      "step": 16253
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4012418564682128e-05,
+      "loss": 0.6358,
+      "step": 16254
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4011765620343191e-05,
+      "loss": 0.5897,
+      "step": 16255
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4011112655619297e-05,
+      "loss": 0.6741,
+      "step": 16256
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4010459670513768e-05,
+      "loss": 0.6024,
+      "step": 16257
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4009806665029915e-05,
+      "loss": 0.6118,
+      "step": 16258
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.400915363917106e-05,
+      "loss": 0.5799,
+      "step": 16259
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4008500592940525e-05,
+      "loss": 0.612,
+      "step": 16260
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.400784752634162e-05,
+      "loss": 0.5845,
+      "step": 16261
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4007194439377671e-05,
+      "loss": 0.6012,
+      "step": 16262
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4006541332051993e-05,
+      "loss": 0.6168,
+      "step": 16263
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4005888204367906e-05,
+      "loss": 0.5693,
+      "step": 16264
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4005235056328723e-05,
+      "loss": 0.6363,
+      "step": 16265
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4004581887937774e-05,
+      "loss": 0.551,
+      "step": 16266
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4003928699198367e-05,
+      "loss": 0.6123,
+      "step": 16267
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.400327549011383e-05,
+      "loss": 0.6095,
+      "step": 16268
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4002622260687475e-05,
+      "loss": 0.575,
+      "step": 16269
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4001969010922625e-05,
+      "loss": 0.614,
+      "step": 16270
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4001315740822597e-05,
+      "loss": 0.664,
+      "step": 16271
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4000662450390712e-05,
+      "loss": 0.5769,
+      "step": 16272
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.4000009139630292e-05,
+      "loss": 0.6314,
+      "step": 16273
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3999355808544655e-05,
+      "loss": 0.6493,
+      "step": 16274
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3998702457137116e-05,
+      "loss": 0.6255,
+      "step": 16275
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3998049085411e-05,
+      "loss": 0.6387,
+      "step": 16276
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3997395693369624e-05,
+      "loss": 0.6083,
+      "step": 16277
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3996742281016312e-05,
+      "loss": 0.6313,
+      "step": 16278
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3996088848354383e-05,
+      "loss": 0.6267,
+      "step": 16279
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3995435395387154e-05,
+      "loss": 0.5974,
+      "step": 16280
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3994781922117949e-05,
+      "loss": 0.6151,
+      "step": 16281
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3994128428550088e-05,
+      "loss": 0.5893,
+      "step": 16282
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3993474914686886e-05,
+      "loss": 0.6046,
+      "step": 16283
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3992821380531671e-05,
+      "loss": 0.6205,
+      "step": 16284
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3992167826087763e-05,
+      "loss": 0.5939,
+      "step": 16285
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.399151425135848e-05,
+      "loss": 0.5795,
+      "step": 16286
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3990860656347144e-05,
+      "loss": 0.5833,
+      "step": 16287
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3990207041057077e-05,
+      "loss": 0.6543,
+      "step": 16288
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3989553405491598e-05,
+      "loss": 0.5616,
+      "step": 16289
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3988899749654033e-05,
+      "loss": 0.6307,
+      "step": 16290
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3988246073547695e-05,
+      "loss": 0.5569,
+      "step": 16291
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3987592377175916e-05,
+      "loss": 0.5806,
+      "step": 16292
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3986938660542009e-05,
+      "loss": 0.5907,
+      "step": 16293
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3986284923649298e-05,
+      "loss": 0.5463,
+      "step": 16294
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.398563116650111e-05,
+      "loss": 0.6364,
+      "step": 16295
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.398497738910076e-05,
+      "loss": 0.5999,
+      "step": 16296
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3984323591451573e-05,
+      "loss": 0.5792,
+      "step": 16297
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.398366977355687e-05,
+      "loss": 0.55,
+      "step": 16298
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3983015935419977e-05,
+      "loss": 0.6071,
+      "step": 16299
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3982362077044212e-05,
+      "loss": 0.6362,
+      "step": 16300
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3981708198432898e-05,
+      "loss": 0.6169,
+      "step": 16301
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3981054299589358e-05,
+      "loss": 0.6481,
+      "step": 16302
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3980400380516918e-05,
+      "loss": 0.5979,
+      "step": 16303
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3979746441218898e-05,
+      "loss": 0.6292,
+      "step": 16304
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3979092481698617e-05,
+      "loss": 0.6159,
+      "step": 16305
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3978438501959406e-05,
+      "loss": 0.6199,
+      "step": 16306
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3977784502004581e-05,
+      "loss": 0.6007,
+      "step": 16307
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.397713048183747e-05,
+      "loss": 0.6135,
+      "step": 16308
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3976476441461394e-05,
+      "loss": 0.6064,
+      "step": 16309
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3975822380879674e-05,
+      "loss": 0.5764,
+      "step": 16310
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3975168300095639e-05,
+      "loss": 0.6053,
+      "step": 16311
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3974514199112607e-05,
+      "loss": 0.5605,
+      "step": 16312
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.397386007793391e-05,
+      "loss": 0.603,
+      "step": 16313
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3973205936562862e-05,
+      "loss": 0.6158,
+      "step": 16314
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.397255177500279e-05,
+      "loss": 0.5402,
+      "step": 16315
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.397189759325702e-05,
+      "loss": 0.6044,
+      "step": 16316
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3971243391328876e-05,
+      "loss": 0.5874,
+      "step": 16317
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.397058916922168e-05,
+      "loss": 0.6453,
+      "step": 16318
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3969934926938757e-05,
+      "loss": 0.6047,
+      "step": 16319
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3969280664483436e-05,
+      "loss": 0.6674,
+      "step": 16320
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3968626381859034e-05,
+      "loss": 0.613,
+      "step": 16321
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3967972079068882e-05,
+      "loss": 0.605,
+      "step": 16322
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3967317756116297e-05,
+      "loss": 0.5593,
+      "step": 16323
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.396666341300461e-05,
+      "loss": 0.628,
+      "step": 16324
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3966009049737149e-05,
+      "loss": 0.5955,
+      "step": 16325
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3965354666317227e-05,
+      "loss": 0.5811,
+      "step": 16326
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3964700262748182e-05,
+      "loss": 0.5911,
+      "step": 16327
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3964045839033333e-05,
+      "loss": 0.659,
+      "step": 16328
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3963391395176004e-05,
+      "loss": 0.654,
+      "step": 16329
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3962736931179527e-05,
+      "loss": 0.6015,
+      "step": 16330
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.396208244704722e-05,
+      "loss": 0.5935,
+      "step": 16331
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3961427942782413e-05,
+      "loss": 0.5921,
+      "step": 16332
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.396077341838843e-05,
+      "loss": 0.5814,
+      "step": 16333
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3960118873868597e-05,
+      "loss": 0.6228,
+      "step": 16334
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.395946430922624e-05,
+      "loss": 0.5849,
+      "step": 16335
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3958809724464688e-05,
+      "loss": 0.5897,
+      "step": 16336
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.395815511958726e-05,
+      "loss": 0.5478,
+      "step": 16337
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.395750049459729e-05,
+      "loss": 0.6085,
+      "step": 16338
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3956845849498099e-05,
+      "loss": 0.5851,
+      "step": 16339
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3956191184293015e-05,
+      "loss": 0.6784,
+      "step": 16340
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.395553649898537e-05,
+      "loss": 0.6092,
+      "step": 16341
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3954881793578479e-05,
+      "loss": 0.6795,
+      "step": 16342
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3954227068075677e-05,
+      "loss": 0.5833,
+      "step": 16343
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3953572322480291e-05,
+      "loss": 0.5939,
+      "step": 16344
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3952917556795644e-05,
+      "loss": 0.6274,
+      "step": 16345
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3952262771025067e-05,
+      "loss": 0.5696,
+      "step": 16346
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3951607965171885e-05,
+      "loss": 0.5512,
+      "step": 16347
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3950953139239425e-05,
+      "loss": 0.6324,
+      "step": 16348
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3950298293231018e-05,
+      "loss": 0.5383,
+      "step": 16349
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3949643427149986e-05,
+      "loss": 0.6358,
+      "step": 16350
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3948988540999659e-05,
+      "loss": 0.6408,
+      "step": 16351
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3948333634783364e-05,
+      "loss": 0.5687,
+      "step": 16352
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3947678708504432e-05,
+      "loss": 0.5559,
+      "step": 16353
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3947023762166185e-05,
+      "loss": 0.6126,
+      "step": 16354
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3946368795771958e-05,
+      "loss": 0.5853,
+      "step": 16355
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3945713809325072e-05,
+      "loss": 0.5913,
+      "step": 16356
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3945058802828859e-05,
+      "loss": 0.6204,
+      "step": 16357
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.394440377628665e-05,
+      "loss": 0.5594,
+      "step": 16358
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3943748729701767e-05,
+      "loss": 0.6028,
+      "step": 16359
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3943093663077545e-05,
+      "loss": 0.6112,
+      "step": 16360
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3942438576417306e-05,
+      "loss": 0.5998,
+      "step": 16361
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3941783469724382e-05,
+      "loss": 0.5924,
+      "step": 16362
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3941128343002103e-05,
+      "loss": 0.5722,
+      "step": 16363
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3940473196253796e-05,
+      "loss": 0.6326,
+      "step": 16364
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3939818029482792e-05,
+      "loss": 0.672,
+      "step": 16365
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.393916284269242e-05,
+      "loss": 0.6781,
+      "step": 16366
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3938507635886004e-05,
+      "loss": 0.6587,
+      "step": 16367
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3937852409066879e-05,
+      "loss": 0.6079,
+      "step": 16368
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3937197162238371e-05,
+      "loss": 0.5723,
+      "step": 16369
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3936541895403817e-05,
+      "loss": 0.5607,
+      "step": 16370
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3935886608566535e-05,
+      "loss": 0.6102,
+      "step": 16371
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3935231301729863e-05,
+      "loss": 0.6451,
+      "step": 16372
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3934575974897126e-05,
+      "loss": 0.5793,
+      "step": 16373
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.393392062807166e-05,
+      "loss": 0.6187,
+      "step": 16374
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3933265261256788e-05,
+      "loss": 0.5869,
+      "step": 16375
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3932609874455845e-05,
+      "loss": 0.6536,
+      "step": 16376
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.393195446767216e-05,
+      "loss": 0.6125,
+      "step": 16377
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.393129904090906e-05,
+      "loss": 0.6347,
+      "step": 16378
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3930643594169881e-05,
+      "loss": 0.6522,
+      "step": 16379
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3929988127457948e-05,
+      "loss": 0.6559,
+      "step": 16380
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3929332640776596e-05,
+      "loss": 0.6297,
+      "step": 16381
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3928677134129154e-05,
+      "loss": 0.6445,
+      "step": 16382
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3928021607518954e-05,
+      "loss": 0.5982,
+      "step": 16383
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3927366060949323e-05,
+      "loss": 0.5541,
+      "step": 16384
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3926710494423596e-05,
+      "loss": 0.5845,
+      "step": 16385
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3926054907945104e-05,
+      "loss": 0.5744,
+      "step": 16386
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3925399301517178e-05,
+      "loss": 0.6314,
+      "step": 16387
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3924743675143147e-05,
+      "loss": 0.6085,
+      "step": 16388
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3924088028826345e-05,
+      "loss": 0.5254,
+      "step": 16389
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3923432362570102e-05,
+      "loss": 0.6365,
+      "step": 16390
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.392277667637775e-05,
+      "loss": 0.5773,
+      "step": 16391
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3922120970252619e-05,
+      "loss": 0.5574,
+      "step": 16392
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3921465244198047e-05,
+      "loss": 0.5875,
+      "step": 16393
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3920809498217358e-05,
+      "loss": 0.6135,
+      "step": 16394
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.392015373231389e-05,
+      "loss": 0.5597,
+      "step": 16395
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3919497946490971e-05,
+      "loss": 0.597,
+      "step": 16396
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3918842140751935e-05,
+      "loss": 0.6185,
+      "step": 16397
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3918186315100117e-05,
+      "loss": 0.5825,
+      "step": 16398
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3917530469538844e-05,
+      "loss": 0.6049,
+      "step": 16399
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3916874604071451e-05,
+      "loss": 0.5847,
+      "step": 16400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3916218718701276e-05,
+      "loss": 0.5568,
+      "step": 16401
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.391556281343164e-05,
+      "loss": 0.6081,
+      "step": 16402
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3914906888265884e-05,
+      "loss": 0.6313,
+      "step": 16403
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3914250943207343e-05,
+      "loss": 0.5793,
+      "step": 16404
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3913594978259344e-05,
+      "loss": 0.5795,
+      "step": 16405
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3912938993425227e-05,
+      "loss": 0.6014,
+      "step": 16406
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3912282988708316e-05,
+      "loss": 0.5774,
+      "step": 16407
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.391162696411195e-05,
+      "loss": 0.5584,
+      "step": 16408
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3910970919639464e-05,
+      "loss": 0.6203,
+      "step": 16409
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.391031485529419e-05,
+      "loss": 0.6277,
+      "step": 16410
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3909658771079459e-05,
+      "loss": 0.5539,
+      "step": 16411
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3909002666998607e-05,
+      "loss": 0.628,
+      "step": 16412
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3908346543054967e-05,
+      "loss": 0.5944,
+      "step": 16413
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3907690399251877e-05,
+      "loss": 0.6449,
+      "step": 16414
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3907034235592662e-05,
+      "loss": 0.5685,
+      "step": 16415
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3906378052080664e-05,
+      "loss": 0.6082,
+      "step": 16416
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3905721848719218e-05,
+      "loss": 0.5839,
+      "step": 16417
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3905065625511654e-05,
+      "loss": 0.5394,
+      "step": 16418
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3904409382461307e-05,
+      "loss": 0.6684,
+      "step": 16419
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3903753119571513e-05,
+      "loss": 0.6333,
+      "step": 16420
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3903096836845604e-05,
+      "loss": 0.5651,
+      "step": 16421
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3902440534286918e-05,
+      "loss": 0.6269,
+      "step": 16422
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3901784211898793e-05,
+      "loss": 0.6019,
+      "step": 16423
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3901127869684553e-05,
+      "loss": 0.5191,
+      "step": 16424
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3900471507647546e-05,
+      "loss": 0.6192,
+      "step": 16425
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3899815125791096e-05,
+      "loss": 0.5876,
+      "step": 16426
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3899158724118544e-05,
+      "loss": 0.5437,
+      "step": 16427
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3898502302633228e-05,
+      "loss": 0.5897,
+      "step": 16428
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3897845861338479e-05,
+      "loss": 0.6075,
+      "step": 16429
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3897189400237631e-05,
+      "loss": 0.645,
+      "step": 16430
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3896532919334025e-05,
+      "loss": 0.634,
+      "step": 16431
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3895876418630994e-05,
+      "loss": 0.62,
+      "step": 16432
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3895219898131872e-05,
+      "loss": 0.6239,
+      "step": 16433
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3894563357839998e-05,
+      "loss": 0.6308,
+      "step": 16434
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3893906797758705e-05,
+      "loss": 0.551,
+      "step": 16435
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3893250217891336e-05,
+      "loss": 0.6041,
+      "step": 16436
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3892593618241219e-05,
+      "loss": 0.5806,
+      "step": 16437
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3891936998811692e-05,
+      "loss": 0.6278,
+      "step": 16438
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3891280359606097e-05,
+      "loss": 0.5529,
+      "step": 16439
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3890623700627766e-05,
+      "loss": 0.5914,
+      "step": 16440
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3889967021880036e-05,
+      "loss": 0.7006,
+      "step": 16441
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3889310323366246e-05,
+      "loss": 0.6302,
+      "step": 16442
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3888653605089725e-05,
+      "loss": 0.6306,
+      "step": 16443
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3887996867053823e-05,
+      "loss": 0.5423,
+      "step": 16444
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3887340109261866e-05,
+      "loss": 0.5534,
+      "step": 16445
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3886683331717198e-05,
+      "loss": 0.61,
+      "step": 16446
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3886026534423156e-05,
+      "loss": 0.615,
+      "step": 16447
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.388536971738307e-05,
+      "loss": 0.5587,
+      "step": 16448
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3884712880600286e-05,
+      "loss": 0.5844,
+      "step": 16449
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3884056024078139e-05,
+      "loss": 0.622,
+      "step": 16450
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3883399147819962e-05,
+      "loss": 0.5927,
+      "step": 16451
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.38827422518291e-05,
+      "loss": 0.6667,
+      "step": 16452
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3882085336108885e-05,
+      "loss": 0.6171,
+      "step": 16453
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3881428400662658e-05,
+      "loss": 0.603,
+      "step": 16454
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.388077144549376e-05,
+      "loss": 0.6079,
+      "step": 16455
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3880114470605521e-05,
+      "loss": 0.5778,
+      "step": 16456
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3879457476001285e-05,
+      "loss": 0.6021,
+      "step": 16457
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3878800461684391e-05,
+      "loss": 0.6395,
+      "step": 16458
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3878143427658174e-05,
+      "loss": 0.6042,
+      "step": 16459
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3877486373925977e-05,
+      "loss": 0.6382,
+      "step": 16460
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3876829300491135e-05,
+      "loss": 0.5362,
+      "step": 16461
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3876172207356985e-05,
+      "loss": 0.5593,
+      "step": 16462
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3875515094526875e-05,
+      "loss": 0.5763,
+      "step": 16463
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3874857962004133e-05,
+      "loss": 0.6209,
+      "step": 16464
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3874200809792102e-05,
+      "loss": 0.6286,
+      "step": 16465
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3873543637894127e-05,
+      "loss": 0.5698,
+      "step": 16466
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3872886446313538e-05,
+      "loss": 0.6389,
+      "step": 16467
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3872229235053683e-05,
+      "loss": 0.5809,
+      "step": 16468
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3871572004117894e-05,
+      "loss": 0.633,
+      "step": 16469
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3870914753509515e-05,
+      "loss": 0.5655,
+      "step": 16470
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3870257483231885e-05,
+      "loss": 0.5899,
+      "step": 16471
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3869600193288341e-05,
+      "loss": 0.5727,
+      "step": 16472
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3868942883682227e-05,
+      "loss": 0.5948,
+      "step": 16473
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.386828555441688e-05,
+      "loss": 0.6609,
+      "step": 16474
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3867628205495642e-05,
+      "loss": 0.6661,
+      "step": 16475
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3866970836921851e-05,
+      "loss": 0.5778,
+      "step": 16476
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3866313448698852e-05,
+      "loss": 0.5987,
+      "step": 16477
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3865656040829978e-05,
+      "loss": 0.5923,
+      "step": 16478
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3864998613318576e-05,
+      "loss": 0.6528,
+      "step": 16479
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3864341166167984e-05,
+      "loss": 0.6036,
+      "step": 16480
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3863683699381541e-05,
+      "loss": 0.5586,
+      "step": 16481
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3863026212962591e-05,
+      "loss": 0.5952,
+      "step": 16482
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3862368706914472e-05,
+      "loss": 0.6006,
+      "step": 16483
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3861711181240529e-05,
+      "loss": 0.5844,
+      "step": 16484
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.38610536359441e-05,
+      "loss": 0.5743,
+      "step": 16485
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3860396071028527e-05,
+      "loss": 0.5826,
+      "step": 16486
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.385973848649715e-05,
+      "loss": 0.6623,
+      "step": 16487
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3859080882353311e-05,
+      "loss": 0.6278,
+      "step": 16488
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3858423258600354e-05,
+      "loss": 0.6497,
+      "step": 16489
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3857765615241618e-05,
+      "loss": 0.6251,
+      "step": 16490
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3857107952280442e-05,
+      "loss": 0.664,
+      "step": 16491
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3856450269720171e-05,
+      "loss": 0.6394,
+      "step": 16492
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3855792567564152e-05,
+      "loss": 0.6158,
+      "step": 16493
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3855134845815717e-05,
+      "loss": 0.624,
+      "step": 16494
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3854477104478212e-05,
+      "loss": 0.5653,
+      "step": 16495
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3853819343554984e-05,
+      "loss": 0.5807,
+      "step": 16496
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.385316156304937e-05,
+      "loss": 0.6131,
+      "step": 16497
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3852503762964714e-05,
+      "loss": 0.6004,
+      "step": 16498
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3851845943304355e-05,
+      "loss": 0.628,
+      "step": 16499
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3851188104071641e-05,
+      "loss": 0.6378,
+      "step": 16500
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3850530245269912e-05,
+      "loss": 0.5911,
+      "step": 16501
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3849872366902512e-05,
+      "loss": 0.6066,
+      "step": 16502
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.384921446897278e-05,
+      "loss": 0.6288,
+      "step": 16503
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3848556551484068e-05,
+      "loss": 0.5689,
+      "step": 16504
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3847898614439707e-05,
+      "loss": 0.5728,
+      "step": 16505
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.384724065784305e-05,
+      "loss": 0.5889,
+      "step": 16506
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3846582681697433e-05,
+      "loss": 0.5545,
+      "step": 16507
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3845924686006207e-05,
+      "loss": 0.628,
+      "step": 16508
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.384526667077271e-05,
+      "loss": 0.6417,
+      "step": 16509
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3844608636000287e-05,
+      "loss": 0.6043,
+      "step": 16510
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.384395058169228e-05,
+      "loss": 0.6258,
+      "step": 16511
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3843292507852036e-05,
+      "loss": 0.5569,
+      "step": 16512
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3842634414482896e-05,
+      "loss": 0.588,
+      "step": 16513
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3841976301588205e-05,
+      "loss": 0.63,
+      "step": 16514
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3841318169171309e-05,
+      "loss": 0.6431,
+      "step": 16515
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3840660017235548e-05,
+      "loss": 0.5261,
+      "step": 16516
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.384000184578427e-05,
+      "loss": 0.5549,
+      "step": 16517
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3839343654820818e-05,
+      "loss": 0.5769,
+      "step": 16518
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3838685444348535e-05,
+      "loss": 0.5611,
+      "step": 16519
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.383802721437077e-05,
+      "loss": 0.5726,
+      "step": 16520
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3837368964890863e-05,
+      "loss": 0.5705,
+      "step": 16521
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3836710695912159e-05,
+      "loss": 0.5627,
+      "step": 16522
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3836052407438008e-05,
+      "loss": 0.5963,
+      "step": 16523
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3835394099471748e-05,
+      "loss": 0.5974,
+      "step": 16524
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.383473577201673e-05,
+      "loss": 0.6219,
+      "step": 16525
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3834077425076294e-05,
+      "loss": 0.6119,
+      "step": 16526
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.383341905865379e-05,
+      "loss": 0.596,
+      "step": 16527
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3832760672752559e-05,
+      "loss": 0.5579,
+      "step": 16528
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3832102267375951e-05,
+      "loss": 0.6204,
+      "step": 16529
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3831443842527305e-05,
+      "loss": 0.6219,
+      "step": 16530
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3830785398209976e-05,
+      "loss": 0.5919,
+      "step": 16531
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3830126934427299e-05,
+      "loss": 0.6252,
+      "step": 16532
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.382946845118263e-05,
+      "loss": 0.6035,
+      "step": 16533
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3828809948479309e-05,
+      "loss": 0.6726,
+      "step": 16534
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.382815142632068e-05,
+      "loss": 0.6273,
+      "step": 16535
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3827492884710097e-05,
+      "loss": 0.5793,
+      "step": 16536
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3826834323650899e-05,
+      "loss": 0.5697,
+      "step": 16537
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3826175743146435e-05,
+      "loss": 0.5817,
+      "step": 16538
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3825517143200054e-05,
+      "loss": 0.5624,
+      "step": 16539
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3824858523815099e-05,
+      "loss": 0.6018,
+      "step": 16540
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3824199884994915e-05,
+      "loss": 0.6537,
+      "step": 16541
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3823541226742854e-05,
+      "loss": 0.6487,
+      "step": 16542
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3822882549062259e-05,
+      "loss": 0.6221,
+      "step": 16543
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.382222385195648e-05,
+      "loss": 0.5915,
+      "step": 16544
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3821565135428863e-05,
+      "loss": 0.6144,
+      "step": 16545
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3820906399482752e-05,
+      "loss": 0.599,
+      "step": 16546
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3820247644121496e-05,
+      "loss": 0.5594,
+      "step": 16547
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3819588869348446e-05,
+      "loss": 0.59,
+      "step": 16548
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3818930075166943e-05,
+      "loss": 0.5583,
+      "step": 16549
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.381827126158034e-05,
+      "loss": 0.645,
+      "step": 16550
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3817612428591982e-05,
+      "loss": 0.6251,
+      "step": 16551
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.381695357620522e-05,
+      "loss": 0.5565,
+      "step": 16552
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3816294704423394e-05,
+      "loss": 0.6218,
+      "step": 16553
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3815635813249857e-05,
+      "loss": 0.5961,
+      "step": 16554
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.381497690268796e-05,
+      "loss": 0.6472,
+      "step": 16555
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3814317972741047e-05,
+      "loss": 0.629,
+      "step": 16556
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3813659023412467e-05,
+      "loss": 0.5923,
+      "step": 16557
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.381300005470557e-05,
+      "loss": 0.6025,
+      "step": 16558
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.38123410666237e-05,
+      "loss": 0.5849,
+      "step": 16559
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3811682059170213e-05,
+      "loss": 0.6207,
+      "step": 16560
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3811023032348449e-05,
+      "loss": 0.6198,
+      "step": 16561
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.381036398616176e-05,
+      "loss": 0.6019,
+      "step": 16562
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.38097049206135e-05,
+      "loss": 0.599,
+      "step": 16563
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3809045835707013e-05,
+      "loss": 0.5672,
+      "step": 16564
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3808386731445649e-05,
+      "loss": 0.5939,
+      "step": 16565
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3807727607832753e-05,
+      "loss": 0.6009,
+      "step": 16566
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.380706846487168e-05,
+      "loss": 0.6281,
+      "step": 16567
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3806409302565778e-05,
+      "loss": 0.5757,
+      "step": 16568
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3805750120918396e-05,
+      "loss": 0.6032,
+      "step": 16569
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.380509091993288e-05,
+      "loss": 0.5702,
+      "step": 16570
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3804431699612586e-05,
+      "loss": 0.6677,
+      "step": 16571
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3803772459960857e-05,
+      "loss": 0.5727,
+      "step": 16572
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.380311320098105e-05,
+      "loss": 0.5387,
+      "step": 16573
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3802453922676508e-05,
+      "loss": 0.6231,
+      "step": 16574
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3801794625050585e-05,
+      "loss": 0.5933,
+      "step": 16575
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3801135308106628e-05,
+      "loss": 0.5406,
+      "step": 16576
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3800475971847993e-05,
+      "loss": 0.6088,
+      "step": 16577
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3799816616278023e-05,
+      "loss": 0.6282,
+      "step": 16578
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3799157241400076e-05,
+      "loss": 0.5172,
+      "step": 16579
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3798497847217496e-05,
+      "loss": 0.5809,
+      "step": 16580
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3797838433733635e-05,
+      "loss": 0.6389,
+      "step": 16581
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3797179000951848e-05,
+      "loss": 0.6387,
+      "step": 16582
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3796519548875477e-05,
+      "loss": 0.6494,
+      "step": 16583
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3795860077507884e-05,
+      "loss": 0.6146,
+      "step": 16584
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3795200586852411e-05,
+      "loss": 0.6097,
+      "step": 16585
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3794541076912415e-05,
+      "loss": 0.5323,
+      "step": 16586
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3793881547691244e-05,
+      "loss": 0.6045,
+      "step": 16587
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3793221999192248e-05,
+      "loss": 0.6279,
+      "step": 16588
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3792562431418781e-05,
+      "loss": 0.5684,
+      "step": 16589
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3791902844374194e-05,
+      "loss": 0.6189,
+      "step": 16590
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3791243238061836e-05,
+      "loss": 0.6009,
+      "step": 16591
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3790583612485062e-05,
+      "loss": 0.6411,
+      "step": 16592
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3789923967647223e-05,
+      "loss": 0.5277,
+      "step": 16593
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378926430355167e-05,
+      "loss": 0.6704,
+      "step": 16594
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3788604620201755e-05,
+      "loss": 0.629,
+      "step": 16595
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378794491760083e-05,
+      "loss": 0.5775,
+      "step": 16596
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378728519575225e-05,
+      "loss": 0.5802,
+      "step": 16597
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3786625454659364e-05,
+      "loss": 0.6227,
+      "step": 16598
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3785965694325523e-05,
+      "loss": 0.592,
+      "step": 16599
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3785305914754082e-05,
+      "loss": 0.5864,
+      "step": 16600
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3784646115948395e-05,
+      "loss": 0.5911,
+      "step": 16601
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3783986297911812e-05,
+      "loss": 0.6115,
+      "step": 16602
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3783326460647684e-05,
+      "loss": 0.5764,
+      "step": 16603
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378266660415937e-05,
+      "loss": 0.5831,
+      "step": 16604
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3782006728450216e-05,
+      "loss": 0.5829,
+      "step": 16605
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378134683352358e-05,
+      "loss": 0.5743,
+      "step": 16606
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3780686919382815e-05,
+      "loss": 0.6407,
+      "step": 16607
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.378002698603127e-05,
+      "loss": 0.6562,
+      "step": 16608
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3779367033472301e-05,
+      "loss": 0.5918,
+      "step": 16609
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3778707061709262e-05,
+      "loss": 0.5823,
+      "step": 16610
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3778047070745504e-05,
+      "loss": 0.6025,
+      "step": 16611
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3777387060584386e-05,
+      "loss": 0.6361,
+      "step": 16612
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3776727031229253e-05,
+      "loss": 0.5361,
+      "step": 16613
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3776066982683467e-05,
+      "loss": 0.6009,
+      "step": 16614
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.377540691495038e-05,
+      "loss": 0.6182,
+      "step": 16615
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3774746828033342e-05,
+      "loss": 0.5812,
+      "step": 16616
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3774086721935713e-05,
+      "loss": 0.6388,
+      "step": 16617
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3773426596660842e-05,
+      "loss": 0.5812,
+      "step": 16618
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3772766452212083e-05,
+      "loss": 0.5414,
+      "step": 16619
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3772106288592797e-05,
+      "loss": 0.5887,
+      "step": 16620
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3771446105806332e-05,
+      "loss": 0.6325,
+      "step": 16621
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3770785903856045e-05,
+      "loss": 0.6127,
+      "step": 16622
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3770125682745292e-05,
+      "loss": 0.6097,
+      "step": 16623
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3769465442477425e-05,
+      "loss": 0.5612,
+      "step": 16624
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3768805183055802e-05,
+      "loss": 0.583,
+      "step": 16625
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3768144904483775e-05,
+      "loss": 0.6069,
+      "step": 16626
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3767484606764699e-05,
+      "loss": 0.606,
+      "step": 16627
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3766824289901933e-05,
+      "loss": 0.585,
+      "step": 16628
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3766163953898826e-05,
+      "loss": 0.6428,
+      "step": 16629
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3765503598758738e-05,
+      "loss": 0.5855,
+      "step": 16630
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3764843224485025e-05,
+      "loss": 0.6219,
+      "step": 16631
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3764182831081039e-05,
+      "loss": 0.5803,
+      "step": 16632
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3763522418550136e-05,
+      "loss": 0.5915,
+      "step": 16633
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.376286198689568e-05,
+      "loss": 0.6182,
+      "step": 16634
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3762201536121015e-05,
+      "loss": 0.572,
+      "step": 16635
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3761541066229505e-05,
+      "loss": 0.6491,
+      "step": 16636
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3760880577224501e-05,
+      "loss": 0.5923,
+      "step": 16637
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.376022006910936e-05,
+      "loss": 0.535,
+      "step": 16638
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3759559541887443e-05,
+      "loss": 0.5596,
+      "step": 16639
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.37588989955621e-05,
+      "loss": 0.603,
+      "step": 16640
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3758238430136692e-05,
+      "loss": 0.5573,
+      "step": 16641
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3757577845614576e-05,
+      "loss": 0.5911,
+      "step": 16642
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3756917241999102e-05,
+      "loss": 0.6283,
+      "step": 16643
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3756256619293635e-05,
+      "loss": 0.6053,
+      "step": 16644
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3755595977501526e-05,
+      "loss": 0.6199,
+      "step": 16645
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3754935316626137e-05,
+      "loss": 0.5949,
+      "step": 16646
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.375427463667082e-05,
+      "loss": 0.5374,
+      "step": 16647
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3753613937638932e-05,
+      "loss": 0.57,
+      "step": 16648
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3752953219533833e-05,
+      "loss": 0.5819,
+      "step": 16649
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3752292482358882e-05,
+      "loss": 0.6207,
+      "step": 16650
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.375163172611743e-05,
+      "loss": 0.6386,
+      "step": 16651
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3750970950812846e-05,
+      "loss": 0.6279,
+      "step": 16652
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3750310156448471e-05,
+      "loss": 0.5273,
+      "step": 16653
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3749649343027676e-05,
+      "loss": 0.6063,
+      "step": 16654
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3748988510553816e-05,
+      "loss": 0.6044,
+      "step": 16655
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3748327659030245e-05,
+      "loss": 0.6439,
+      "step": 16656
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3747666788460323e-05,
+      "loss": 0.5965,
+      "step": 16657
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3747005898847412e-05,
+      "loss": 0.5301,
+      "step": 16658
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3746344990194862e-05,
+      "loss": 0.5666,
+      "step": 16659
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.374568406250604e-05,
+      "loss": 0.6083,
+      "step": 16660
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3745023115784299e-05,
+      "loss": 0.6135,
+      "step": 16661
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3744362150032997e-05,
+      "loss": 0.6301,
+      "step": 16662
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3743701165255496e-05,
+      "loss": 0.6367,
+      "step": 16663
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3743040161455152e-05,
+      "loss": 0.5869,
+      "step": 16664
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.374237913863533e-05,
+      "loss": 0.6251,
+      "step": 16665
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.374171809679938e-05,
+      "loss": 0.6089,
+      "step": 16666
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.374105703595066e-05,
+      "loss": 0.5922,
+      "step": 16667
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3740395956092538e-05,
+      "loss": 0.6209,
+      "step": 16668
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3739734857228369e-05,
+      "loss": 0.6019,
+      "step": 16669
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3739073739361509e-05,
+      "loss": 0.6191,
+      "step": 16670
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3738412602495323e-05,
+      "loss": 0.6309,
+      "step": 16671
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3737751446633167e-05,
+      "loss": 0.6222,
+      "step": 16672
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3737090271778399e-05,
+      "loss": 0.5703,
+      "step": 16673
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3736429077934386e-05,
+      "loss": 0.614,
+      "step": 16674
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3735767865104476e-05,
+      "loss": 0.6019,
+      "step": 16675
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.373510663329204e-05,
+      "loss": 0.6436,
+      "step": 16676
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.3734445382500435e-05,
+      "loss": 0.5819,
+      "step": 16677
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3733784112733014e-05,
+      "loss": 0.5663,
+      "step": 16678
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3733122823993144e-05,
+      "loss": 0.5806,
+      "step": 16679
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3732461516284185e-05,
+      "loss": 0.6058,
+      "step": 16680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3731800189609497e-05,
+      "loss": 0.5843,
+      "step": 16681
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3731138843972438e-05,
+      "loss": 0.6258,
+      "step": 16682
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.373047747937637e-05,
+      "loss": 0.5887,
+      "step": 16683
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3729816095824653e-05,
+      "loss": 0.6104,
+      "step": 16684
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3729154693320648e-05,
+      "loss": 0.6239,
+      "step": 16685
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3728493271867718e-05,
+      "loss": 0.5903,
+      "step": 16686
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3727831831469219e-05,
+      "loss": 0.6209,
+      "step": 16687
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3727170372128515e-05,
+      "loss": 0.5929,
+      "step": 16688
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3726508893848966e-05,
+      "loss": 0.5228,
+      "step": 16689
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3725847396633938e-05,
+      "loss": 0.5128,
+      "step": 16690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3725185880486784e-05,
+      "loss": 0.6366,
+      "step": 16691
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3724524345410872e-05,
+      "loss": 0.5957,
+      "step": 16692
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3723862791409561e-05,
+      "loss": 0.6153,
+      "step": 16693
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3723201218486211e-05,
+      "loss": 0.5867,
+      "step": 16694
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3722539626644185e-05,
+      "loss": 0.5795,
+      "step": 16695
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3721878015886846e-05,
+      "loss": 0.5625,
+      "step": 16696
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3721216386217554e-05,
+      "loss": 0.5832,
+      "step": 16697
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3720554737639673e-05,
+      "loss": 0.603,
+      "step": 16698
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3719893070156562e-05,
+      "loss": 0.5638,
+      "step": 16699
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3719231383771588e-05,
+      "loss": 0.6604,
+      "step": 16700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3718569678488107e-05,
+      "loss": 0.6048,
+      "step": 16701
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3717907954309486e-05,
+      "loss": 0.5819,
+      "step": 16702
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3717246211239086e-05,
+      "loss": 0.526,
+      "step": 16703
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3716584449280268e-05,
+      "loss": 0.6008,
+      "step": 16704
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3715922668436398e-05,
+      "loss": 0.5621,
+      "step": 16705
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3715260868710836e-05,
+      "loss": 0.599,
+      "step": 16706
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3714599050106945e-05,
+      "loss": 0.6155,
+      "step": 16707
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3713937212628088e-05,
+      "loss": 0.5844,
+      "step": 16708
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3713275356277629e-05,
+      "loss": 0.5773,
+      "step": 16709
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3712613481058929e-05,
+      "loss": 0.6754,
+      "step": 16710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3711951586975355e-05,
+      "loss": 0.6113,
+      "step": 16711
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3711289674030266e-05,
+      "loss": 0.5969,
+      "step": 16712
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3710627742227027e-05,
+      "loss": 0.6214,
+      "step": 16713
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3709965791569001e-05,
+      "loss": 0.5919,
+      "step": 16714
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3709303822059557e-05,
+      "loss": 0.5326,
+      "step": 16715
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.370864183370205e-05,
+      "loss": 0.6319,
+      "step": 16716
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3707979826499848e-05,
+      "loss": 0.5397,
+      "step": 16717
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3707317800456315e-05,
+      "loss": 0.5801,
+      "step": 16718
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3706655755574814e-05,
+      "loss": 0.5926,
+      "step": 16719
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3705993691858712e-05,
+      "loss": 0.5148,
+      "step": 16720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3705331609311367e-05,
+      "loss": 0.5974,
+      "step": 16721
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3704669507936151e-05,
+      "loss": 0.6,
+      "step": 16722
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.370400738773642e-05,
+      "loss": 0.6145,
+      "step": 16723
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3703345248715548e-05,
+      "loss": 0.5781,
+      "step": 16724
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.370268309087689e-05,
+      "loss": 0.5547,
+      "step": 16725
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3702020914223815e-05,
+      "loss": 0.6379,
+      "step": 16726
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3701358718759688e-05,
+      "loss": 0.5471,
+      "step": 16727
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3700696504487873e-05,
+      "loss": 0.6047,
+      "step": 16728
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3700034271411737e-05,
+      "loss": 0.5854,
+      "step": 16729
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.369937201953464e-05,
+      "loss": 0.5707,
+      "step": 16730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3698709748859954e-05,
+      "loss": 0.6074,
+      "step": 16731
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.369804745939104e-05,
+      "loss": 0.6022,
+      "step": 16732
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3697385151131262e-05,
+      "loss": 0.6301,
+      "step": 16733
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.369672282408399e-05,
+      "loss": 0.5491,
+      "step": 16734
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3696060478252584e-05,
+      "loss": 0.6055,
+      "step": 16735
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3695398113640414e-05,
+      "loss": 0.595,
+      "step": 16736
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3694735730250841e-05,
+      "loss": 0.6055,
+      "step": 16737
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3694073328087236e-05,
+      "loss": 0.5851,
+      "step": 16738
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3693410907152965e-05,
+      "loss": 0.5886,
+      "step": 16739
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3692748467451386e-05,
+      "loss": 0.5564,
+      "step": 16740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3692086008985875e-05,
+      "loss": 0.5642,
+      "step": 16741
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3691423531759792e-05,
+      "loss": 0.5255,
+      "step": 16742
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3690761035776506e-05,
+      "loss": 0.5918,
+      "step": 16743
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.369009852103938e-05,
+      "loss": 0.5944,
+      "step": 16744
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3689435987551787e-05,
+      "loss": 0.6083,
+      "step": 16745
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3688773435317084e-05,
+      "loss": 0.6157,
+      "step": 16746
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3688110864338646e-05,
+      "loss": 0.5913,
+      "step": 16747
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3687448274619835e-05,
+      "loss": 0.5946,
+      "step": 16748
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.368678566616402e-05,
+      "loss": 0.5921,
+      "step": 16749
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3686123038974568e-05,
+      "loss": 0.6108,
+      "step": 16750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3685460393054847e-05,
+      "loss": 0.535,
+      "step": 16751
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3684797728408217e-05,
+      "loss": 0.6042,
+      "step": 16752
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3684135045038054e-05,
+      "loss": 0.6846,
+      "step": 16753
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3683472342947722e-05,
+      "loss": 0.6106,
+      "step": 16754
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3682809622140589e-05,
+      "loss": 0.5607,
+      "step": 16755
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.368214688262002e-05,
+      "loss": 0.62,
+      "step": 16756
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3681484124389384e-05,
+      "loss": 0.6145,
+      "step": 16757
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3680821347452051e-05,
+      "loss": 0.6661,
+      "step": 16758
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3680158551811387e-05,
+      "loss": 0.5674,
+      "step": 16759
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3679495737470757e-05,
+      "loss": 0.6607,
+      "step": 16760
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3678832904433534e-05,
+      "loss": 0.6063,
+      "step": 16761
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3678170052703083e-05,
+      "loss": 0.621,
+      "step": 16762
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3677507182282773e-05,
+      "loss": 0.5717,
+      "step": 16763
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3676844293175973e-05,
+      "loss": 0.5824,
+      "step": 16764
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3676181385386046e-05,
+      "loss": 0.5899,
+      "step": 16765
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3675518458916368e-05,
+      "loss": 0.6395,
+      "step": 16766
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3674855513770304e-05,
+      "loss": 0.6572,
+      "step": 16767
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3674192549951222e-05,
+      "loss": 0.6367,
+      "step": 16768
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3673529567462494e-05,
+      "loss": 0.6564,
+      "step": 16769
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3672866566307483e-05,
+      "loss": 0.5837,
+      "step": 16770
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3672203546489561e-05,
+      "loss": 0.5706,
+      "step": 16771
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3671540508012102e-05,
+      "loss": 0.6089,
+      "step": 16772
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3670877450878465e-05,
+      "loss": 0.5881,
+      "step": 16773
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.367021437509203e-05,
+      "loss": 0.6233,
+      "step": 16774
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3669551280656157e-05,
+      "loss": 0.6308,
+      "step": 16775
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3668888167574221e-05,
+      "loss": 0.6382,
+      "step": 16776
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.366822503584959e-05,
+      "loss": 0.6104,
+      "step": 16777
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3667561885485632e-05,
+      "loss": 0.5927,
+      "step": 16778
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3666898716485716e-05,
+      "loss": 0.6301,
+      "step": 16779
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3666235528853217e-05,
+      "loss": 0.6517,
+      "step": 16780
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.36655723225915e-05,
+      "loss": 0.6124,
+      "step": 16781
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.366490909770394e-05,
+      "loss": 0.6108,
+      "step": 16782
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3664245854193903e-05,
+      "loss": 0.5752,
+      "step": 16783
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3663582592064754e-05,
+      "loss": 0.6436,
+      "step": 16784
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3662919311319875e-05,
+      "loss": 0.6305,
+      "step": 16785
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3662256011962627e-05,
+      "loss": 0.5549,
+      "step": 16786
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3661592693996384e-05,
+      "loss": 0.6291,
+      "step": 16787
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3660929357424516e-05,
+      "loss": 0.6319,
+      "step": 16788
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3660266002250394e-05,
+      "loss": 0.6218,
+      "step": 16789
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.365960262847739e-05,
+      "loss": 0.6403,
+      "step": 16790
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3658939236108873e-05,
+      "loss": 0.5393,
+      "step": 16791
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3658275825148212e-05,
+      "loss": 0.5531,
+      "step": 16792
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3657612395598783e-05,
+      "loss": 0.5738,
+      "step": 16793
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3656948947463952e-05,
+      "loss": 0.6502,
+      "step": 16794
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.365628548074709e-05,
+      "loss": 0.5995,
+      "step": 16795
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3655621995451576e-05,
+      "loss": 0.5705,
+      "step": 16796
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3654958491580774e-05,
+      "loss": 0.6093,
+      "step": 16797
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3654294969138055e-05,
+      "loss": 0.5701,
+      "step": 16798
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3653631428126798e-05,
+      "loss": 0.5873,
+      "step": 16799
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3652967868550367e-05,
+      "loss": 0.619,
+      "step": 16800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3652304290412136e-05,
+      "loss": 0.6324,
+      "step": 16801
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.365164069371548e-05,
+      "loss": 0.5953,
+      "step": 16802
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3650977078463764e-05,
+      "loss": 0.5637,
+      "step": 16803
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3650313444660369e-05,
+      "loss": 0.5809,
+      "step": 16804
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3649649792308659e-05,
+      "loss": 0.5757,
+      "step": 16805
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.364898612141201e-05,
+      "loss": 0.5602,
+      "step": 16806
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3648322431973796e-05,
+      "loss": 0.5431,
+      "step": 16807
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3647658723997381e-05,
+      "loss": 0.5623,
+      "step": 16808
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.364699499748615e-05,
+      "loss": 0.6025,
+      "step": 16809
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3646331252443465e-05,
+      "loss": 0.6349,
+      "step": 16810
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3645667488872708e-05,
+      "loss": 0.5511,
+      "step": 16811
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3645003706777243e-05,
+      "loss": 0.5828,
+      "step": 16812
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3644339906160449e-05,
+      "loss": 0.5895,
+      "step": 16813
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3643676087025695e-05,
+      "loss": 0.5858,
+      "step": 16814
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3643012249376355e-05,
+      "loss": 0.4794,
+      "step": 16815
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3642348393215806e-05,
+      "loss": 0.6242,
+      "step": 16816
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3641684518547414e-05,
+      "loss": 0.5791,
+      "step": 16817
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3641020625374559e-05,
+      "loss": 0.6281,
+      "step": 16818
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.364035671370061e-05,
+      "loss": 0.6021,
+      "step": 16819
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3639692783528946e-05,
+      "loss": 0.5537,
+      "step": 16820
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3639028834862935e-05,
+      "loss": 0.6099,
+      "step": 16821
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3638364867705953e-05,
+      "loss": 0.587,
+      "step": 16822
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.3637700882061373e-05,
+      "loss": 0.5859,
+      "step": 16823
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.363703687793257e-05,
+      "loss": 0.6569,
+      "step": 16824
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3636372855322917e-05,
+      "loss": 0.6154,
+      "step": 16825
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3635708814235788e-05,
+      "loss": 0.6016,
+      "step": 16826
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.363504475467456e-05,
+      "loss": 0.606,
+      "step": 16827
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3634380676642605e-05,
+      "loss": 0.5793,
+      "step": 16828
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3633716580143293e-05,
+      "loss": 0.6032,
+      "step": 16829
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3633052465180005e-05,
+      "loss": 0.6165,
+      "step": 16830
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3632388331756116e-05,
+      "loss": 0.5874,
+      "step": 16831
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3631724179874994e-05,
+      "loss": 0.5444,
+      "step": 16832
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.363106000954002e-05,
+      "loss": 0.6566,
+      "step": 16833
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3630395820754569e-05,
+      "loss": 0.5877,
+      "step": 16834
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.362973161352201e-05,
+      "loss": 0.6,
+      "step": 16835
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3629067387845725e-05,
+      "loss": 0.6324,
+      "step": 16836
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3628403143729084e-05,
+      "loss": 0.6731,
+      "step": 16837
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.362773888117546e-05,
+      "loss": 0.5953,
+      "step": 16838
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.362707460018824e-05,
+      "loss": 0.5887,
+      "step": 16839
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3626410300770787e-05,
+      "loss": 0.5985,
+      "step": 16840
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3625745982926484e-05,
+      "loss": 0.6183,
+      "step": 16841
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.36250816466587e-05,
+      "loss": 0.5709,
+      "step": 16842
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3624417291970818e-05,
+      "loss": 0.6747,
+      "step": 16843
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3623752918866209e-05,
+      "loss": 0.6245,
+      "step": 16844
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.362308852734825e-05,
+      "loss": 0.6121,
+      "step": 16845
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3622424117420316e-05,
+      "loss": 0.6166,
+      "step": 16846
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3621759689085785e-05,
+      "loss": 0.6296,
+      "step": 16847
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3621095242348033e-05,
+      "loss": 0.5945,
+      "step": 16848
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3620430777210434e-05,
+      "loss": 0.6605,
+      "step": 16849
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3619766293676369e-05,
+      "loss": 0.6241,
+      "step": 16850
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3619101791749208e-05,
+      "loss": 0.5747,
+      "step": 16851
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3618437271432333e-05,
+      "loss": 0.6147,
+      "step": 16852
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3617772732729118e-05,
+      "loss": 0.6108,
+      "step": 16853
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.361710817564294e-05,
+      "loss": 0.58,
+      "step": 16854
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3616443600177177e-05,
+      "loss": 0.5843,
+      "step": 16855
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3615779006335204e-05,
+      "loss": 0.5884,
+      "step": 16856
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3615114394120396e-05,
+      "loss": 0.6084,
+      "step": 16857
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3614449763536138e-05,
+      "loss": 0.6176,
+      "step": 16858
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3613785114585798e-05,
+      "loss": 0.5676,
+      "step": 16859
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3613120447272761e-05,
+      "loss": 0.617,
+      "step": 16860
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.36124557616004e-05,
+      "loss": 0.6117,
+      "step": 16861
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3611791057572092e-05,
+      "loss": 0.6415,
+      "step": 16862
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3611126335191219e-05,
+      "loss": 0.6315,
+      "step": 16863
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3610461594461152e-05,
+      "loss": 0.5746,
+      "step": 16864
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3609796835385274e-05,
+      "loss": 0.5703,
+      "step": 16865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3609132057966963e-05,
+      "loss": 0.6363,
+      "step": 16866
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3608467262209591e-05,
+      "loss": 0.6177,
+      "step": 16867
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.360780244811654e-05,
+      "loss": 0.6329,
+      "step": 16868
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.360713761569119e-05,
+      "loss": 0.5871,
+      "step": 16869
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3606472764936918e-05,
+      "loss": 0.6162,
+      "step": 16870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3605807895857099e-05,
+      "loss": 0.576,
+      "step": 16871
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3605143008455119e-05,
+      "loss": 0.5656,
+      "step": 16872
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3604478102734346e-05,
+      "loss": 0.6592,
+      "step": 16873
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3603813178698169e-05,
+      "loss": 0.6305,
+      "step": 16874
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.360314823634996e-05,
+      "loss": 0.5971,
+      "step": 16875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3602483275693098e-05,
+      "loss": 0.6463,
+      "step": 16876
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3601818296730964e-05,
+      "loss": 0.5866,
+      "step": 16877
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3601153299466936e-05,
+      "loss": 0.6246,
+      "step": 16878
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3600488283904395e-05,
+      "loss": 0.6314,
+      "step": 16879
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3599823250046717e-05,
+      "loss": 0.6133,
+      "step": 16880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3599158197897287e-05,
+      "loss": 0.6118,
+      "step": 16881
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3598493127459476e-05,
+      "loss": 0.608,
+      "step": 16882
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3597828038736669e-05,
+      "loss": 0.6458,
+      "step": 16883
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3597162931732243e-05,
+      "loss": 0.5813,
+      "step": 16884
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.359649780644958e-05,
+      "loss": 0.5715,
+      "step": 16885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3595832662892057e-05,
+      "loss": 0.6661,
+      "step": 16886
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3595167501063053e-05,
+      "loss": 0.5558,
+      "step": 16887
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3594502320965956e-05,
+      "loss": 0.6508,
+      "step": 16888
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3593837122604136e-05,
+      "loss": 0.5319,
+      "step": 16889
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3593171905980978e-05,
+      "loss": 0.5853,
+      "step": 16890
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3592506671099861e-05,
+      "loss": 0.6517,
+      "step": 16891
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3591841417964164e-05,
+      "loss": 0.6293,
+      "step": 16892
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3591176146577272e-05,
+      "loss": 0.5951,
+      "step": 16893
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.359051085694256e-05,
+      "loss": 0.6434,
+      "step": 16894
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358984554906341e-05,
+      "loss": 0.6002,
+      "step": 16895
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3589180222943207e-05,
+      "loss": 0.5655,
+      "step": 16896
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3588514878585323e-05,
+      "loss": 0.6033,
+      "step": 16897
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358784951599315e-05,
+      "loss": 0.6057,
+      "step": 16898
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358718413517006e-05,
+      "loss": 0.5862,
+      "step": 16899
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3586518736119436e-05,
+      "loss": 0.6603,
+      "step": 16900
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358585331884466e-05,
+      "loss": 0.6024,
+      "step": 16901
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3585187883349115e-05,
+      "loss": 0.5447,
+      "step": 16902
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358452242963618e-05,
+      "loss": 0.6541,
+      "step": 16903
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3583856957709237e-05,
+      "loss": 0.6071,
+      "step": 16904
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3583191467571663e-05,
+      "loss": 0.6084,
+      "step": 16905
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.358252595922685e-05,
+      "loss": 0.659,
+      "step": 16906
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3581860432678168e-05,
+      "loss": 0.6166,
+      "step": 16907
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3581194887929007e-05,
+      "loss": 0.6203,
+      "step": 16908
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3580529324982744e-05,
+      "loss": 0.601,
+      "step": 16909
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3579863743842764e-05,
+      "loss": 0.6293,
+      "step": 16910
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3579198144512449e-05,
+      "loss": 0.5749,
+      "step": 16911
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.357853252699518e-05,
+      "loss": 0.6409,
+      "step": 16912
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3577866891294338e-05,
+      "loss": 0.5682,
+      "step": 16913
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3577201237413307e-05,
+      "loss": 0.5731,
+      "step": 16914
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3576535565355468e-05,
+      "loss": 0.6508,
+      "step": 16915
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3575869875124206e-05,
+      "loss": 0.6074,
+      "step": 16916
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3575204166722899e-05,
+      "loss": 0.6082,
+      "step": 16917
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3574538440154936e-05,
+      "loss": 0.6176,
+      "step": 16918
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3573872695423698e-05,
+      "loss": 0.5791,
+      "step": 16919
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3573206932532562e-05,
+      "loss": 0.6423,
+      "step": 16920
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3572541151484917e-05,
+      "loss": 0.6164,
+      "step": 16921
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3571875352284143e-05,
+      "loss": 0.6005,
+      "step": 16922
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3571209534933625e-05,
+      "loss": 0.565,
+      "step": 16923
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3570543699436744e-05,
+      "loss": 0.6191,
+      "step": 16924
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3569877845796886e-05,
+      "loss": 0.5489,
+      "step": 16925
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3569211974017434e-05,
+      "loss": 0.5985,
+      "step": 16926
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3568546084101767e-05,
+      "loss": 0.5848,
+      "step": 16927
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3567880176053279e-05,
+      "loss": 0.5973,
+      "step": 16928
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3567214249875342e-05,
+      "loss": 0.6367,
+      "step": 16929
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3566548305571345e-05,
+      "loss": 0.5883,
+      "step": 16930
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3565882343144674e-05,
+      "loss": 0.5759,
+      "step": 16931
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3565216362598705e-05,
+      "loss": 0.5427,
+      "step": 16932
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3564550363936832e-05,
+      "loss": 0.5943,
+      "step": 16933
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3563884347162435e-05,
+      "loss": 0.6254,
+      "step": 16934
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3563218312278896e-05,
+      "loss": 0.6019,
+      "step": 16935
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.35625522592896e-05,
+      "loss": 0.576,
+      "step": 16936
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3561886188197937e-05,
+      "loss": 0.5859,
+      "step": 16937
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3561220099007284e-05,
+      "loss": 0.5627,
+      "step": 16938
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.356055399172103e-05,
+      "loss": 0.6102,
+      "step": 16939
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.355988786634256e-05,
+      "loss": 0.5777,
+      "step": 16940
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3559221722875256e-05,
+      "loss": 0.5567,
+      "step": 16941
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3558555561322501e-05,
+      "loss": 0.5773,
+      "step": 16942
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3557889381687685e-05,
+      "loss": 0.6089,
+      "step": 16943
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3557223183974193e-05,
+      "loss": 0.6246,
+      "step": 16944
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3556556968185406e-05,
+      "loss": 0.5974,
+      "step": 16945
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3555890734324713e-05,
+      "loss": 0.6179,
+      "step": 16946
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3555224482395497e-05,
+      "loss": 0.5577,
+      "step": 16947
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3554558212401144e-05,
+      "loss": 0.6234,
+      "step": 16948
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3553891924345037e-05,
+      "loss": 0.5822,
+      "step": 16949
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3553225618230567e-05,
+      "loss": 0.5951,
+      "step": 16950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3552559294061119e-05,
+      "loss": 0.6415,
+      "step": 16951
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3551892951840074e-05,
+      "loss": 0.6234,
+      "step": 16952
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.355122659157082e-05,
+      "loss": 0.6171,
+      "step": 16953
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3550560213256747e-05,
+      "loss": 0.5816,
+      "step": 16954
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3549893816901234e-05,
+      "loss": 0.6071,
+      "step": 16955
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3549227402507674e-05,
+      "loss": 0.6051,
+      "step": 16956
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3548560970079448e-05,
+      "loss": 0.6226,
+      "step": 16957
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3547894519619949e-05,
+      "loss": 0.6655,
+      "step": 16958
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3547228051132555e-05,
+      "loss": 0.5806,
+      "step": 16959
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3546561564620655e-05,
+      "loss": 0.5419,
+      "step": 16960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3545895060087641e-05,
+      "loss": 0.5543,
+      "step": 16961
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3545228537536892e-05,
+      "loss": 0.5854,
+      "step": 16962
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.35445619969718e-05,
+      "loss": 0.5976,
+      "step": 16963
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3543895438395751e-05,
+      "loss": 0.5667,
+      "step": 16964
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.354322886181213e-05,
+      "loss": 0.605,
+      "step": 16965
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3542562267224328e-05,
+      "loss": 0.5526,
+      "step": 16966
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3541895654635727e-05,
+      "loss": 0.5426,
+      "step": 16967
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3541229024049716e-05,
+      "loss": 0.6009,
+      "step": 16968
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3540562375469688e-05,
+      "loss": 0.5712,
+      "step": 16969
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.353989570889902e-05,
+      "loss": 0.5954,
+      "step": 16970
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.3539229024341108e-05,
+      "loss": 0.5917,
+      "step": 16971
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3538562321799339e-05,
+      "loss": 0.5944,
+      "step": 16972
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3537895601277096e-05,
+      "loss": 0.6053,
+      "step": 16973
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.353722886277777e-05,
+      "loss": 0.6267,
+      "step": 16974
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3536562106304747e-05,
+      "loss": 0.568,
+      "step": 16975
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3535895331861415e-05,
+      "loss": 0.6032,
+      "step": 16976
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3535228539451166e-05,
+      "loss": 0.5807,
+      "step": 16977
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3534561729077384e-05,
+      "loss": 0.5493,
+      "step": 16978
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.353389490074346e-05,
+      "loss": 0.5536,
+      "step": 16979
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3533228054452783e-05,
+      "loss": 0.6006,
+      "step": 16980
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3532561190208738e-05,
+      "loss": 0.6178,
+      "step": 16981
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3531894308014712e-05,
+      "loss": 0.5652,
+      "step": 16982
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3531227407874098e-05,
+      "loss": 0.5985,
+      "step": 16983
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3530560489790286e-05,
+      "loss": 0.5107,
+      "step": 16984
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.352989355376666e-05,
+      "loss": 0.6337,
+      "step": 16985
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.352922659980661e-05,
+      "loss": 0.5634,
+      "step": 16986
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3528559627913529e-05,
+      "loss": 0.6275,
+      "step": 16987
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.35278926380908e-05,
+      "loss": 0.6258,
+      "step": 16988
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3527225630341817e-05,
+      "loss": 0.6528,
+      "step": 16989
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3526558604669969e-05,
+      "loss": 0.5675,
+      "step": 16990
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3525891561078642e-05,
+      "loss": 0.5829,
+      "step": 16991
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3525224499571228e-05,
+      "loss": 0.6749,
+      "step": 16992
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3524557420151117e-05,
+      "loss": 0.5699,
+      "step": 16993
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3523890322821694e-05,
+      "loss": 0.6002,
+      "step": 16994
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3523223207586353e-05,
+      "loss": 0.5783,
+      "step": 16995
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3522556074448487e-05,
+      "loss": 0.6486,
+      "step": 16996
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3521888923411477e-05,
+      "loss": 0.6417,
+      "step": 16997
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3521221754478724e-05,
+      "loss": 0.5266,
+      "step": 16998
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3520554567653607e-05,
+      "loss": 0.5927,
+      "step": 16999
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3519887362939523e-05,
+      "loss": 0.5357,
+      "step": 17000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.351922014033986e-05,
+      "loss": 0.6239,
+      "step": 17001
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3518552899858008e-05,
+      "loss": 0.575,
+      "step": 17002
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3517885641497359e-05,
+      "loss": 0.5894,
+      "step": 17003
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3517218365261304e-05,
+      "loss": 0.6006,
+      "step": 17004
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3516551071153232e-05,
+      "loss": 0.5844,
+      "step": 17005
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.351588375917653e-05,
+      "loss": 0.6399,
+      "step": 17006
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3515216429334598e-05,
+      "loss": 0.6257,
+      "step": 17007
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3514549081630821e-05,
+      "loss": 0.6166,
+      "step": 17008
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3513881716068592e-05,
+      "loss": 0.6032,
+      "step": 17009
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3513214332651298e-05,
+      "loss": 0.63,
+      "step": 17010
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3512546931382334e-05,
+      "loss": 0.5587,
+      "step": 17011
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3511879512265092e-05,
+      "loss": 0.604,
+      "step": 17012
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3511212075302957e-05,
+      "loss": 0.5746,
+      "step": 17013
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3510544620499328e-05,
+      "loss": 0.6286,
+      "step": 17014
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3509877147857595e-05,
+      "loss": 0.5814,
+      "step": 17015
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3509209657381148e-05,
+      "loss": 0.6475,
+      "step": 17016
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3508542149073378e-05,
+      "loss": 0.6625,
+      "step": 17017
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3507874622937678e-05,
+      "loss": 0.6662,
+      "step": 17018
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.350720707897744e-05,
+      "loss": 0.5985,
+      "step": 17019
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3506539517196058e-05,
+      "loss": 0.6576,
+      "step": 17020
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.350587193759692e-05,
+      "loss": 0.6058,
+      "step": 17021
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3505204340183417e-05,
+      "loss": 0.56,
+      "step": 17022
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3504536724958948e-05,
+      "loss": 0.6389,
+      "step": 17023
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.35038690919269e-05,
+      "loss": 0.5559,
+      "step": 17024
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3503201441090665e-05,
+      "loss": 0.6032,
+      "step": 17025
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3502533772453641e-05,
+      "loss": 0.6142,
+      "step": 17026
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3501866086019215e-05,
+      "loss": 0.6017,
+      "step": 17027
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.350119838179078e-05,
+      "loss": 0.6155,
+      "step": 17028
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3500530659771736e-05,
+      "loss": 0.5333,
+      "step": 17029
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3499862919965466e-05,
+      "loss": 0.6248,
+      "step": 17030
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3499195162375371e-05,
+      "loss": 0.5857,
+      "step": 17031
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3498527387004839e-05,
+      "loss": 0.6072,
+      "step": 17032
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3497859593857264e-05,
+      "loss": 0.7021,
+      "step": 17033
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3497191782936042e-05,
+      "loss": 0.5375,
+      "step": 17034
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.349652395424456e-05,
+      "loss": 0.6122,
+      "step": 17035
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.349585610778622e-05,
+      "loss": 0.5804,
+      "step": 17036
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.349518824356441e-05,
+      "loss": 0.6056,
+      "step": 17037
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3494520361582526e-05,
+      "loss": 0.6114,
+      "step": 17038
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3493852461843959e-05,
+      "loss": 0.6045,
+      "step": 17039
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3493184544352106e-05,
+      "loss": 0.5842,
+      "step": 17040
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3492516609110358e-05,
+      "loss": 0.5897,
+      "step": 17041
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.349184865612211e-05,
+      "loss": 0.6041,
+      "step": 17042
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3491180685390757e-05,
+      "loss": 0.6352,
+      "step": 17043
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3490512696919692e-05,
+      "loss": 0.5737,
+      "step": 17044
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.348984469071231e-05,
+      "loss": 0.6258,
+      "step": 17045
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3489176666772006e-05,
+      "loss": 0.5922,
+      "step": 17046
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3488508625102172e-05,
+      "loss": 0.5453,
+      "step": 17047
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3487840565706205e-05,
+      "loss": 0.6376,
+      "step": 17048
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3487172488587498e-05,
+      "loss": 0.553,
+      "step": 17049
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3486504393749447e-05,
+      "loss": 0.5537,
+      "step": 17050
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3485836281195446e-05,
+      "loss": 0.5878,
+      "step": 17051
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.348516815092889e-05,
+      "loss": 0.595,
+      "step": 17052
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3484500002953173e-05,
+      "loss": 0.5983,
+      "step": 17053
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.348383183727169e-05,
+      "loss": 0.5879,
+      "step": 17054
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3483163653887841e-05,
+      "loss": 0.5941,
+      "step": 17055
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3482495452805013e-05,
+      "loss": 0.5594,
+      "step": 17056
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.348182723402661e-05,
+      "loss": 0.5382,
+      "step": 17057
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.348115899755602e-05,
+      "loss": 0.5969,
+      "step": 17058
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3480490743396643e-05,
+      "loss": 0.6185,
+      "step": 17059
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3479822471551874e-05,
+      "loss": 0.654,
+      "step": 17060
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3479154182025109e-05,
+      "loss": 0.6139,
+      "step": 17061
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3478485874819738e-05,
+      "loss": 0.5532,
+      "step": 17062
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3477817549939163e-05,
+      "loss": 0.6321,
+      "step": 17063
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3477149207386782e-05,
+      "loss": 0.5279,
+      "step": 17064
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3476480847165985e-05,
+      "loss": 0.6007,
+      "step": 17065
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3475812469280171e-05,
+      "loss": 0.6343,
+      "step": 17066
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3475144073732735e-05,
+      "loss": 0.5442,
+      "step": 17067
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3474475660527073e-05,
+      "loss": 0.6072,
+      "step": 17068
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3473807229666587e-05,
+      "loss": 0.6284,
+      "step": 17069
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3473138781154664e-05,
+      "loss": 0.5792,
+      "step": 17070
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.347247031499471e-05,
+      "loss": 0.6448,
+      "step": 17071
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3471801831190117e-05,
+      "loss": 0.6112,
+      "step": 17072
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3471133329744279e-05,
+      "loss": 0.6183,
+      "step": 17073
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3470464810660598e-05,
+      "loss": 0.6215,
+      "step": 17074
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.346979627394247e-05,
+      "loss": 0.5999,
+      "step": 17075
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3469127719593291e-05,
+      "loss": 0.6367,
+      "step": 17076
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3468459147616455e-05,
+      "loss": 0.6395,
+      "step": 17077
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3467790558015366e-05,
+      "loss": 0.5999,
+      "step": 17078
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3467121950793414e-05,
+      "loss": 0.5697,
+      "step": 17079
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3466453325954003e-05,
+      "loss": 0.5907,
+      "step": 17080
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3465784683500525e-05,
+      "loss": 0.5494,
+      "step": 17081
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.346511602343638e-05,
+      "loss": 0.5988,
+      "step": 17082
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3464447345764968e-05,
+      "loss": 0.5771,
+      "step": 17083
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3463778650489683e-05,
+      "loss": 0.6246,
+      "step": 17084
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3463109937613925e-05,
+      "loss": 0.6325,
+      "step": 17085
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.346244120714109e-05,
+      "loss": 0.6329,
+      "step": 17086
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3461772459074578e-05,
+      "loss": 0.6079,
+      "step": 17087
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3461103693417787e-05,
+      "loss": 0.6493,
+      "step": 17088
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3460434910174112e-05,
+      "loss": 0.5666,
+      "step": 17089
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3459766109346957e-05,
+      "loss": 0.6378,
+      "step": 17090
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3459097290939715e-05,
+      "loss": 0.6356,
+      "step": 17091
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3458428454955787e-05,
+      "loss": 0.584,
+      "step": 17092
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.345775960139857e-05,
+      "loss": 0.5753,
+      "step": 17093
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3457090730271468e-05,
+      "loss": 0.6614,
+      "step": 17094
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3456421841577872e-05,
+      "loss": 0.6154,
+      "step": 17095
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3455752935321185e-05,
+      "loss": 0.5606,
+      "step": 17096
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3455084011504807e-05,
+      "loss": 0.6239,
+      "step": 17097
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3454415070132131e-05,
+      "loss": 0.5919,
+      "step": 17098
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3453746111206565e-05,
+      "loss": 0.6473,
+      "step": 17099
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.34530771347315e-05,
+      "loss": 0.6114,
+      "step": 17100
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.345240814071034e-05,
+      "loss": 0.6175,
+      "step": 17101
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3451739129146485e-05,
+      "loss": 0.6291,
+      "step": 17102
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3451070100043331e-05,
+      "loss": 0.6015,
+      "step": 17103
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.345040105340428e-05,
+      "loss": 0.6109,
+      "step": 17104
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.344973198923273e-05,
+      "loss": 0.6,
+      "step": 17105
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.344906290753208e-05,
+      "loss": 0.6018,
+      "step": 17106
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3448393808305733e-05,
+      "loss": 0.6311,
+      "step": 17107
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3447724691557088e-05,
+      "loss": 0.6499,
+      "step": 17108
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3447055557289541e-05,
+      "loss": 0.6083,
+      "step": 17109
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3446386405506499e-05,
+      "loss": 0.6157,
+      "step": 17110
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3445717236211355e-05,
+      "loss": 0.5795,
+      "step": 17111
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3445048049407517e-05,
+      "loss": 0.5619,
+      "step": 17112
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3444378845098376e-05,
+      "loss": 0.6179,
+      "step": 17113
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3443709623287337e-05,
+      "loss": 0.6366,
+      "step": 17114
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3443040383977805e-05,
+      "loss": 0.666,
+      "step": 17115
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.3442371127173177e-05,
+      "loss": 0.5643,
+      "step": 17116
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.344170185287685e-05,
+      "loss": 0.5691,
+      "step": 17117
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.344103256109223e-05,
+      "loss": 0.6369,
+      "step": 17118
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3440363251822714e-05,
+      "loss": 0.6346,
+      "step": 17119
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3439693925071704e-05,
+      "loss": 0.5852,
+      "step": 17120
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3439024580842605e-05,
+      "loss": 0.5891,
+      "step": 17121
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3438355219138813e-05,
+      "loss": 0.6093,
+      "step": 17122
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3437685839963733e-05,
+      "loss": 0.5734,
+      "step": 17123
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.343701644332076e-05,
+      "loss": 0.6487,
+      "step": 17124
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3436347029213303e-05,
+      "loss": 0.5837,
+      "step": 17125
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3435677597644761e-05,
+      "loss": 0.5643,
+      "step": 17126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3435008148618534e-05,
+      "loss": 0.6144,
+      "step": 17127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3434338682138025e-05,
+      "loss": 0.618,
+      "step": 17128
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3433669198206633e-05,
+      "loss": 0.5518,
+      "step": 17129
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3432999696827764e-05,
+      "loss": 0.6012,
+      "step": 17130
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.343233017800482e-05,
+      "loss": 0.6075,
+      "step": 17131
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.34316606417412e-05,
+      "loss": 0.5581,
+      "step": 17132
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3430991088040304e-05,
+      "loss": 0.5841,
+      "step": 17133
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3430321516905542e-05,
+      "loss": 0.5892,
+      "step": 17134
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3429651928340311e-05,
+      "loss": 0.6405,
+      "step": 17135
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3428982322348013e-05,
+      "loss": 0.5607,
+      "step": 17136
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3428312698932053e-05,
+      "loss": 0.6193,
+      "step": 17137
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.342764305809583e-05,
+      "loss": 0.6335,
+      "step": 17138
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3426973399842748e-05,
+      "loss": 0.575,
+      "step": 17139
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3426303724176214e-05,
+      "loss": 0.5569,
+      "step": 17140
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3425634031099623e-05,
+      "loss": 0.6345,
+      "step": 17141
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3424964320616386e-05,
+      "loss": 0.6185,
+      "step": 17142
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.34242945927299e-05,
+      "loss": 0.5539,
+      "step": 17143
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3423624847443572e-05,
+      "loss": 0.5929,
+      "step": 17144
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3422955084760805e-05,
+      "loss": 0.6103,
+      "step": 17145
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3422285304684996e-05,
+      "loss": 0.6202,
+      "step": 17146
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3421615507219556e-05,
+      "loss": 0.5971,
+      "step": 17147
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3420945692367886e-05,
+      "loss": 0.6154,
+      "step": 17148
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3420275860133387e-05,
+      "loss": 0.547,
+      "step": 17149
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3419606010519467e-05,
+      "loss": 0.587,
+      "step": 17150
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3418936143529526e-05,
+      "loss": 0.6011,
+      "step": 17151
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.341826625916697e-05,
+      "loss": 0.6748,
+      "step": 17152
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.34175963574352e-05,
+      "loss": 0.628,
+      "step": 17153
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3416926438337625e-05,
+      "loss": 0.5557,
+      "step": 17154
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3416256501877645e-05,
+      "loss": 0.5885,
+      "step": 17155
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3415586548058666e-05,
+      "loss": 0.5869,
+      "step": 17156
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3414916576884089e-05,
+      "loss": 0.5442,
+      "step": 17157
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3414246588357322e-05,
+      "loss": 0.6383,
+      "step": 17158
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3413576582481768e-05,
+      "loss": 0.5544,
+      "step": 17159
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3412906559260832e-05,
+      "loss": 0.5969,
+      "step": 17160
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3412236518697919e-05,
+      "loss": 0.6334,
+      "step": 17161
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3411566460796432e-05,
+      "loss": 0.6535,
+      "step": 17162
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3410896385559775e-05,
+      "loss": 0.6152,
+      "step": 17163
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3410226292991358e-05,
+      "loss": 0.5937,
+      "step": 17164
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340955618309458e-05,
+      "loss": 0.6771,
+      "step": 17165
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340888605587285e-05,
+      "loss": 0.5953,
+      "step": 17166
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3408215911329572e-05,
+      "loss": 0.6687,
+      "step": 17167
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3407545749468146e-05,
+      "loss": 0.6697,
+      "step": 17168
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3406875570291988e-05,
+      "loss": 0.6315,
+      "step": 17169
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3406205373804495e-05,
+      "loss": 0.5499,
+      "step": 17170
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3405535160009075e-05,
+      "loss": 0.6634,
+      "step": 17171
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3404864928909133e-05,
+      "loss": 0.5848,
+      "step": 17172
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3404194680508079e-05,
+      "loss": 0.5746,
+      "step": 17173
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340352441480931e-05,
+      "loss": 0.5565,
+      "step": 17174
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3402854131816239e-05,
+      "loss": 0.6055,
+      "step": 17175
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340218383153227e-05,
+      "loss": 0.6492,
+      "step": 17176
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340151351396081e-05,
+      "loss": 0.5655,
+      "step": 17177
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3400843179105259e-05,
+      "loss": 0.5926,
+      "step": 17178
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.340017282696903e-05,
+      "loss": 0.5257,
+      "step": 17179
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.339950245755553e-05,
+      "loss": 0.6218,
+      "step": 17180
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3398832070868157e-05,
+      "loss": 0.6758,
+      "step": 17181
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3398161666910324e-05,
+      "loss": 0.5832,
+      "step": 17182
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3397491245685439e-05,
+      "loss": 0.6179,
+      "step": 17183
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3396820807196905e-05,
+      "loss": 0.5835,
+      "step": 17184
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3396150351448128e-05,
+      "loss": 0.6864,
+      "step": 17185
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.339547987844252e-05,
+      "loss": 0.5892,
+      "step": 17186
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.339480938818348e-05,
+      "loss": 0.5852,
+      "step": 17187
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3394138880674423e-05,
+      "loss": 0.6627,
+      "step": 17188
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3393468355918749e-05,
+      "loss": 0.5911,
+      "step": 17189
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3392797813919871e-05,
+      "loss": 0.5855,
+      "step": 17190
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3392127254681193e-05,
+      "loss": 0.585,
+      "step": 17191
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3391456678206123e-05,
+      "loss": 0.6193,
+      "step": 17192
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.339078608449807e-05,
+      "loss": 0.5879,
+      "step": 17193
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3390115473560437e-05,
+      "loss": 0.6055,
+      "step": 17194
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3389444845396637e-05,
+      "loss": 0.6079,
+      "step": 17195
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3388774200010075e-05,
+      "loss": 0.615,
+      "step": 17196
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3388103537404156e-05,
+      "loss": 0.666,
+      "step": 17197
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3387432857582293e-05,
+      "loss": 0.5886,
+      "step": 17198
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.338676216054789e-05,
+      "loss": 0.571,
+      "step": 17199
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3386091446304356e-05,
+      "loss": 0.5651,
+      "step": 17200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3385420714855101e-05,
+      "loss": 0.5455,
+      "step": 17201
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3384749966203534e-05,
+      "loss": 0.6428,
+      "step": 17202
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3384079200353057e-05,
+      "loss": 0.6117,
+      "step": 17203
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3383408417307087e-05,
+      "loss": 0.5562,
+      "step": 17204
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3382737617069026e-05,
+      "loss": 0.5764,
+      "step": 17205
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3382066799642281e-05,
+      "loss": 0.6702,
+      "step": 17206
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3381395965030271e-05,
+      "loss": 0.6486,
+      "step": 17207
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3380725113236393e-05,
+      "loss": 0.6095,
+      "step": 17208
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.338005424426406e-05,
+      "loss": 0.6326,
+      "step": 17209
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3379383358116685e-05,
+      "loss": 0.6267,
+      "step": 17210
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.337871245479767e-05,
+      "loss": 0.6135,
+      "step": 17211
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3378041534310428e-05,
+      "loss": 0.6708,
+      "step": 17212
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.337737059665837e-05,
+      "loss": 0.5893,
+      "step": 17213
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3376699641844904e-05,
+      "loss": 0.5818,
+      "step": 17214
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3376028669873433e-05,
+      "loss": 0.6509,
+      "step": 17215
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3375357680747376e-05,
+      "loss": 0.5553,
+      "step": 17216
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3374686674470136e-05,
+      "loss": 0.59,
+      "step": 17217
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3374015651045128e-05,
+      "loss": 0.6613,
+      "step": 17218
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3373344610475754e-05,
+      "loss": 0.5756,
+      "step": 17219
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3372673552765429e-05,
+      "loss": 0.6669,
+      "step": 17220
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3372002477917563e-05,
+      "loss": 0.6424,
+      "step": 17221
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3371331385935565e-05,
+      "loss": 0.6423,
+      "step": 17222
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3370660276822845e-05,
+      "loss": 0.6431,
+      "step": 17223
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3369989150582811e-05,
+      "loss": 0.6005,
+      "step": 17224
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3369318007218877e-05,
+      "loss": 0.6008,
+      "step": 17225
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3368646846734451e-05,
+      "loss": 0.5513,
+      "step": 17226
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3367975669132945e-05,
+      "loss": 0.5611,
+      "step": 17227
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3367304474417764e-05,
+      "loss": 0.5721,
+      "step": 17228
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3366633262592328e-05,
+      "loss": 0.6467,
+      "step": 17229
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.336596203366004e-05,
+      "loss": 0.6001,
+      "step": 17230
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3365290787624314e-05,
+      "loss": 0.585,
+      "step": 17231
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3364619524488558e-05,
+      "loss": 0.594,
+      "step": 17232
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3363948244256188e-05,
+      "loss": 0.6119,
+      "step": 17233
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.336327694693061e-05,
+      "loss": 0.6518,
+      "step": 17234
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3362605632515238e-05,
+      "loss": 0.588,
+      "step": 17235
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.336193430101348e-05,
+      "loss": 0.6456,
+      "step": 17236
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3361262952428753e-05,
+      "loss": 0.5921,
+      "step": 17237
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.336059158676446e-05,
+      "loss": 0.616,
+      "step": 17238
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3359920204024017e-05,
+      "loss": 0.6838,
+      "step": 17239
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3359248804210838e-05,
+      "loss": 0.5697,
+      "step": 17240
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.335857738732833e-05,
+      "loss": 0.6413,
+      "step": 17241
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.335790595337991e-05,
+      "loss": 0.5736,
+      "step": 17242
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3357234502368983e-05,
+      "loss": 0.6155,
+      "step": 17243
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3356563034298967e-05,
+      "loss": 0.5614,
+      "step": 17244
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.335589154917327e-05,
+      "loss": 0.5325,
+      "step": 17245
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3355220046995302e-05,
+      "loss": 0.6291,
+      "step": 17246
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3354548527768483e-05,
+      "loss": 0.6547,
+      "step": 17247
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3353876991496218e-05,
+      "loss": 0.6172,
+      "step": 17248
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3353205438181922e-05,
+      "loss": 0.6391,
+      "step": 17249
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3352533867829009e-05,
+      "loss": 0.6195,
+      "step": 17250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3351862280440889e-05,
+      "loss": 0.6076,
+      "step": 17251
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3351190676020976e-05,
+      "loss": 0.6469,
+      "step": 17252
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3350519054572677e-05,
+      "loss": 0.6424,
+      "step": 17253
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3349847416099415e-05,
+      "loss": 0.5601,
+      "step": 17254
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3349175760604593e-05,
+      "loss": 0.5661,
+      "step": 17255
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3348504088091632e-05,
+      "loss": 0.6009,
+      "step": 17256
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3347832398563939e-05,
+      "loss": 0.6138,
+      "step": 17257
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3347160692024927e-05,
+      "loss": 0.6246,
+      "step": 17258
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3346488968478014e-05,
+      "loss": 0.6049,
+      "step": 17259
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.334581722792661e-05,
+      "loss": 0.6002,
+      "step": 17260
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.334514547037413e-05,
+      "loss": 0.6005,
+      "step": 17261
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3344473695823983e-05,
+      "loss": 0.6526,
+      "step": 17262
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3343801904279588e-05,
+      "loss": 0.6135,
+      "step": 17263
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3343130095744357e-05,
+      "loss": 0.5698,
+      "step": 17264
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.3342458270221701e-05,
+      "loss": 0.6817,
+      "step": 17265
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3341786427715036e-05,
+      "loss": 0.5699,
+      "step": 17266
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3341114568227775e-05,
+      "loss": 0.6133,
+      "step": 17267
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3340442691763333e-05,
+      "loss": 0.5644,
+      "step": 17268
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3339770798325124e-05,
+      "loss": 0.6237,
+      "step": 17269
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3339098887916561e-05,
+      "loss": 0.5815,
+      "step": 17270
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3338426960541058e-05,
+      "loss": 0.5667,
+      "step": 17271
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3337755016202032e-05,
+      "loss": 0.531,
+      "step": 17272
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3337083054902895e-05,
+      "loss": 0.6227,
+      "step": 17273
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.333641107664706e-05,
+      "loss": 0.6096,
+      "step": 17274
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3335739081437944e-05,
+      "loss": 0.5855,
+      "step": 17275
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.333506706927896e-05,
+      "loss": 0.5782,
+      "step": 17276
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3334395040173523e-05,
+      "loss": 0.5732,
+      "step": 17277
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3333722994125051e-05,
+      "loss": 0.617,
+      "step": 17278
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3333050931136952e-05,
+      "loss": 0.5878,
+      "step": 17279
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3332378851212651e-05,
+      "loss": 0.6345,
+      "step": 17280
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3331706754355552e-05,
+      "loss": 0.6176,
+      "step": 17281
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3331034640569076e-05,
+      "loss": 0.6152,
+      "step": 17282
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.333036250985664e-05,
+      "loss": 0.5892,
+      "step": 17283
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3329690362221654e-05,
+      "loss": 0.6265,
+      "step": 17284
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3329018197667539e-05,
+      "loss": 0.6269,
+      "step": 17285
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3328346016197706e-05,
+      "loss": 0.6313,
+      "step": 17286
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.332767381781557e-05,
+      "loss": 0.6459,
+      "step": 17287
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3327001602524552e-05,
+      "loss": 0.5874,
+      "step": 17288
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3326329370328064e-05,
+      "loss": 0.6186,
+      "step": 17289
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3325657121229521e-05,
+      "loss": 0.5613,
+      "step": 17290
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3324984855232341e-05,
+      "loss": 0.6244,
+      "step": 17291
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3324312572339939e-05,
+      "loss": 0.602,
+      "step": 17292
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3323640272555731e-05,
+      "loss": 0.5918,
+      "step": 17293
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3322967955883134e-05,
+      "loss": 0.5894,
+      "step": 17294
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3322295622325563e-05,
+      "loss": 0.6293,
+      "step": 17295
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3321623271886437e-05,
+      "loss": 0.5427,
+      "step": 17296
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3320950904569168e-05,
+      "loss": 0.5959,
+      "step": 17297
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3320278520377175e-05,
+      "loss": 0.5583,
+      "step": 17298
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3319606119313874e-05,
+      "loss": 0.5968,
+      "step": 17299
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3318933701382683e-05,
+      "loss": 0.5687,
+      "step": 17300
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3318261266587017e-05,
+      "loss": 0.575,
+      "step": 17301
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3317588814930295e-05,
+      "loss": 0.5645,
+      "step": 17302
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3316916346415933e-05,
+      "loss": 0.5974,
+      "step": 17303
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3316243861047347e-05,
+      "loss": 0.5502,
+      "step": 17304
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3315571358827955e-05,
+      "loss": 0.6009,
+      "step": 17305
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3314898839761173e-05,
+      "loss": 0.6123,
+      "step": 17306
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.331422630385042e-05,
+      "loss": 0.5921,
+      "step": 17307
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.331355375109911e-05,
+      "loss": 0.6372,
+      "step": 17308
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3312881181510665e-05,
+      "loss": 0.5584,
+      "step": 17309
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3312208595088502e-05,
+      "loss": 0.5797,
+      "step": 17310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3311535991836037e-05,
+      "loss": 0.5577,
+      "step": 17311
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3310863371756688e-05,
+      "loss": 0.5933,
+      "step": 17312
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3310190734853872e-05,
+      "loss": 0.571,
+      "step": 17313
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3309518081131006e-05,
+      "loss": 0.6273,
+      "step": 17314
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3308845410591512e-05,
+      "loss": 0.5481,
+      "step": 17315
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3308172723238802e-05,
+      "loss": 0.6313,
+      "step": 17316
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3307500019076298e-05,
+      "loss": 0.6609,
+      "step": 17317
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3306827298107418e-05,
+      "loss": 0.5783,
+      "step": 17318
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.330615456033558e-05,
+      "loss": 0.546,
+      "step": 17319
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3305481805764204e-05,
+      "loss": 0.5785,
+      "step": 17320
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3304809034396708e-05,
+      "loss": 0.6485,
+      "step": 17321
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3304136246236505e-05,
+      "loss": 0.5729,
+      "step": 17322
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3303463441287021e-05,
+      "loss": 0.6045,
+      "step": 17323
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3302790619551673e-05,
+      "loss": 0.579,
+      "step": 17324
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3302117781033874e-05,
+      "loss": 0.5406,
+      "step": 17325
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3301444925737053e-05,
+      "loss": 0.6175,
+      "step": 17326
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3300772053664621e-05,
+      "loss": 0.5846,
+      "step": 17327
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3300099164819998e-05,
+      "loss": 0.6329,
+      "step": 17328
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3299426259206606e-05,
+      "loss": 0.6378,
+      "step": 17329
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3298753336827863e-05,
+      "loss": 0.6084,
+      "step": 17330
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3298080397687189e-05,
+      "loss": 0.5625,
+      "step": 17331
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3297407441788004e-05,
+      "loss": 0.6552,
+      "step": 17332
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3296734469133723e-05,
+      "loss": 0.603,
+      "step": 17333
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3296061479727772e-05,
+      "loss": 0.6155,
+      "step": 17334
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3295388473573563e-05,
+      "loss": 0.5735,
+      "step": 17335
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3294715450674521e-05,
+      "loss": 0.6184,
+      "step": 17336
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3294042411034068e-05,
+      "loss": 0.5575,
+      "step": 17337
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3293369354655616e-05,
+      "loss": 0.5637,
+      "step": 17338
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3292696281542594e-05,
+      "loss": 0.5233,
+      "step": 17339
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3292023191698418e-05,
+      "loss": 0.5791,
+      "step": 17340
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3291350085126504e-05,
+      "loss": 0.5754,
+      "step": 17341
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.329067696183028e-05,
+      "loss": 0.5815,
+      "step": 17342
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3290003821813163e-05,
+      "loss": 0.5348,
+      "step": 17343
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3289330665078568e-05,
+      "loss": 0.6287,
+      "step": 17344
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3288657491629926e-05,
+      "loss": 0.6177,
+      "step": 17345
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3287984301470649e-05,
+      "loss": 0.6336,
+      "step": 17346
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3287311094604162e-05,
+      "loss": 0.6031,
+      "step": 17347
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3286637871033885e-05,
+      "loss": 0.6238,
+      "step": 17348
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3285964630763238e-05,
+      "loss": 0.5582,
+      "step": 17349
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3285291373795643e-05,
+      "loss": 0.7016,
+      "step": 17350
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3284618100134518e-05,
+      "loss": 0.5871,
+      "step": 17351
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3283944809783291e-05,
+      "loss": 0.6203,
+      "step": 17352
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3283271502745377e-05,
+      "loss": 0.5682,
+      "step": 17353
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3282598179024198e-05,
+      "loss": 0.6297,
+      "step": 17354
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3281924838623175e-05,
+      "loss": 0.5697,
+      "step": 17355
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3281251481545732e-05,
+      "loss": 0.6572,
+      "step": 17356
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3280578107795288e-05,
+      "loss": 0.6661,
+      "step": 17357
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3279904717375267e-05,
+      "loss": 0.613,
+      "step": 17358
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3279231310289092e-05,
+      "loss": 0.6003,
+      "step": 17359
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3278557886540178e-05,
+      "loss": 0.6537,
+      "step": 17360
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3277884446131955e-05,
+      "loss": 0.478,
+      "step": 17361
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3277210989067838e-05,
+      "loss": 0.5514,
+      "step": 17362
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3276537515351254e-05,
+      "loss": 0.6183,
+      "step": 17363
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3275864024985623e-05,
+      "loss": 0.6147,
+      "step": 17364
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3275190517974368e-05,
+      "loss": 0.5108,
+      "step": 17365
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3274516994320908e-05,
+      "loss": 0.6005,
+      "step": 17366
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3273843454028671e-05,
+      "loss": 0.5361,
+      "step": 17367
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3273169897101075e-05,
+      "loss": 0.6125,
+      "step": 17368
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3272496323541545e-05,
+      "loss": 0.5606,
+      "step": 17369
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3271822733353503e-05,
+      "loss": 0.6163,
+      "step": 17370
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3271149126540373e-05,
+      "loss": 0.5696,
+      "step": 17371
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3270475503105576e-05,
+      "loss": 0.6102,
+      "step": 17372
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3269801863052533e-05,
+      "loss": 0.6363,
+      "step": 17373
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.326912820638467e-05,
+      "loss": 0.5791,
+      "step": 17374
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3268454533105411e-05,
+      "loss": 0.6122,
+      "step": 17375
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3267780843218174e-05,
+      "loss": 0.6066,
+      "step": 17376
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3267107136726384e-05,
+      "loss": 0.5585,
+      "step": 17377
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3266433413633472e-05,
+      "loss": 0.6284,
+      "step": 17378
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.326575967394285e-05,
+      "loss": 0.6844,
+      "step": 17379
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3265085917657951e-05,
+      "loss": 0.5802,
+      "step": 17380
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.326441214478219e-05,
+      "loss": 0.6263,
+      "step": 17381
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3263738355318995e-05,
+      "loss": 0.632,
+      "step": 17382
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3263064549271793e-05,
+      "loss": 0.5562,
+      "step": 17383
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3262390726644003e-05,
+      "loss": 0.6014,
+      "step": 17384
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3261716887439049e-05,
+      "loss": 0.6322,
+      "step": 17385
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3261043031660358e-05,
+      "loss": 0.614,
+      "step": 17386
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3260369159311352e-05,
+      "loss": 0.5973,
+      "step": 17387
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3259695270395455e-05,
+      "loss": 0.6129,
+      "step": 17388
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3259021364916092e-05,
+      "loss": 0.622,
+      "step": 17389
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3258347442876689e-05,
+      "loss": 0.6114,
+      "step": 17390
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3257673504280663e-05,
+      "loss": 0.5635,
+      "step": 17391
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3256999549131449e-05,
+      "loss": 0.5674,
+      "step": 17392
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3256325577432466e-05,
+      "loss": 0.6617,
+      "step": 17393
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3255651589187138e-05,
+      "loss": 0.6402,
+      "step": 17394
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3254977584398887e-05,
+      "loss": 0.5967,
+      "step": 17395
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3254303563071145e-05,
+      "loss": 0.5359,
+      "step": 17396
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3253629525207335e-05,
+      "loss": 0.6061,
+      "step": 17397
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3252955470810877e-05,
+      "loss": 0.6103,
+      "step": 17398
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3252281399885205e-05,
+      "loss": 0.6087,
+      "step": 17399
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3251607312433735e-05,
+      "loss": 0.6153,
+      "step": 17400
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3250933208459893e-05,
+      "loss": 0.5964,
+      "step": 17401
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3250259087967113e-05,
+      "loss": 0.6084,
+      "step": 17402
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3249584950958812e-05,
+      "loss": 0.5903,
+      "step": 17403
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3248910797438416e-05,
+      "loss": 0.6304,
+      "step": 17404
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3248236627409357e-05,
+      "loss": 0.581,
+      "step": 17405
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3247562440875053e-05,
+      "loss": 0.5729,
+      "step": 17406
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3246888237838938e-05,
+      "loss": 0.6056,
+      "step": 17407
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3246214018304428e-05,
+      "loss": 0.5655,
+      "step": 17408
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3245539782274956e-05,
+      "loss": 0.5661,
+      "step": 17409
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3244865529753945e-05,
+      "loss": 0.6262,
+      "step": 17410
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3244191260744823e-05,
+      "loss": 0.5989,
+      "step": 17411
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.3243516975251015e-05,
+      "loss": 0.5637,
+      "step": 17412
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3242842673275949e-05,
+      "loss": 0.6571,
+      "step": 17413
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3242168354823046e-05,
+      "loss": 0.649,
+      "step": 17414
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3241494019895735e-05,
+      "loss": 0.6069,
+      "step": 17415
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3240819668497449e-05,
+      "loss": 0.7424,
+      "step": 17416
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3240145300631604e-05,
+      "loss": 0.6115,
+      "step": 17417
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3239470916301635e-05,
+      "loss": 0.6079,
+      "step": 17418
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3238796515510965e-05,
+      "loss": 0.6498,
+      "step": 17419
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.323812209826302e-05,
+      "loss": 0.5904,
+      "step": 17420
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3237447664561232e-05,
+      "loss": 0.604,
+      "step": 17421
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.323677321440902e-05,
+      "loss": 0.6197,
+      "step": 17422
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3236098747809815e-05,
+      "loss": 0.5569,
+      "step": 17423
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3235424264767049e-05,
+      "loss": 0.584,
+      "step": 17424
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.323474976528414e-05,
+      "loss": 0.6161,
+      "step": 17425
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3234075249364524e-05,
+      "loss": 0.6266,
+      "step": 17426
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3233400717011622e-05,
+      "loss": 0.6198,
+      "step": 17427
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3232726168228867e-05,
+      "loss": 0.5999,
+      "step": 17428
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.323205160301968e-05,
+      "loss": 0.5819,
+      "step": 17429
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3231377021387494e-05,
+      "loss": 0.5544,
+      "step": 17430
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3230702423335734e-05,
+      "loss": 0.5941,
+      "step": 17431
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.323002780886783e-05,
+      "loss": 0.6115,
+      "step": 17432
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.322935317798721e-05,
+      "loss": 0.6281,
+      "step": 17433
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.32286785306973e-05,
+      "loss": 0.6406,
+      "step": 17434
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3228003867001526e-05,
+      "loss": 0.6159,
+      "step": 17435
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3227329186903321e-05,
+      "loss": 0.5848,
+      "step": 17436
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3226654490406112e-05,
+      "loss": 0.5852,
+      "step": 17437
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3225979777513326e-05,
+      "loss": 0.6238,
+      "step": 17438
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.322530504822839e-05,
+      "loss": 0.6214,
+      "step": 17439
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3224630302554737e-05,
+      "loss": 0.5974,
+      "step": 17440
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.322395554049579e-05,
+      "loss": 0.6326,
+      "step": 17441
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3223280762054981e-05,
+      "loss": 0.5755,
+      "step": 17442
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3222605967235745e-05,
+      "loss": 0.5978,
+      "step": 17443
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3221931156041498e-05,
+      "loss": 0.6064,
+      "step": 17444
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3221256328475678e-05,
+      "loss": 0.5896,
+      "step": 17445
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3220581484541711e-05,
+      "loss": 0.5972,
+      "step": 17446
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3219906624243021e-05,
+      "loss": 0.5629,
+      "step": 17447
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.321923174758305e-05,
+      "loss": 0.6279,
+      "step": 17448
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3218556854565217e-05,
+      "loss": 0.5628,
+      "step": 17449
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3217881945192952e-05,
+      "loss": 0.5681,
+      "step": 17450
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3217207019469691e-05,
+      "loss": 0.5849,
+      "step": 17451
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3216532077398854e-05,
+      "loss": 0.5977,
+      "step": 17452
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.321585711898388e-05,
+      "loss": 0.5786,
+      "step": 17453
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3215182144228189e-05,
+      "loss": 0.5655,
+      "step": 17454
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3214507153135218e-05,
+      "loss": 0.5916,
+      "step": 17455
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3213832145708396e-05,
+      "loss": 0.5366,
+      "step": 17456
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.321315712195115e-05,
+      "loss": 0.602,
+      "step": 17457
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3212482081866913e-05,
+      "loss": 0.6202,
+      "step": 17458
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3211807025459114e-05,
+      "loss": 0.6057,
+      "step": 17459
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3211131952731183e-05,
+      "loss": 0.5889,
+      "step": 17460
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3210456863686546e-05,
+      "loss": 0.5806,
+      "step": 17461
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3209781758328641e-05,
+      "loss": 0.5933,
+      "step": 17462
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3209106636660895e-05,
+      "loss": 0.6594,
+      "step": 17463
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.320843149868674e-05,
+      "loss": 0.5728,
+      "step": 17464
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.32077563444096e-05,
+      "loss": 0.5491,
+      "step": 17465
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3207081173832913e-05,
+      "loss": 0.593,
+      "step": 17466
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3206405986960112e-05,
+      "loss": 0.6471,
+      "step": 17467
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3205730783794618e-05,
+      "loss": 0.5904,
+      "step": 17468
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3205055564339869e-05,
+      "loss": 0.6259,
+      "step": 17469
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3204380328599294e-05,
+      "loss": 0.5882,
+      "step": 17470
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3203705076576321e-05,
+      "loss": 0.6144,
+      "step": 17471
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3203029808274388e-05,
+      "loss": 0.5699,
+      "step": 17472
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3202354523696919e-05,
+      "loss": 0.5854,
+      "step": 17473
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.320167922284735e-05,
+      "loss": 0.5676,
+      "step": 17474
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3201003905729113e-05,
+      "loss": 0.6578,
+      "step": 17475
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3200328572345636e-05,
+      "loss": 0.5616,
+      "step": 17476
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3199653222700352e-05,
+      "loss": 0.6129,
+      "step": 17477
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3198977856796695e-05,
+      "loss": 0.6185,
+      "step": 17478
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3198302474638091e-05,
+      "loss": 0.556,
+      "step": 17479
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.319762707622798e-05,
+      "loss": 0.5872,
+      "step": 17480
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3196951661569788e-05,
+      "loss": 0.608,
+      "step": 17481
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3196276230666945e-05,
+      "loss": 0.6446,
+      "step": 17482
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.319560078352289e-05,
+      "loss": 0.6011,
+      "step": 17483
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.319492532014105e-05,
+      "loss": 0.5943,
+      "step": 17484
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.319424984052486e-05,
+      "loss": 0.6429,
+      "step": 17485
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3193574344677751e-05,
+      "loss": 0.5913,
+      "step": 17486
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3192898832603153e-05,
+      "loss": 0.5535,
+      "step": 17487
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3192223304304503e-05,
+      "loss": 0.6249,
+      "step": 17488
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3191547759785234e-05,
+      "loss": 0.5746,
+      "step": 17489
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3190872199048773e-05,
+      "loss": 0.5921,
+      "step": 17490
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3190196622098556e-05,
+      "loss": 0.5948,
+      "step": 17491
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3189521028938015e-05,
+      "loss": 0.6072,
+      "step": 17492
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3188845419570584e-05,
+      "loss": 0.6189,
+      "step": 17493
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3188169793999696e-05,
+      "loss": 0.5843,
+      "step": 17494
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3187494152228784e-05,
+      "loss": 0.5417,
+      "step": 17495
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.318681849426128e-05,
+      "loss": 0.6148,
+      "step": 17496
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3186142820100619e-05,
+      "loss": 0.578,
+      "step": 17497
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.318546712975023e-05,
+      "loss": 0.5769,
+      "step": 17498
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3184791423213554e-05,
+      "loss": 0.6057,
+      "step": 17499
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3184115700494019e-05,
+      "loss": 0.6076,
+      "step": 17500
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3183439961595056e-05,
+      "loss": 0.631,
+      "step": 17501
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3182764206520107e-05,
+      "loss": 0.5607,
+      "step": 17502
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3182088435272599e-05,
+      "loss": 0.6105,
+      "step": 17503
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3181412647855967e-05,
+      "loss": 0.5767,
+      "step": 17504
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3180736844273647e-05,
+      "loss": 0.6306,
+      "step": 17505
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3180061024529069e-05,
+      "loss": 0.6178,
+      "step": 17506
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3179385188625672e-05,
+      "loss": 0.6149,
+      "step": 17507
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3178709336566888e-05,
+      "loss": 0.6153,
+      "step": 17508
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.317803346835615e-05,
+      "loss": 0.5871,
+      "step": 17509
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3177357583996897e-05,
+      "loss": 0.7038,
+      "step": 17510
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3176681683492554e-05,
+      "loss": 0.5808,
+      "step": 17511
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3176005766846562e-05,
+      "loss": 0.5478,
+      "step": 17512
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3175329834062355e-05,
+      "loss": 0.5804,
+      "step": 17513
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.317465388514337e-05,
+      "loss": 0.6436,
+      "step": 17514
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3173977920093035e-05,
+      "loss": 0.6143,
+      "step": 17515
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3173301938914791e-05,
+      "loss": 0.5867,
+      "step": 17516
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.317262594161207e-05,
+      "loss": 0.6144,
+      "step": 17517
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3171949928188308e-05,
+      "loss": 0.5624,
+      "step": 17518
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3171273898646937e-05,
+      "loss": 0.6014,
+      "step": 17519
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3170597852991396e-05,
+      "loss": 0.6225,
+      "step": 17520
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.316992179122512e-05,
+      "loss": 0.5325,
+      "step": 17521
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3169245713351542e-05,
+      "loss": 0.632,
+      "step": 17522
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3168569619374096e-05,
+      "loss": 0.5549,
+      "step": 17523
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3167893509296224e-05,
+      "loss": 0.6742,
+      "step": 17524
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3167217383121354e-05,
+      "loss": 0.6179,
+      "step": 17525
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3166541240852928e-05,
+      "loss": 0.622,
+      "step": 17526
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3165865082494376e-05,
+      "loss": 0.5972,
+      "step": 17527
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3165188908049138e-05,
+      "loss": 0.63,
+      "step": 17528
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.316451271752065e-05,
+      "loss": 0.6307,
+      "step": 17529
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3163836510912341e-05,
+      "loss": 0.6142,
+      "step": 17530
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3163160288227655e-05,
+      "loss": 0.5518,
+      "step": 17531
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3162484049470025e-05,
+      "loss": 0.6241,
+      "step": 17532
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.316180779464289e-05,
+      "loss": 0.6172,
+      "step": 17533
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3161131523749677e-05,
+      "loss": 0.5977,
+      "step": 17534
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3160455236793834e-05,
+      "loss": 0.6016,
+      "step": 17535
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3159778933778791e-05,
+      "loss": 0.6103,
+      "step": 17536
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3159102614707988e-05,
+      "loss": 0.5954,
+      "step": 17537
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3158426279584858e-05,
+      "loss": 0.5444,
+      "step": 17538
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3157749928412837e-05,
+      "loss": 0.6126,
+      "step": 17539
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3157073561195366e-05,
+      "loss": 0.5867,
+      "step": 17540
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3156397177935878e-05,
+      "loss": 0.6209,
+      "step": 17541
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3155720778637813e-05,
+      "loss": 0.5985,
+      "step": 17542
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3155044363304605e-05,
+      "loss": 0.5623,
+      "step": 17543
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3154367931939693e-05,
+      "loss": 0.5693,
+      "step": 17544
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3153691484546518e-05,
+      "loss": 0.6137,
+      "step": 17545
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3153015021128509e-05,
+      "loss": 0.6215,
+      "step": 17546
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3152338541689106e-05,
+      "loss": 0.6403,
+      "step": 17547
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.315166204623175e-05,
+      "loss": 0.5055,
+      "step": 17548
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3150985534759878e-05,
+      "loss": 0.6098,
+      "step": 17549
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3150309007276922e-05,
+      "loss": 0.6416,
+      "step": 17550
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3149632463786324e-05,
+      "loss": 0.6144,
+      "step": 17551
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3148955904291523e-05,
+      "loss": 0.5604,
+      "step": 17552
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3148279328795952e-05,
+      "loss": 0.6089,
+      "step": 17553
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3147602737303054e-05,
+      "loss": 0.6082,
+      "step": 17554
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3146926129816262e-05,
+      "loss": 0.5814,
+      "step": 17555
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.314624950633902e-05,
+      "loss": 0.6368,
+      "step": 17556
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.314557286687476e-05,
+      "loss": 0.5321,
+      "step": 17557
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3144896211426926e-05,
+      "loss": 0.5525,
+      "step": 17558
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.3144219539998951e-05,
+      "loss": 0.5919,
+      "step": 17559
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3143542852594276e-05,
+      "loss": 0.6172,
+      "step": 17560
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3142866149216337e-05,
+      "loss": 0.6007,
+      "step": 17561
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.314218942986858e-05,
+      "loss": 0.6643,
+      "step": 17562
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3141512694554434e-05,
+      "loss": 0.5629,
+      "step": 17563
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3140835943277346e-05,
+      "loss": 0.5471,
+      "step": 17564
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3140159176040746e-05,
+      "loss": 0.5732,
+      "step": 17565
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3139482392848081e-05,
+      "loss": 0.5667,
+      "step": 17566
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3138805593702786e-05,
+      "loss": 0.5925,
+      "step": 17567
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.31381287786083e-05,
+      "loss": 0.5904,
+      "step": 17568
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3137451947568062e-05,
+      "loss": 0.524,
+      "step": 17569
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3136775100585512e-05,
+      "loss": 0.5909,
+      "step": 17570
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.313609823766409e-05,
+      "loss": 0.5937,
+      "step": 17571
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3135421358807234e-05,
+      "loss": 0.6477,
+      "step": 17572
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3134744464018382e-05,
+      "loss": 0.6089,
+      "step": 17573
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3134067553300978e-05,
+      "loss": 0.6055,
+      "step": 17574
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.313339062665846e-05,
+      "loss": 0.5911,
+      "step": 17575
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3132713684094262e-05,
+      "loss": 0.5639,
+      "step": 17576
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.313203672561183e-05,
+      "loss": 0.6196,
+      "step": 17577
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3131359751214606e-05,
+      "loss": 0.5797,
+      "step": 17578
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3130682760906019e-05,
+      "loss": 0.6357,
+      "step": 17579
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3130005754689519e-05,
+      "loss": 0.6531,
+      "step": 17580
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3129328732568545e-05,
+      "loss": 0.5859,
+      "step": 17581
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3128651694546535e-05,
+      "loss": 0.5871,
+      "step": 17582
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.312797464062693e-05,
+      "loss": 0.5924,
+      "step": 17583
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3127297570813166e-05,
+      "loss": 0.6459,
+      "step": 17584
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3126620485108692e-05,
+      "loss": 0.6264,
+      "step": 17585
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3125943383516942e-05,
+      "loss": 0.5997,
+      "step": 17586
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3125266266041356e-05,
+      "loss": 0.5879,
+      "step": 17587
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.312458913268538e-05,
+      "loss": 0.6105,
+      "step": 17588
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.312391198345245e-05,
+      "loss": 0.5811,
+      "step": 17589
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3123234818346006e-05,
+      "loss": 0.6282,
+      "step": 17590
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3122557637369495e-05,
+      "loss": 0.6039,
+      "step": 17591
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3121880440526353e-05,
+      "loss": 0.6496,
+      "step": 17592
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.312120322782002e-05,
+      "loss": 0.576,
+      "step": 17593
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3120525999253944e-05,
+      "loss": 0.5899,
+      "step": 17594
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3119848754831557e-05,
+      "loss": 0.6299,
+      "step": 17595
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3119171494556304e-05,
+      "loss": 0.5993,
+      "step": 17596
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3118494218431632e-05,
+      "loss": 0.6322,
+      "step": 17597
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3117816926460974e-05,
+      "loss": 0.6736,
+      "step": 17598
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3117139618647774e-05,
+      "loss": 0.5959,
+      "step": 17599
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.311646229499548e-05,
+      "loss": 0.6221,
+      "step": 17600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3115784955507524e-05,
+      "loss": 0.5669,
+      "step": 17601
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3115107600187353e-05,
+      "loss": 0.574,
+      "step": 17602
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3114430229038408e-05,
+      "loss": 0.6087,
+      "step": 17603
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3113752842064132e-05,
+      "loss": 0.6182,
+      "step": 17604
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3113075439267966e-05,
+      "loss": 0.6091,
+      "step": 17605
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3112398020653351e-05,
+      "loss": 0.5698,
+      "step": 17606
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.311172058622373e-05,
+      "loss": 0.5977,
+      "step": 17607
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3111043135982549e-05,
+      "loss": 0.6212,
+      "step": 17608
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.311036566993324e-05,
+      "loss": 0.6458,
+      "step": 17609
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.310968818807926e-05,
+      "loss": 0.5985,
+      "step": 17610
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3109010690424038e-05,
+      "loss": 0.6279,
+      "step": 17611
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3108333176971022e-05,
+      "loss": 0.5842,
+      "step": 17612
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3107655647723655e-05,
+      "loss": 0.579,
+      "step": 17613
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3106978102685382e-05,
+      "loss": 0.5786,
+      "step": 17614
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.310630054185964e-05,
+      "loss": 0.6602,
+      "step": 17615
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3105622965249879e-05,
+      "loss": 0.5796,
+      "step": 17616
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3104945372859535e-05,
+      "loss": 0.6407,
+      "step": 17617
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3104267764692059e-05,
+      "loss": 0.6068,
+      "step": 17618
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3103590140750884e-05,
+      "loss": 0.6171,
+      "step": 17619
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.310291250103946e-05,
+      "loss": 0.6173,
+      "step": 17620
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.310223484556123e-05,
+      "loss": 0.5576,
+      "step": 17621
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3101557174319635e-05,
+      "loss": 0.6467,
+      "step": 17622
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3100879487318118e-05,
+      "loss": 0.5953,
+      "step": 17623
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3100201784560129e-05,
+      "loss": 0.5712,
+      "step": 17624
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3099524066049105e-05,
+      "loss": 0.6674,
+      "step": 17625
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.309884633178849e-05,
+      "loss": 0.6254,
+      "step": 17626
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3098168581781732e-05,
+      "loss": 0.6102,
+      "step": 17627
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.309749081603227e-05,
+      "loss": 0.7056,
+      "step": 17628
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3096813034543549e-05,
+      "loss": 0.6067,
+      "step": 17629
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3096135237319017e-05,
+      "loss": 0.6315,
+      "step": 17630
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3095457424362112e-05,
+      "loss": 0.6056,
+      "step": 17631
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3094779595676284e-05,
+      "loss": 0.5646,
+      "step": 17632
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3094101751264971e-05,
+      "loss": 0.6168,
+      "step": 17633
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3093423891131622e-05,
+      "loss": 0.5495,
+      "step": 17634
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3092746015279683e-05,
+      "loss": 0.6126,
+      "step": 17635
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3092068123712593e-05,
+      "loss": 0.6039,
+      "step": 17636
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3091390216433802e-05,
+      "loss": 0.5971,
+      "step": 17637
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3090712293446748e-05,
+      "loss": 0.6072,
+      "step": 17638
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.309003435475488e-05,
+      "loss": 0.5947,
+      "step": 17639
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3089356400361647e-05,
+      "loss": 0.5984,
+      "step": 17640
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3088678430270484e-05,
+      "loss": 0.5569,
+      "step": 17641
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3088000444484844e-05,
+      "loss": 0.6608,
+      "step": 17642
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3087322443008173e-05,
+      "loss": 0.5485,
+      "step": 17643
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3086644425843908e-05,
+      "loss": 0.5576,
+      "step": 17644
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.30859663929955e-05,
+      "loss": 0.5912,
+      "step": 17645
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3085288344466394e-05,
+      "loss": 0.5426,
+      "step": 17646
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3084610280260033e-05,
+      "loss": 0.6051,
+      "step": 17647
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3083932200379865e-05,
+      "loss": 0.6202,
+      "step": 17648
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3083254104829331e-05,
+      "loss": 0.5615,
+      "step": 17649
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3082575993611881e-05,
+      "loss": 0.6359,
+      "step": 17650
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3081897866730962e-05,
+      "loss": 0.6496,
+      "step": 17651
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3081219724190015e-05,
+      "loss": 0.6074,
+      "step": 17652
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.308054156599249e-05,
+      "loss": 0.6079,
+      "step": 17653
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3079863392141832e-05,
+      "loss": 0.6123,
+      "step": 17654
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3079185202641483e-05,
+      "loss": 0.5892,
+      "step": 17655
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3078506997494896e-05,
+      "loss": 0.6332,
+      "step": 17656
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.307782877670551e-05,
+      "loss": 0.6331,
+      "step": 17657
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3077150540276773e-05,
+      "loss": 0.6103,
+      "step": 17658
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3076472288212138e-05,
+      "loss": 0.596,
+      "step": 17659
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3075794020515043e-05,
+      "loss": 0.6007,
+      "step": 17660
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3075115737188938e-05,
+      "loss": 0.5684,
+      "step": 17661
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.307443743823727e-05,
+      "loss": 0.5882,
+      "step": 17662
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3073759123663486e-05,
+      "loss": 0.6333,
+      "step": 17663
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3073080793471031e-05,
+      "loss": 0.5963,
+      "step": 17664
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3072402447663353e-05,
+      "loss": 0.6135,
+      "step": 17665
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3071724086243895e-05,
+      "loss": 0.541,
+      "step": 17666
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3071045709216113e-05,
+      "loss": 0.6086,
+      "step": 17667
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3070367316583444e-05,
+      "loss": 0.5797,
+      "step": 17668
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.306968890834934e-05,
+      "loss": 0.6036,
+      "step": 17669
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3069010484517251e-05,
+      "loss": 0.5496,
+      "step": 17670
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3068332045090617e-05,
+      "loss": 0.6018,
+      "step": 17671
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3067653590072889e-05,
+      "loss": 0.6048,
+      "step": 17672
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3066975119467518e-05,
+      "loss": 0.6712,
+      "step": 17673
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3066296633277947e-05,
+      "loss": 0.6282,
+      "step": 17674
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3065618131507623e-05,
+      "loss": 0.5992,
+      "step": 17675
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3064939614159998e-05,
+      "loss": 0.6037,
+      "step": 17676
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3064261081238514e-05,
+      "loss": 0.5627,
+      "step": 17677
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3063582532746626e-05,
+      "loss": 0.5624,
+      "step": 17678
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3062903968687775e-05,
+      "loss": 0.6,
+      "step": 17679
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3062225389065411e-05,
+      "loss": 0.6276,
+      "step": 17680
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3061546793882986e-05,
+      "loss": 0.6381,
+      "step": 17681
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3060868183143942e-05,
+      "loss": 0.6243,
+      "step": 17682
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3060189556851734e-05,
+      "loss": 0.6521,
+      "step": 17683
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3059510915009806e-05,
+      "loss": 0.6109,
+      "step": 17684
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3058832257621607e-05,
+      "loss": 0.6658,
+      "step": 17685
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3058153584690585e-05,
+      "loss": 0.5971,
+      "step": 17686
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3057474896220188e-05,
+      "loss": 0.6413,
+      "step": 17687
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3056796192213865e-05,
+      "loss": 0.5105,
+      "step": 17688
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3056117472675068e-05,
+      "loss": 0.6269,
+      "step": 17689
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.305543873760724e-05,
+      "loss": 0.5534,
+      "step": 17690
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3054759987013834e-05,
+      "loss": 0.5731,
+      "step": 17691
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3054081220898299e-05,
+      "loss": 0.6159,
+      "step": 17692
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3053402439264083e-05,
+      "loss": 0.6037,
+      "step": 17693
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3052723642114633e-05,
+      "loss": 0.5695,
+      "step": 17694
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.30520448294534e-05,
+      "loss": 0.5987,
+      "step": 17695
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3051366001283836e-05,
+      "loss": 0.589,
+      "step": 17696
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3050687157609387e-05,
+      "loss": 0.5775,
+      "step": 17697
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3050008298433503e-05,
+      "loss": 0.6067,
+      "step": 17698
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3049329423759631e-05,
+      "loss": 0.5675,
+      "step": 17699
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3048650533591227e-05,
+      "loss": 0.6282,
+      "step": 17700
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3047971627931734e-05,
+      "loss": 0.6054,
+      "step": 17701
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3047292706784607e-05,
+      "loss": 0.6461,
+      "step": 17702
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3046613770153291e-05,
+      "loss": 0.649,
+      "step": 17703
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.304593481804124e-05,
+      "loss": 0.5898,
+      "step": 17704
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3045255850451903e-05,
+      "loss": 0.6192,
+      "step": 17705
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.3044576867388728e-05,
+      "loss": 0.6222,
+      "step": 17706
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3043897868855165e-05,
+      "loss": 0.6316,
+      "step": 17707
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3043218854854666e-05,
+      "loss": 0.6459,
+      "step": 17708
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.304253982539068e-05,
+      "loss": 0.5923,
+      "step": 17709
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3041860780466663e-05,
+      "loss": 0.531,
+      "step": 17710
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3041181720086054e-05,
+      "loss": 0.6947,
+      "step": 17711
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3040502644252311e-05,
+      "loss": 0.6432,
+      "step": 17712
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3039823552968886e-05,
+      "loss": 0.6388,
+      "step": 17713
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3039144446239227e-05,
+      "loss": 0.5903,
+      "step": 17714
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3038465324066784e-05,
+      "loss": 0.5725,
+      "step": 17715
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.303778618645501e-05,
+      "loss": 0.5668,
+      "step": 17716
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3037107033407352e-05,
+      "loss": 0.6484,
+      "step": 17717
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3036427864927265e-05,
+      "loss": 0.6013,
+      "step": 17718
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.30357486810182e-05,
+      "loss": 0.5862,
+      "step": 17719
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3035069481683609e-05,
+      "loss": 0.5767,
+      "step": 17720
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3034390266926939e-05,
+      "loss": 0.6348,
+      "step": 17721
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3033711036751642e-05,
+      "loss": 0.6318,
+      "step": 17722
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3033031791161172e-05,
+      "loss": 0.559,
+      "step": 17723
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3032352530158978e-05,
+      "loss": 0.6337,
+      "step": 17724
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3031673253748517e-05,
+      "loss": 0.68,
+      "step": 17725
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3030993961933231e-05,
+      "loss": 0.5958,
+      "step": 17726
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3030314654716581e-05,
+      "loss": 0.5995,
+      "step": 17727
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.302963533210201e-05,
+      "loss": 0.5842,
+      "step": 17728
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.302895599409298e-05,
+      "loss": 0.6185,
+      "step": 17729
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3028276640692934e-05,
+      "loss": 0.6063,
+      "step": 17730
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3027597271905329e-05,
+      "loss": 0.6018,
+      "step": 17731
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3026917887733614e-05,
+      "loss": 0.5941,
+      "step": 17732
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3026238488181244e-05,
+      "loss": 0.6283,
+      "step": 17733
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3025559073251667e-05,
+      "loss": 0.5892,
+      "step": 17734
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3024879642948343e-05,
+      "loss": 0.5487,
+      "step": 17735
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3024200197274717e-05,
+      "loss": 0.6103,
+      "step": 17736
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3023520736234242e-05,
+      "loss": 0.5944,
+      "step": 17737
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.302284125983038e-05,
+      "loss": 0.6365,
+      "step": 17738
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.302216176806657e-05,
+      "loss": 0.5578,
+      "step": 17739
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3021482260946275e-05,
+      "loss": 0.6546,
+      "step": 17740
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.302080273847294e-05,
+      "loss": 0.5845,
+      "step": 17741
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3020123200650024e-05,
+      "loss": 0.5525,
+      "step": 17742
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3019443647480976e-05,
+      "loss": 0.6183,
+      "step": 17743
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3018764078969252e-05,
+      "loss": 0.6231,
+      "step": 17744
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3018084495118301e-05,
+      "loss": 0.6031,
+      "step": 17745
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3017404895931583e-05,
+      "loss": 0.5614,
+      "step": 17746
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3016725281412543e-05,
+      "loss": 0.618,
+      "step": 17747
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3016045651564642e-05,
+      "loss": 0.5994,
+      "step": 17748
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3015366006391325e-05,
+      "loss": 0.544,
+      "step": 17749
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3014686345896052e-05,
+      "loss": 0.5544,
+      "step": 17750
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3014006670082278e-05,
+      "loss": 0.5925,
+      "step": 17751
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.301332697895345e-05,
+      "loss": 0.6467,
+      "step": 17752
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3012647272513027e-05,
+      "loss": 0.6073,
+      "step": 17753
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.301196755076446e-05,
+      "loss": 0.5686,
+      "step": 17754
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3011287813711203e-05,
+      "loss": 0.6052,
+      "step": 17755
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3010608061356713e-05,
+      "loss": 0.5991,
+      "step": 17756
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.300992829370444e-05,
+      "loss": 0.6182,
+      "step": 17757
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3009248510757839e-05,
+      "loss": 0.6407,
+      "step": 17758
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3008568712520368e-05,
+      "loss": 0.5794,
+      "step": 17759
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3007888898995476e-05,
+      "loss": 0.5892,
+      "step": 17760
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3007209070186622e-05,
+      "loss": 0.5527,
+      "step": 17761
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3006529226097254e-05,
+      "loss": 0.6425,
+      "step": 17762
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3005849366730833e-05,
+      "loss": 0.6353,
+      "step": 17763
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.300516949209081e-05,
+      "loss": 0.5454,
+      "step": 17764
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3004489602180644e-05,
+      "loss": 0.591,
+      "step": 17765
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3003809697003782e-05,
+      "loss": 0.6171,
+      "step": 17766
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3003129776563687e-05,
+      "loss": 0.6027,
+      "step": 17767
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3002449840863806e-05,
+      "loss": 0.6082,
+      "step": 17768
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.30017698899076e-05,
+      "loss": 0.619,
+      "step": 17769
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3001089923698524e-05,
+      "loss": 0.5591,
+      "step": 17770
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.3000409942240028e-05,
+      "loss": 0.5692,
+      "step": 17771
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.299972994553557e-05,
+      "loss": 0.6183,
+      "step": 17772
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.299904993358861e-05,
+      "loss": 0.6514,
+      "step": 17773
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2998369906402595e-05,
+      "loss": 0.5792,
+      "step": 17774
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2997689863980986e-05,
+      "loss": 0.6567,
+      "step": 17775
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2997009806327236e-05,
+      "loss": 0.6133,
+      "step": 17776
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2996329733444803e-05,
+      "loss": 0.6341,
+      "step": 17777
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.299564964533714e-05,
+      "loss": 0.5608,
+      "step": 17778
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2994969542007705e-05,
+      "loss": 0.6188,
+      "step": 17779
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.299428942345995e-05,
+      "loss": 0.6056,
+      "step": 17780
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2993609289697338e-05,
+      "loss": 0.6647,
+      "step": 17781
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2992929140723318e-05,
+      "loss": 0.5583,
+      "step": 17782
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2992248976541348e-05,
+      "loss": 0.6611,
+      "step": 17783
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2991568797154886e-05,
+      "loss": 0.6286,
+      "step": 17784
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2990888602567384e-05,
+      "loss": 0.608,
+      "step": 17785
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2990208392782305e-05,
+      "loss": 0.5801,
+      "step": 17786
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2989528167803098e-05,
+      "loss": 0.5574,
+      "step": 17787
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2988847927633223e-05,
+      "loss": 0.5703,
+      "step": 17788
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.298816767227614e-05,
+      "loss": 0.5332,
+      "step": 17789
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2987487401735298e-05,
+      "loss": 0.6416,
+      "step": 17790
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2986807116014157e-05,
+      "loss": 0.6468,
+      "step": 17791
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2986126815116179e-05,
+      "loss": 0.6324,
+      "step": 17792
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2985446499044811e-05,
+      "loss": 0.647,
+      "step": 17793
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.298476616780352e-05,
+      "loss": 0.6589,
+      "step": 17794
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2984085821395756e-05,
+      "loss": 0.6234,
+      "step": 17795
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2983405459824977e-05,
+      "loss": 0.6218,
+      "step": 17796
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.298272508309464e-05,
+      "loss": 0.6088,
+      "step": 17797
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2982044691208205e-05,
+      "loss": 0.5711,
+      "step": 17798
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2981364284169127e-05,
+      "loss": 0.6298,
+      "step": 17799
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2980683861980866e-05,
+      "loss": 0.5169,
+      "step": 17800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2980003424646876e-05,
+      "loss": 0.5719,
+      "step": 17801
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2979322972170616e-05,
+      "loss": 0.5091,
+      "step": 17802
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2978642504555542e-05,
+      "loss": 0.6401,
+      "step": 17803
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2977962021805114e-05,
+      "loss": 0.5644,
+      "step": 17804
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2977281523922792e-05,
+      "loss": 0.6023,
+      "step": 17805
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2976601010912027e-05,
+      "loss": 0.5656,
+      "step": 17806
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2975920482776278e-05,
+      "loss": 0.6075,
+      "step": 17807
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.297523993951901e-05,
+      "loss": 0.6415,
+      "step": 17808
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2974559381143675e-05,
+      "loss": 0.6178,
+      "step": 17809
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2973878807653733e-05,
+      "loss": 0.5996,
+      "step": 17810
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2973198219052643e-05,
+      "loss": 0.547,
+      "step": 17811
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2972517615343859e-05,
+      "loss": 0.6208,
+      "step": 17812
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2971836996530844e-05,
+      "loss": 0.6855,
+      "step": 17813
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2971156362617054e-05,
+      "loss": 0.5701,
+      "step": 17814
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.297047571360595e-05,
+      "loss": 0.5952,
+      "step": 17815
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.296979504950099e-05,
+      "loss": 0.6891,
+      "step": 17816
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.296911437030563e-05,
+      "loss": 0.5949,
+      "step": 17817
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2968433676023329e-05,
+      "loss": 0.6116,
+      "step": 17818
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.296775296665755e-05,
+      "loss": 0.5795,
+      "step": 17819
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2967072242211746e-05,
+      "loss": 0.5999,
+      "step": 17820
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2966391502689382e-05,
+      "loss": 0.5919,
+      "step": 17821
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2965710748093913e-05,
+      "loss": 0.594,
+      "step": 17822
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2965029978428797e-05,
+      "loss": 0.6215,
+      "step": 17823
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.29643491936975e-05,
+      "loss": 0.5628,
+      "step": 17824
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2963668393903472e-05,
+      "loss": 0.6246,
+      "step": 17825
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2962987579050178e-05,
+      "loss": 0.6145,
+      "step": 17826
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2962306749141078e-05,
+      "loss": 0.6329,
+      "step": 17827
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2961625904179629e-05,
+      "loss": 0.5555,
+      "step": 17828
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2960945044169292e-05,
+      "loss": 0.5925,
+      "step": 17829
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2960264169113526e-05,
+      "loss": 0.5353,
+      "step": 17830
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2959583279015789e-05,
+      "loss": 0.605,
+      "step": 17831
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2958902373879543e-05,
+      "loss": 0.544,
+      "step": 17832
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2958221453708248e-05,
+      "loss": 0.6287,
+      "step": 17833
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2957540518505363e-05,
+      "loss": 0.5807,
+      "step": 17834
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.295685956827435e-05,
+      "loss": 0.5737,
+      "step": 17835
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2956178603018665e-05,
+      "loss": 0.573,
+      "step": 17836
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2955497622741772e-05,
+      "loss": 0.5508,
+      "step": 17837
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2954816627447134e-05,
+      "loss": 0.6299,
+      "step": 17838
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2954135617138201e-05,
+      "loss": 0.5785,
+      "step": 17839
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2953454591818444e-05,
+      "loss": 0.6016,
+      "step": 17840
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2952773551491317e-05,
+      "loss": 0.6051,
+      "step": 17841
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2952092496160285e-05,
+      "loss": 0.6147,
+      "step": 17842
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2951411425828805e-05,
+      "loss": 0.5935,
+      "step": 17843
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.295073034050034e-05,
+      "loss": 0.5285,
+      "step": 17844
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2950049240178349e-05,
+      "loss": 0.641,
+      "step": 17845
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2949368124866295e-05,
+      "loss": 0.5276,
+      "step": 17846
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2948686994567637e-05,
+      "loss": 0.5495,
+      "step": 17847
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2948005849285839e-05,
+      "loss": 0.6205,
+      "step": 17848
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2947324689024355e-05,
+      "loss": 0.6039,
+      "step": 17849
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2946643513786654e-05,
+      "loss": 0.5528,
+      "step": 17850
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2945962323576193e-05,
+      "loss": 0.5947,
+      "step": 17851
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2945281118396435e-05,
+      "loss": 0.5937,
+      "step": 17852
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.2944599898250843e-05,
+      "loss": 0.6076,
+      "step": 17853
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2943918663142876e-05,
+      "loss": 0.5552,
+      "step": 17854
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2943237413075996e-05,
+      "loss": 0.6077,
+      "step": 17855
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2942556148053661e-05,
+      "loss": 0.5474,
+      "step": 17856
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2941874868079342e-05,
+      "loss": 0.5436,
+      "step": 17857
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2941193573156492e-05,
+      "loss": 0.6511,
+      "step": 17858
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2940512263288577e-05,
+      "loss": 0.5956,
+      "step": 17859
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.293983093847906e-05,
+      "loss": 0.5952,
+      "step": 17860
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2939149598731397e-05,
+      "loss": 0.5824,
+      "step": 17861
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2938468244049061e-05,
+      "loss": 0.6141,
+      "step": 17862
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.29377868744355e-05,
+      "loss": 0.5852,
+      "step": 17863
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2937105489894184e-05,
+      "loss": 0.5715,
+      "step": 17864
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.293642409042858e-05,
+      "loss": 0.6462,
+      "step": 17865
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.293574267604214e-05,
+      "loss": 0.6317,
+      "step": 17866
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2935061246738333e-05,
+      "loss": 0.6419,
+      "step": 17867
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2934379802520622e-05,
+      "loss": 0.6076,
+      "step": 17868
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2933698343392467e-05,
+      "loss": 0.606,
+      "step": 17869
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2933016869357332e-05,
+      "loss": 0.6273,
+      "step": 17870
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2932335380418679e-05,
+      "loss": 0.5947,
+      "step": 17871
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2931653876579969e-05,
+      "loss": 0.5821,
+      "step": 17872
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2930972357844671e-05,
+      "loss": 0.6018,
+      "step": 17873
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2930290824216241e-05,
+      "loss": 0.5768,
+      "step": 17874
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2929609275698145e-05,
+      "loss": 0.604,
+      "step": 17875
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2928927712293849e-05,
+      "loss": 0.6378,
+      "step": 17876
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.292824613400681e-05,
+      "loss": 0.5673,
+      "step": 17877
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2927564540840497e-05,
+      "loss": 0.5962,
+      "step": 17878
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2926882932798371e-05,
+      "loss": 0.6102,
+      "step": 17879
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2926201309883896e-05,
+      "loss": 0.5716,
+      "step": 17880
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2925519672100536e-05,
+      "loss": 0.5964,
+      "step": 17881
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2924838019451753e-05,
+      "loss": 0.5931,
+      "step": 17882
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2924156351941011e-05,
+      "loss": 0.6967,
+      "step": 17883
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2923474669571774e-05,
+      "loss": 0.5536,
+      "step": 17884
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2922792972347507e-05,
+      "loss": 0.5907,
+      "step": 17885
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2922111260271672e-05,
+      "loss": 0.577,
+      "step": 17886
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2921429533347733e-05,
+      "loss": 0.6202,
+      "step": 17887
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2920747791579155e-05,
+      "loss": 0.5689,
+      "step": 17888
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2920066034969405e-05,
+      "loss": 0.5895,
+      "step": 17889
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2919384263521941e-05,
+      "loss": 0.6268,
+      "step": 17890
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.291870247724023e-05,
+      "loss": 0.6154,
+      "step": 17891
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2918020676127743e-05,
+      "loss": 0.5981,
+      "step": 17892
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2917338860187931e-05,
+      "loss": 0.617,
+      "step": 17893
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2916657029424271e-05,
+      "loss": 0.617,
+      "step": 17894
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2915975183840219e-05,
+      "loss": 0.5727,
+      "step": 17895
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2915293323439244e-05,
+      "loss": 0.6197,
+      "step": 17896
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2914611448224813e-05,
+      "loss": 0.5702,
+      "step": 17897
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2913929558200385e-05,
+      "loss": 0.6282,
+      "step": 17898
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2913247653369427e-05,
+      "loss": 0.6202,
+      "step": 17899
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2912565733735404e-05,
+      "loss": 0.6339,
+      "step": 17900
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2911883799301786e-05,
+      "loss": 0.5487,
+      "step": 17901
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2911201850072029e-05,
+      "loss": 0.6203,
+      "step": 17902
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2910519886049605e-05,
+      "loss": 0.5761,
+      "step": 17903
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2909837907237975e-05,
+      "loss": 0.5769,
+      "step": 17904
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.290915591364061e-05,
+      "loss": 0.6257,
+      "step": 17905
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2908473905260968e-05,
+      "loss": 0.6046,
+      "step": 17906
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2907791882102518e-05,
+      "loss": 0.5615,
+      "step": 17907
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.290710984416873e-05,
+      "loss": 0.6068,
+      "step": 17908
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2906427791463064e-05,
+      "loss": 0.6179,
+      "step": 17909
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2905745723988983e-05,
+      "loss": 0.5867,
+      "step": 17910
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2905063641749962e-05,
+      "loss": 0.6262,
+      "step": 17911
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2904381544749458e-05,
+      "loss": 0.6162,
+      "step": 17912
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2903699432990946e-05,
+      "loss": 0.6481,
+      "step": 17913
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2903017306477882e-05,
+      "loss": 0.6477,
+      "step": 17914
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2902335165213737e-05,
+      "loss": 0.5396,
+      "step": 17915
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2901653009201982e-05,
+      "loss": 0.6136,
+      "step": 17916
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2900970838446071e-05,
+      "loss": 0.6359,
+      "step": 17917
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2900288652949484e-05,
+      "loss": 0.5959,
+      "step": 17918
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2899606452715679e-05,
+      "loss": 0.6064,
+      "step": 17919
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2898924237748125e-05,
+      "loss": 0.5462,
+      "step": 17920
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2898242008050285e-05,
+      "loss": 0.631,
+      "step": 17921
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2897559763625633e-05,
+      "loss": 0.6158,
+      "step": 17922
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2896877504477626e-05,
+      "loss": 0.618,
+      "step": 17923
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2896195230609741e-05,
+      "loss": 0.5947,
+      "step": 17924
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2895512942025436e-05,
+      "loss": 0.6188,
+      "step": 17925
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2894830638728183e-05,
+      "loss": 0.6725,
+      "step": 17926
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2894148320721448e-05,
+      "loss": 0.6238,
+      "step": 17927
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2893465988008698e-05,
+      "loss": 0.6178,
+      "step": 17928
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2892783640593401e-05,
+      "loss": 0.6455,
+      "step": 17929
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2892101278479021e-05,
+      "loss": 0.6065,
+      "step": 17930
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2891418901669028e-05,
+      "loss": 0.6225,
+      "step": 17931
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.289073651016689e-05,
+      "loss": 0.5488,
+      "step": 17932
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.289005410397607e-05,
+      "loss": 0.6086,
+      "step": 17933
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288937168310004e-05,
+      "loss": 0.5664,
+      "step": 17934
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288868924754227e-05,
+      "loss": 0.6066,
+      "step": 17935
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288800679730622e-05,
+      "loss": 0.5803,
+      "step": 17936
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2887324332395363e-05,
+      "loss": 0.594,
+      "step": 17937
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2886641852813165e-05,
+      "loss": 0.6344,
+      "step": 17938
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2885959358563097e-05,
+      "loss": 0.5718,
+      "step": 17939
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288527684964862e-05,
+      "loss": 0.6216,
+      "step": 17940
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288459432607321e-05,
+      "loss": 0.6724,
+      "step": 17941
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2883911787840329e-05,
+      "loss": 0.6282,
+      "step": 17942
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288322923495345e-05,
+      "loss": 0.5937,
+      "step": 17943
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2882546667416036e-05,
+      "loss": 0.6359,
+      "step": 17944
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288186408523156e-05,
+      "loss": 0.5833,
+      "step": 17945
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.288118148840349e-05,
+      "loss": 0.5755,
+      "step": 17946
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2880498876935288e-05,
+      "loss": 0.5916,
+      "step": 17947
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.287981625083043e-05,
+      "loss": 0.5844,
+      "step": 17948
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2879133610092384e-05,
+      "loss": 0.5851,
+      "step": 17949
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2878450954724615e-05,
+      "loss": 0.6017,
+      "step": 17950
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2877768284730596e-05,
+      "loss": 0.6555,
+      "step": 17951
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.287708560011379e-05,
+      "loss": 0.5906,
+      "step": 17952
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2876402900877671e-05,
+      "loss": 0.5589,
+      "step": 17953
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2875720187025709e-05,
+      "loss": 0.5913,
+      "step": 17954
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2875037458561368e-05,
+      "loss": 0.5909,
+      "step": 17955
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.287435471548812e-05,
+      "loss": 0.6037,
+      "step": 17956
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2873671957809435e-05,
+      "loss": 0.6661,
+      "step": 17957
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2872989185528781e-05,
+      "loss": 0.5743,
+      "step": 17958
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2872306398649626e-05,
+      "loss": 0.5877,
+      "step": 17959
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2871623597175441e-05,
+      "loss": 0.63,
+      "step": 17960
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2870940781109698e-05,
+      "loss": 0.6139,
+      "step": 17961
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2870257950455865e-05,
+      "loss": 0.5833,
+      "step": 17962
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2869575105217407e-05,
+      "loss": 0.6424,
+      "step": 17963
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2868892245397797e-05,
+      "loss": 0.6096,
+      "step": 17964
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2868209371000508e-05,
+      "loss": 0.548,
+      "step": 17965
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2867526482029008e-05,
+      "loss": 0.6091,
+      "step": 17966
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.286684357848676e-05,
+      "loss": 0.5972,
+      "step": 17967
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2866160660377248e-05,
+      "loss": 0.5531,
+      "step": 17968
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.286547772770393e-05,
+      "loss": 0.5764,
+      "step": 17969
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2864794780470282e-05,
+      "loss": 0.5852,
+      "step": 17970
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2864111818679769e-05,
+      "loss": 0.5894,
+      "step": 17971
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.286342884233587e-05,
+      "loss": 0.6277,
+      "step": 17972
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.286274585144205e-05,
+      "loss": 0.6067,
+      "step": 17973
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2862062846001777e-05,
+      "loss": 0.5822,
+      "step": 17974
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2861379826018525e-05,
+      "loss": 0.641,
+      "step": 17975
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2860696791495766e-05,
+      "loss": 0.5847,
+      "step": 17976
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2860013742436968e-05,
+      "loss": 0.6282,
+      "step": 17977
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2859330678845604e-05,
+      "loss": 0.5561,
+      "step": 17978
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2858647600725142e-05,
+      "loss": 0.5949,
+      "step": 17979
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2857964508079054e-05,
+      "loss": 0.6261,
+      "step": 17980
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2857281400910812e-05,
+      "loss": 0.5747,
+      "step": 17981
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2856598279223887e-05,
+      "loss": 0.6171,
+      "step": 17982
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2855915143021746e-05,
+      "loss": 0.5553,
+      "step": 17983
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2855231992307868e-05,
+      "loss": 0.5777,
+      "step": 17984
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2854548827085718e-05,
+      "loss": 0.529,
+      "step": 17985
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2853865647358772e-05,
+      "loss": 0.6129,
+      "step": 17986
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2853182453130495e-05,
+      "loss": 0.556,
+      "step": 17987
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2852499244404362e-05,
+      "loss": 0.6445,
+      "step": 17988
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.285181602118385e-05,
+      "loss": 0.6538,
+      "step": 17989
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2851132783472422e-05,
+      "loss": 0.5766,
+      "step": 17990
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2850449531273553e-05,
+      "loss": 0.6204,
+      "step": 17991
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2849766264590718e-05,
+      "loss": 0.6145,
+      "step": 17992
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.284908298342738e-05,
+      "loss": 0.6776,
+      "step": 17993
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.284839968778702e-05,
+      "loss": 0.6725,
+      "step": 17994
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2847716377673112e-05,
+      "loss": 0.6081,
+      "step": 17995
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2847033053089118e-05,
+      "loss": 0.6173,
+      "step": 17996
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2846349714038517e-05,
+      "loss": 0.5815,
+      "step": 17997
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.284566636052478e-05,
+      "loss": 0.5911,
+      "step": 17998
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2844982992551379e-05,
+      "loss": 0.5783,
+      "step": 17999
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.2844299610121785e-05,
+      "loss": 0.5817,
+      "step": 18000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.284361621323947e-05,
+      "loss": 0.5642,
+      "step": 18001
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2842932801907909e-05,
+      "loss": 0.6432,
+      "step": 18002
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2842249376130575e-05,
+      "loss": 0.6437,
+      "step": 18003
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.284156593591094e-05,
+      "loss": 0.593,
+      "step": 18004
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2840882481252477e-05,
+      "loss": 0.6153,
+      "step": 18005
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2840199012158653e-05,
+      "loss": 0.6379,
+      "step": 18006
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2839515528632951e-05,
+      "loss": 0.6351,
+      "step": 18007
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2838832030678838e-05,
+      "loss": 0.5558,
+      "step": 18008
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2838148518299787e-05,
+      "loss": 0.5965,
+      "step": 18009
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2837464991499273e-05,
+      "loss": 0.5534,
+      "step": 18010
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2836781450280768e-05,
+      "loss": 0.5889,
+      "step": 18011
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2836097894647745e-05,
+      "loss": 0.5978,
+      "step": 18012
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2835414324603679e-05,
+      "loss": 0.5996,
+      "step": 18013
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2834730740152044e-05,
+      "loss": 0.5918,
+      "step": 18014
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.283404714129631e-05,
+      "loss": 0.5732,
+      "step": 18015
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2833363528039953e-05,
+      "loss": 0.5789,
+      "step": 18016
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2832679900386444e-05,
+      "loss": 0.5361,
+      "step": 18017
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2831996258339263e-05,
+      "loss": 0.6232,
+      "step": 18018
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.283131260190188e-05,
+      "loss": 0.6309,
+      "step": 18019
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2830628931077764e-05,
+      "loss": 0.6372,
+      "step": 18020
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2829945245870394e-05,
+      "loss": 0.6404,
+      "step": 18021
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2829261546283246e-05,
+      "loss": 0.6041,
+      "step": 18022
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.282857783231979e-05,
+      "loss": 0.6068,
+      "step": 18023
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2827894103983503e-05,
+      "loss": 0.5764,
+      "step": 18024
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2827210361277857e-05,
+      "loss": 0.5807,
+      "step": 18025
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2826526604206326e-05,
+      "loss": 0.5533,
+      "step": 18026
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2825842832772385e-05,
+      "loss": 0.5989,
+      "step": 18027
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2825159046979512e-05,
+      "loss": 0.6242,
+      "step": 18028
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2824475246831176e-05,
+      "loss": 0.6078,
+      "step": 18029
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2823791432330857e-05,
+      "loss": 0.5924,
+      "step": 18030
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2823107603482023e-05,
+      "loss": 0.5771,
+      "step": 18031
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2822423760288155e-05,
+      "loss": 0.6692,
+      "step": 18032
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2821739902752724e-05,
+      "loss": 0.6346,
+      "step": 18033
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2821056030879206e-05,
+      "loss": 0.6326,
+      "step": 18034
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2820372144671078e-05,
+      "loss": 0.632,
+      "step": 18035
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2819688244131811e-05,
+      "loss": 0.5556,
+      "step": 18036
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2819004329264883e-05,
+      "loss": 0.5778,
+      "step": 18037
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2818320400073772e-05,
+      "loss": 0.6178,
+      "step": 18038
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2817636456561945e-05,
+      "loss": 0.6321,
+      "step": 18039
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2816952498732882e-05,
+      "loss": 0.6363,
+      "step": 18040
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2816268526590062e-05,
+      "loss": 0.6013,
+      "step": 18041
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2815584540136952e-05,
+      "loss": 0.6068,
+      "step": 18042
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2814900539377038e-05,
+      "loss": 0.6529,
+      "step": 18043
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2814216524313784e-05,
+      "loss": 0.6063,
+      "step": 18044
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2813532494950676e-05,
+      "loss": 0.6451,
+      "step": 18045
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2812848451291184e-05,
+      "loss": 0.6372,
+      "step": 18046
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2812164393338784e-05,
+      "loss": 0.6332,
+      "step": 18047
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2811480321096954e-05,
+      "loss": 0.6327,
+      "step": 18048
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.281079623456917e-05,
+      "loss": 0.578,
+      "step": 18049
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2810112133758905e-05,
+      "loss": 0.5678,
+      "step": 18050
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2809428018669641e-05,
+      "loss": 0.6405,
+      "step": 18051
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2808743889304847e-05,
+      "loss": 0.601,
+      "step": 18052
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2808059745668003e-05,
+      "loss": 0.6045,
+      "step": 18053
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2807375587762586e-05,
+      "loss": 0.6043,
+      "step": 18054
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2806691415592071e-05,
+      "loss": 0.7348,
+      "step": 18055
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2806007229159937e-05,
+      "loss": 0.6141,
+      "step": 18056
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2805323028469655e-05,
+      "loss": 0.5381,
+      "step": 18057
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.280463881352471e-05,
+      "loss": 0.5946,
+      "step": 18058
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2803954584328571e-05,
+      "loss": 0.6012,
+      "step": 18059
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.280327034088472e-05,
+      "loss": 0.5461,
+      "step": 18060
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2802586083196626e-05,
+      "loss": 0.5732,
+      "step": 18061
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2801901811267773e-05,
+      "loss": 0.5928,
+      "step": 18062
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2801217525101638e-05,
+      "loss": 0.57,
+      "step": 18063
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2800533224701697e-05,
+      "loss": 0.5822,
+      "step": 18064
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2799848910071426e-05,
+      "loss": 0.576,
+      "step": 18065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.27991645812143e-05,
+      "loss": 0.6265,
+      "step": 18066
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.27984802381338e-05,
+      "loss": 0.6121,
+      "step": 18067
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2797795880833405e-05,
+      "loss": 0.601,
+      "step": 18068
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.279711150931659e-05,
+      "loss": 0.5866,
+      "step": 18069
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.279642712358683e-05,
+      "loss": 0.6587,
+      "step": 18070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2795742723647604e-05,
+      "loss": 0.5745,
+      "step": 18071
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.279505830950239e-05,
+      "loss": 0.575,
+      "step": 18072
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2794373881154672e-05,
+      "loss": 0.5715,
+      "step": 18073
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2793689438607915e-05,
+      "loss": 0.6152,
+      "step": 18074
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2793004981865609e-05,
+      "loss": 0.656,
+      "step": 18075
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2792320510931222e-05,
+      "loss": 0.5694,
+      "step": 18076
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.279163602580824e-05,
+      "loss": 0.6208,
+      "step": 18077
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2790951526500138e-05,
+      "loss": 0.5974,
+      "step": 18078
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2790267013010394e-05,
+      "loss": 0.6306,
+      "step": 18079
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2789582485342482e-05,
+      "loss": 0.6297,
+      "step": 18080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2788897943499889e-05,
+      "loss": 0.5792,
+      "step": 18081
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2788213387486085e-05,
+      "loss": 0.5705,
+      "step": 18082
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2787528817304555e-05,
+      "loss": 0.6013,
+      "step": 18083
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2786844232958775e-05,
+      "loss": 0.633,
+      "step": 18084
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2786159634452222e-05,
+      "loss": 0.6121,
+      "step": 18085
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2785475021788374e-05,
+      "loss": 0.5922,
+      "step": 18086
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2784790394970717e-05,
+      "loss": 0.5967,
+      "step": 18087
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.278410575400272e-05,
+      "loss": 0.6309,
+      "step": 18088
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2783421098887869e-05,
+      "loss": 0.613,
+      "step": 18089
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2782736429629638e-05,
+      "loss": 0.5458,
+      "step": 18090
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2782051746231507e-05,
+      "loss": 0.6015,
+      "step": 18091
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.278136704869696e-05,
+      "loss": 0.6431,
+      "step": 18092
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.278068233702947e-05,
+      "loss": 0.6394,
+      "step": 18093
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2779997611232522e-05,
+      "loss": 0.6235,
+      "step": 18094
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2779312871309588e-05,
+      "loss": 0.6333,
+      "step": 18095
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2778628117264154e-05,
+      "loss": 0.6121,
+      "step": 18096
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2777943349099695e-05,
+      "loss": 0.5998,
+      "step": 18097
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2777258566819695e-05,
+      "loss": 0.6245,
+      "step": 18098
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2776573770427626e-05,
+      "loss": 0.5569,
+      "step": 18099
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2775888959926979e-05,
+      "loss": 0.6064,
+      "step": 18100
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2775204135321222e-05,
+      "loss": 0.5721,
+      "step": 18101
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2774519296613841e-05,
+      "loss": 0.6388,
+      "step": 18102
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2773834443808318e-05,
+      "loss": 0.5499,
+      "step": 18103
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2773149576908123e-05,
+      "loss": 0.5789,
+      "step": 18104
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2772464695916748e-05,
+      "loss": 0.6042,
+      "step": 18105
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2771779800837667e-05,
+      "loss": 0.6615,
+      "step": 18106
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2771094891674363e-05,
+      "loss": 0.605,
+      "step": 18107
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2770409968430313e-05,
+      "loss": 0.7157,
+      "step": 18108
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2769725031108997e-05,
+      "loss": 0.6386,
+      "step": 18109
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2769040079713897e-05,
+      "loss": 0.5834,
+      "step": 18110
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2768355114248493e-05,
+      "loss": 0.6365,
+      "step": 18111
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.276767013471627e-05,
+      "loss": 0.5935,
+      "step": 18112
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.27669851411207e-05,
+      "loss": 0.6212,
+      "step": 18113
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.276630013346527e-05,
+      "loss": 0.6315,
+      "step": 18114
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2765615111753463e-05,
+      "loss": 0.5511,
+      "step": 18115
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2764930075988748e-05,
+      "loss": 0.5869,
+      "step": 18116
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.276424502617462e-05,
+      "loss": 0.5746,
+      "step": 18117
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2763559962314549e-05,
+      "loss": 0.5754,
+      "step": 18118
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2762874884412022e-05,
+      "loss": 0.6431,
+      "step": 18119
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2762189792470518e-05,
+      "loss": 0.6343,
+      "step": 18120
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.276150468649352e-05,
+      "loss": 0.6129,
+      "step": 18121
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2760819566484509e-05,
+      "loss": 0.5708,
+      "step": 18122
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2760134432446962e-05,
+      "loss": 0.6022,
+      "step": 18123
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2759449284384367e-05,
+      "loss": 0.5867,
+      "step": 18124
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2758764122300198e-05,
+      "loss": 0.6194,
+      "step": 18125
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2758078946197943e-05,
+      "loss": 0.6506,
+      "step": 18126
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.275739375608108e-05,
+      "loss": 0.5553,
+      "step": 18127
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2756708551953092e-05,
+      "loss": 0.6273,
+      "step": 18128
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2756023333817462e-05,
+      "loss": 0.6128,
+      "step": 18129
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.275533810167767e-05,
+      "loss": 0.6033,
+      "step": 18130
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2754652855537197e-05,
+      "loss": 0.6112,
+      "step": 18131
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2753967595399526e-05,
+      "loss": 0.5659,
+      "step": 18132
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2753282321268141e-05,
+      "loss": 0.579,
+      "step": 18133
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.275259703314652e-05,
+      "loss": 0.5808,
+      "step": 18134
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2751911731038149e-05,
+      "loss": 0.5892,
+      "step": 18135
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.275122641494651e-05,
+      "loss": 0.5208,
+      "step": 18136
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2750541084875081e-05,
+      "loss": 0.5325,
+      "step": 18137
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2749855740827348e-05,
+      "loss": 0.6291,
+      "step": 18138
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2749170382806792e-05,
+      "loss": 0.5881,
+      "step": 18139
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2748485010816896e-05,
+      "loss": 0.5826,
+      "step": 18140
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2747799624861146e-05,
+      "loss": 0.6259,
+      "step": 18141
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2747114224943018e-05,
+      "loss": 0.5614,
+      "step": 18142
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2746428811066001e-05,
+      "loss": 0.6286,
+      "step": 18143
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.274574338323357e-05,
+      "loss": 0.5924,
+      "step": 18144
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2745057941449215e-05,
+      "loss": 0.5895,
+      "step": 18145
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.2744372485716418e-05,
+      "loss": 0.6356,
+      "step": 18146
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.274368701603866e-05,
+      "loss": 0.5715,
+      "step": 18147
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2743001532419423e-05,
+      "loss": 0.5993,
+      "step": 18148
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2742316034862196e-05,
+      "loss": 0.6098,
+      "step": 18149
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2741630523370454e-05,
+      "loss": 0.5601,
+      "step": 18150
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2740944997947687e-05,
+      "loss": 0.5676,
+      "step": 18151
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2740259458597375e-05,
+      "loss": 0.601,
+      "step": 18152
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2739573905323002e-05,
+      "loss": 0.5753,
+      "step": 18153
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2738888338128053e-05,
+      "loss": 0.5827,
+      "step": 18154
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.273820275701601e-05,
+      "loss": 0.6243,
+      "step": 18155
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2737517161990353e-05,
+      "loss": 0.5627,
+      "step": 18156
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2736831553054575e-05,
+      "loss": 0.6298,
+      "step": 18157
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2736145930212152e-05,
+      "loss": 0.5861,
+      "step": 18158
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2735460293466568e-05,
+      "loss": 0.645,
+      "step": 18159
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2734774642821313e-05,
+      "loss": 0.5314,
+      "step": 18160
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2734088978279863e-05,
+      "loss": 0.6007,
+      "step": 18161
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.273340329984571e-05,
+      "loss": 0.5518,
+      "step": 18162
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.273271760752233e-05,
+      "loss": 0.6108,
+      "step": 18163
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2732031901313215e-05,
+      "loss": 0.5856,
+      "step": 18164
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2731346181221846e-05,
+      "loss": 0.5802,
+      "step": 18165
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2730660447251706e-05,
+      "loss": 0.5944,
+      "step": 18166
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.272997469940628e-05,
+      "loss": 0.5924,
+      "step": 18167
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2729288937689055e-05,
+      "loss": 0.6118,
+      "step": 18168
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.272860316210351e-05,
+      "loss": 0.6489,
+      "step": 18169
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2727917372653138e-05,
+      "loss": 0.6418,
+      "step": 18170
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2727231569341414e-05,
+      "loss": 0.5675,
+      "step": 18171
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.272654575217183e-05,
+      "loss": 0.657,
+      "step": 18172
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2725859921147872e-05,
+      "loss": 0.5857,
+      "step": 18173
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2725174076273015e-05,
+      "loss": 0.6558,
+      "step": 18174
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2724488217550755e-05,
+      "loss": 0.5949,
+      "step": 18175
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2723802344984573e-05,
+      "loss": 0.6749,
+      "step": 18176
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2723116458577951e-05,
+      "loss": 0.6435,
+      "step": 18177
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2722430558334375e-05,
+      "loss": 0.6195,
+      "step": 18178
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2721744644257336e-05,
+      "loss": 0.6274,
+      "step": 18179
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2721058716350315e-05,
+      "loss": 0.6116,
+      "step": 18180
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2720372774616797e-05,
+      "loss": 0.6598,
+      "step": 18181
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2719686819060266e-05,
+      "loss": 0.6023,
+      "step": 18182
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2719000849684211e-05,
+      "loss": 0.6184,
+      "step": 18183
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2718314866492118e-05,
+      "loss": 0.5917,
+      "step": 18184
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2717628869487469e-05,
+      "loss": 0.6222,
+      "step": 18185
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2716942858673753e-05,
+      "loss": 0.6316,
+      "step": 18186
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2716256834054455e-05,
+      "loss": 0.5794,
+      "step": 18187
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2715570795633059e-05,
+      "loss": 0.5603,
+      "step": 18188
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2714884743413054e-05,
+      "loss": 0.6269,
+      "step": 18189
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2714198677397924e-05,
+      "loss": 0.5858,
+      "step": 18190
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2713512597591155e-05,
+      "loss": 0.6519,
+      "step": 18191
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2712826503996236e-05,
+      "loss": 0.6103,
+      "step": 18192
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2712140396616648e-05,
+      "loss": 0.5849,
+      "step": 18193
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2711454275455883e-05,
+      "loss": 0.6674,
+      "step": 18194
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2710768140517424e-05,
+      "loss": 0.5683,
+      "step": 18195
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.271008199180476e-05,
+      "loss": 0.5667,
+      "step": 18196
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2709395829321372e-05,
+      "loss": 0.6107,
+      "step": 18197
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2708709653070753e-05,
+      "loss": 0.614,
+      "step": 18198
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2708023463056382e-05,
+      "loss": 0.5825,
+      "step": 18199
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2707337259281758e-05,
+      "loss": 0.5987,
+      "step": 18200
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2706651041750355e-05,
+      "loss": 0.6305,
+      "step": 18201
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2705964810465667e-05,
+      "loss": 0.6149,
+      "step": 18202
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.270527856543118e-05,
+      "loss": 0.5664,
+      "step": 18203
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.270459230665038e-05,
+      "loss": 0.5875,
+      "step": 18204
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2703906034126754e-05,
+      "loss": 0.5848,
+      "step": 18205
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2703219747863791e-05,
+      "loss": 0.54,
+      "step": 18206
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2702533447864975e-05,
+      "loss": 0.6302,
+      "step": 18207
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2701847134133796e-05,
+      "loss": 0.6146,
+      "step": 18208
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.270116080667374e-05,
+      "loss": 0.5906,
+      "step": 18209
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2700474465488295e-05,
+      "loss": 0.5969,
+      "step": 18210
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2699788110580948e-05,
+      "loss": 0.5633,
+      "step": 18211
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.269910174195519e-05,
+      "loss": 0.6222,
+      "step": 18212
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2698415359614502e-05,
+      "loss": 0.6282,
+      "step": 18213
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2697728963562375e-05,
+      "loss": 0.5856,
+      "step": 18214
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26970425538023e-05,
+      "loss": 0.624,
+      "step": 18215
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2696356130337761e-05,
+      "loss": 0.6668,
+      "step": 18216
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2695669693172246e-05,
+      "loss": 0.6264,
+      "step": 18217
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2694983242309245e-05,
+      "loss": 0.6196,
+      "step": 18218
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2694296777752244e-05,
+      "loss": 0.5557,
+      "step": 18219
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.269361029950473e-05,
+      "loss": 0.5784,
+      "step": 18220
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2692923807570195e-05,
+      "loss": 0.6122,
+      "step": 18221
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2692237301952126e-05,
+      "loss": 0.5903,
+      "step": 18222
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.269155078265401e-05,
+      "loss": 0.5244,
+      "step": 18223
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2690864249679335e-05,
+      "loss": 0.6054,
+      "step": 18224
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2690177703031595e-05,
+      "loss": 0.584,
+      "step": 18225
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.268949114271427e-05,
+      "loss": 0.6009,
+      "step": 18226
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2688804568730854e-05,
+      "loss": 0.6337,
+      "step": 18227
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2688117981084835e-05,
+      "loss": 0.5673,
+      "step": 18228
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26874313797797e-05,
+      "loss": 0.5833,
+      "step": 18229
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2686744764818944e-05,
+      "loss": 0.6294,
+      "step": 18230
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2686058136206046e-05,
+      "loss": 0.6178,
+      "step": 18231
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26853714939445e-05,
+      "loss": 0.5643,
+      "step": 18232
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2684684838037798e-05,
+      "loss": 0.582,
+      "step": 18233
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2683998168489424e-05,
+      "loss": 0.5697,
+      "step": 18234
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2683311485302871e-05,
+      "loss": 0.6307,
+      "step": 18235
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2682624788481626e-05,
+      "loss": 0.571,
+      "step": 18236
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2681938078029177e-05,
+      "loss": 0.5973,
+      "step": 18237
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2681251353949015e-05,
+      "loss": 0.5978,
+      "step": 18238
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2680564616244631e-05,
+      "loss": 0.6078,
+      "step": 18239
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2679877864919512e-05,
+      "loss": 0.7016,
+      "step": 18240
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2679191099977151e-05,
+      "loss": 0.6436,
+      "step": 18241
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2678504321421033e-05,
+      "loss": 0.6374,
+      "step": 18242
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.267781752925465e-05,
+      "loss": 0.5992,
+      "step": 18243
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2677130723481494e-05,
+      "loss": 0.5967,
+      "step": 18244
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2676443904105052e-05,
+      "loss": 0.5449,
+      "step": 18245
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2675757071128816e-05,
+      "loss": 0.6292,
+      "step": 18246
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2675070224556269e-05,
+      "loss": 0.5886,
+      "step": 18247
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2674383364390911e-05,
+      "loss": 0.5561,
+      "step": 18248
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.267369649063623e-05,
+      "loss": 0.5798,
+      "step": 18249
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2673009603295712e-05,
+      "loss": 0.5758,
+      "step": 18250
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.267232270237285e-05,
+      "loss": 0.5852,
+      "step": 18251
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2671635787871131e-05,
+      "loss": 0.631,
+      "step": 18252
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2670948859794051e-05,
+      "loss": 0.5428,
+      "step": 18253
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2670261918145097e-05,
+      "loss": 0.5816,
+      "step": 18254
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.266957496292776e-05,
+      "loss": 0.6029,
+      "step": 18255
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.266888799414553e-05,
+      "loss": 0.6722,
+      "step": 18256
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2668201011801901e-05,
+      "loss": 0.5561,
+      "step": 18257
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2667514015900359e-05,
+      "loss": 0.5341,
+      "step": 18258
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2666827006444397e-05,
+      "loss": 0.633,
+      "step": 18259
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2666139983437506e-05,
+      "loss": 0.5912,
+      "step": 18260
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2665452946883176e-05,
+      "loss": 0.6021,
+      "step": 18261
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2664765896784903e-05,
+      "loss": 0.5955,
+      "step": 18262
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2664078833146172e-05,
+      "loss": 0.6014,
+      "step": 18263
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2663391755970473e-05,
+      "loss": 0.5794,
+      "step": 18264
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2662704665261306e-05,
+      "loss": 0.586,
+      "step": 18265
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2662017561022152e-05,
+      "loss": 0.6149,
+      "step": 18266
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2661330443256508e-05,
+      "loss": 0.5856,
+      "step": 18267
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2660643311967867e-05,
+      "loss": 0.6288,
+      "step": 18268
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2659956167159715e-05,
+      "loss": 0.6423,
+      "step": 18269
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.265926900883555e-05,
+      "loss": 0.5856,
+      "step": 18270
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2658581836998859e-05,
+      "loss": 0.6152,
+      "step": 18271
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2657894651653136e-05,
+      "loss": 0.6279,
+      "step": 18272
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.265720745280187e-05,
+      "loss": 0.5736,
+      "step": 18273
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2656520240448557e-05,
+      "loss": 0.5508,
+      "step": 18274
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2655833014596684e-05,
+      "loss": 0.5972,
+      "step": 18275
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.265514577524975e-05,
+      "loss": 0.6001,
+      "step": 18276
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2654458522411237e-05,
+      "loss": 0.6308,
+      "step": 18277
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2653771256084646e-05,
+      "loss": 0.5543,
+      "step": 18278
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2653083976273466e-05,
+      "loss": 0.6308,
+      "step": 18279
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2652396682981187e-05,
+      "loss": 0.5983,
+      "step": 18280
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2651709376211307e-05,
+      "loss": 0.6677,
+      "step": 18281
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2651022055967312e-05,
+      "loss": 0.5729,
+      "step": 18282
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26503347222527e-05,
+      "loss": 0.607,
+      "step": 18283
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2649647375070961e-05,
+      "loss": 0.5456,
+      "step": 18284
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2648960014425587e-05,
+      "loss": 0.6071,
+      "step": 18285
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.264827264032007e-05,
+      "loss": 0.6407,
+      "step": 18286
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2647585252757906e-05,
+      "loss": 0.6021,
+      "step": 18287
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2646897851742584e-05,
+      "loss": 0.5942,
+      "step": 18288
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26462104372776e-05,
+      "loss": 0.6552,
+      "step": 18289
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2645523009366447e-05,
+      "loss": 0.5578,
+      "step": 18290
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2644835568012617e-05,
+      "loss": 0.6013,
+      "step": 18291
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.26441481132196e-05,
+      "loss": 0.6292,
+      "step": 18292
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2643460644990897e-05,
+      "loss": 0.5882,
+      "step": 18293
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.2642773163329992e-05,
+      "loss": 0.5975,
+      "step": 18294
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2642085668240385e-05,
+      "loss": 0.5836,
+      "step": 18295
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2641398159725564e-05,
+      "loss": 0.5838,
+      "step": 18296
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2640710637789026e-05,
+      "loss": 0.5943,
+      "step": 18297
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2640023102434266e-05,
+      "loss": 0.6036,
+      "step": 18298
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2639335553664773e-05,
+      "loss": 0.5167,
+      "step": 18299
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2638647991484047e-05,
+      "loss": 0.6111,
+      "step": 18300
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2637960415895573e-05,
+      "loss": 0.562,
+      "step": 18301
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.263727282690285e-05,
+      "loss": 0.6514,
+      "step": 18302
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2636585224509377e-05,
+      "loss": 0.6474,
+      "step": 18303
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2635897608718637e-05,
+      "loss": 0.5383,
+      "step": 18304
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.263520997953413e-05,
+      "loss": 0.5699,
+      "step": 18305
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2634522336959352e-05,
+      "loss": 0.5912,
+      "step": 18306
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2633834680997792e-05,
+      "loss": 0.5252,
+      "step": 18307
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2633147011652947e-05,
+      "loss": 0.6077,
+      "step": 18308
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.263245932892831e-05,
+      "loss": 0.6047,
+      "step": 18309
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2631771632827377e-05,
+      "loss": 0.5835,
+      "step": 18310
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2631083923353642e-05,
+      "loss": 0.5703,
+      "step": 18311
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2630396200510602e-05,
+      "loss": 0.5516,
+      "step": 18312
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2629708464301744e-05,
+      "loss": 0.5507,
+      "step": 18313
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2629020714730569e-05,
+      "loss": 0.6165,
+      "step": 18314
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.262833295180057e-05,
+      "loss": 0.56,
+      "step": 18315
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.262764517551524e-05,
+      "loss": 0.6055,
+      "step": 18316
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2626957385878075e-05,
+      "loss": 0.6268,
+      "step": 18317
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2626269582892572e-05,
+      "loss": 0.6115,
+      "step": 18318
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2625581766562225e-05,
+      "loss": 0.5743,
+      "step": 18319
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2624893936890525e-05,
+      "loss": 0.6766,
+      "step": 18320
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2624206093880971e-05,
+      "loss": 0.627,
+      "step": 18321
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.262351823753706e-05,
+      "loss": 0.6056,
+      "step": 18322
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2622830367862281e-05,
+      "loss": 0.5829,
+      "step": 18323
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2622142484860133e-05,
+      "loss": 0.5754,
+      "step": 18324
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2621454588534115e-05,
+      "loss": 0.6261,
+      "step": 18325
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2620766678887715e-05,
+      "loss": 0.6379,
+      "step": 18326
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2620078755924435e-05,
+      "loss": 0.5983,
+      "step": 18327
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2619390819647763e-05,
+      "loss": 0.5988,
+      "step": 18328
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2618702870061202e-05,
+      "loss": 0.5789,
+      "step": 18329
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2618014907168246e-05,
+      "loss": 0.7188,
+      "step": 18330
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2617326930972389e-05,
+      "loss": 0.636,
+      "step": 18331
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2616638941477124e-05,
+      "loss": 0.6235,
+      "step": 18332
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2615950938685955e-05,
+      "loss": 0.5399,
+      "step": 18333
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.261526292260237e-05,
+      "loss": 0.6012,
+      "step": 18334
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2614574893229867e-05,
+      "loss": 0.6101,
+      "step": 18335
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2613886850571949e-05,
+      "loss": 0.6086,
+      "step": 18336
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.26131987946321e-05,
+      "loss": 0.5284,
+      "step": 18337
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2612510725413826e-05,
+      "loss": 0.5926,
+      "step": 18338
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.261182264292062e-05,
+      "loss": 0.5983,
+      "step": 18339
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2611134547155974e-05,
+      "loss": 0.6357,
+      "step": 18340
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2610446438123394e-05,
+      "loss": 0.5716,
+      "step": 18341
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2609758315826368e-05,
+      "loss": 0.5916,
+      "step": 18342
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2609070180268397e-05,
+      "loss": 0.5812,
+      "step": 18343
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2608382031452976e-05,
+      "loss": 0.5927,
+      "step": 18344
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2607693869383601e-05,
+      "loss": 0.5613,
+      "step": 18345
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2607005694063773e-05,
+      "loss": 0.5796,
+      "step": 18346
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260631750549698e-05,
+      "loss": 0.5727,
+      "step": 18347
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260562930368673e-05,
+      "loss": 0.6016,
+      "step": 18348
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2604941088636513e-05,
+      "loss": 0.6321,
+      "step": 18349
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2604252860349826e-05,
+      "loss": 0.6062,
+      "step": 18350
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2603564618830168e-05,
+      "loss": 0.6164,
+      "step": 18351
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2602876364081038e-05,
+      "loss": 0.6102,
+      "step": 18352
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260218809610593e-05,
+      "loss": 0.5808,
+      "step": 18353
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260149981490834e-05,
+      "loss": 0.5551,
+      "step": 18354
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260081152049177e-05,
+      "loss": 0.6119,
+      "step": 18355
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.260012321285971e-05,
+      "loss": 0.5746,
+      "step": 18356
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2599434892015667e-05,
+      "loss": 0.5535,
+      "step": 18357
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2598746557963133e-05,
+      "loss": 0.606,
+      "step": 18358
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2598058210705606e-05,
+      "loss": 0.5874,
+      "step": 18359
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2597369850246586e-05,
+      "loss": 0.5808,
+      "step": 18360
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2596681476589568e-05,
+      "loss": 0.6366,
+      "step": 18361
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.259599308973805e-05,
+      "loss": 0.6268,
+      "step": 18362
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2595304689695533e-05,
+      "loss": 0.6223,
+      "step": 18363
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2594616276465513e-05,
+      "loss": 0.658,
+      "step": 18364
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2593927850051487e-05,
+      "loss": 0.6005,
+      "step": 18365
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2593239410456952e-05,
+      "loss": 0.652,
+      "step": 18366
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2592550957685408e-05,
+      "loss": 0.6049,
+      "step": 18367
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2591862491740356e-05,
+      "loss": 0.6139,
+      "step": 18368
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.259117401262529e-05,
+      "loss": 0.636,
+      "step": 18369
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2590485520343712e-05,
+      "loss": 0.6564,
+      "step": 18370
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.258979701489912e-05,
+      "loss": 0.6288,
+      "step": 18371
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2589108496295006e-05,
+      "loss": 0.5891,
+      "step": 18372
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2588419964534877e-05,
+      "loss": 0.6304,
+      "step": 18373
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2587731419622224e-05,
+      "loss": 0.5724,
+      "step": 18374
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2587042861560551e-05,
+      "loss": 0.6443,
+      "step": 18375
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.258635429035336e-05,
+      "loss": 0.6499,
+      "step": 18376
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.258566570600414e-05,
+      "loss": 0.6389,
+      "step": 18377
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2584977108516396e-05,
+      "loss": 0.6,
+      "step": 18378
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2584288497893632e-05,
+      "loss": 0.5836,
+      "step": 18379
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2583599874139335e-05,
+      "loss": 0.5833,
+      "step": 18380
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2582911237257015e-05,
+      "loss": 0.5694,
+      "step": 18381
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2582222587250161e-05,
+      "loss": 0.6229,
+      "step": 18382
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.258153392412228e-05,
+      "loss": 0.5613,
+      "step": 18383
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2580845247876872e-05,
+      "loss": 0.5749,
+      "step": 18384
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2580156558517432e-05,
+      "loss": 0.5891,
+      "step": 18385
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2579467856047458e-05,
+      "loss": 0.6241,
+      "step": 18386
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2578779140470455e-05,
+      "loss": 0.6108,
+      "step": 18387
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2578090411789917e-05,
+      "loss": 0.6353,
+      "step": 18388
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2577401670009353e-05,
+      "loss": 0.5478,
+      "step": 18389
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2576712915132252e-05,
+      "loss": 0.5936,
+      "step": 18390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.257602414716212e-05,
+      "loss": 0.6257,
+      "step": 18391
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2575335366102453e-05,
+      "loss": 0.5974,
+      "step": 18392
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2574646571956754e-05,
+      "loss": 0.6008,
+      "step": 18393
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2573957764728519e-05,
+      "loss": 0.6068,
+      "step": 18394
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2573268944421253e-05,
+      "loss": 0.6291,
+      "step": 18395
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2572580111038453e-05,
+      "loss": 0.6136,
+      "step": 18396
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.257189126458362e-05,
+      "loss": 0.5987,
+      "step": 18397
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2571202405060257e-05,
+      "loss": 0.5836,
+      "step": 18398
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.257051353247186e-05,
+      "loss": 0.747,
+      "step": 18399
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2569824646821931e-05,
+      "loss": 0.607,
+      "step": 18400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2569135748113967e-05,
+      "loss": 0.5554,
+      "step": 18401
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2568446836351476e-05,
+      "loss": 0.6175,
+      "step": 18402
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2567757911537953e-05,
+      "loss": 0.6214,
+      "step": 18403
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.25670689736769e-05,
+      "loss": 0.5935,
+      "step": 18404
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2566380022771818e-05,
+      "loss": 0.6348,
+      "step": 18405
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2565691058826209e-05,
+      "loss": 0.6318,
+      "step": 18406
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2565002081843569e-05,
+      "loss": 0.6137,
+      "step": 18407
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2564313091827404e-05,
+      "loss": 0.6328,
+      "step": 18408
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2563624088781213e-05,
+      "loss": 0.6641,
+      "step": 18409
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2562935072708497e-05,
+      "loss": 0.5805,
+      "step": 18410
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2562246043612758e-05,
+      "loss": 0.6048,
+      "step": 18411
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2561557001497498e-05,
+      "loss": 0.6596,
+      "step": 18412
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2560867946366214e-05,
+      "loss": 0.6233,
+      "step": 18413
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2560178878222412e-05,
+      "loss": 0.6269,
+      "step": 18414
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2559489797069588e-05,
+      "loss": 0.5937,
+      "step": 18415
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.255880070291125e-05,
+      "loss": 0.6483,
+      "step": 18416
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2558111595750895e-05,
+      "loss": 0.6385,
+      "step": 18417
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2557422475592025e-05,
+      "loss": 0.6046,
+      "step": 18418
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2556733342438146e-05,
+      "loss": 0.6269,
+      "step": 18419
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2556044196292752e-05,
+      "loss": 0.6192,
+      "step": 18420
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.255535503715935e-05,
+      "loss": 0.5941,
+      "step": 18421
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2554665865041442e-05,
+      "loss": 0.6134,
+      "step": 18422
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2553976679942525e-05,
+      "loss": 0.5684,
+      "step": 18423
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2553287481866109e-05,
+      "loss": 0.5779,
+      "step": 18424
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2552598270815688e-05,
+      "loss": 0.6523,
+      "step": 18425
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.255190904679477e-05,
+      "loss": 0.5947,
+      "step": 18426
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2551219809806857e-05,
+      "loss": 0.6344,
+      "step": 18427
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2550530559855442e-05,
+      "loss": 0.5866,
+      "step": 18428
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.254984129694404e-05,
+      "loss": 0.5733,
+      "step": 18429
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2549152021076145e-05,
+      "loss": 0.5977,
+      "step": 18430
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2548462732255265e-05,
+      "loss": 0.5729,
+      "step": 18431
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2547773430484895e-05,
+      "loss": 0.582,
+      "step": 18432
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2547084115768547e-05,
+      "loss": 0.6298,
+      "step": 18433
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2546394788109714e-05,
+      "loss": 0.5573,
+      "step": 18434
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2545705447511907e-05,
+      "loss": 0.5655,
+      "step": 18435
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2545016093978624e-05,
+      "loss": 0.6637,
+      "step": 18436
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2544326727513367e-05,
+      "loss": 0.5991,
+      "step": 18437
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2543637348119644e-05,
+      "loss": 0.5844,
+      "step": 18438
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2542947955800953e-05,
+      "loss": 0.5944,
+      "step": 18439
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2542258550560798e-05,
+      "loss": 0.5491,
+      "step": 18440
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.2541569132402685e-05,
+      "loss": 0.6255,
+      "step": 18441
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.254087970133011e-05,
+      "loss": 0.6371,
+      "step": 18442
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2540190257346585e-05,
+      "loss": 0.5665,
+      "step": 18443
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.253950080045561e-05,
+      "loss": 0.5994,
+      "step": 18444
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2538811330660684e-05,
+      "loss": 0.542,
+      "step": 18445
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2538121847965319e-05,
+      "loss": 0.5773,
+      "step": 18446
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.253743235237301e-05,
+      "loss": 0.5798,
+      "step": 18447
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2536742843887266e-05,
+      "loss": 0.5893,
+      "step": 18448
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2536053322511588e-05,
+      "loss": 0.6108,
+      "step": 18449
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2535363788249481e-05,
+      "loss": 0.6549,
+      "step": 18450
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2534674241104448e-05,
+      "loss": 0.6459,
+      "step": 18451
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.253398468107999e-05,
+      "loss": 0.5668,
+      "step": 18452
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2533295108179618e-05,
+      "loss": 0.5861,
+      "step": 18453
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2532605522406826e-05,
+      "loss": 0.6609,
+      "step": 18454
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.253191592376513e-05,
+      "loss": 0.5863,
+      "step": 18455
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2531226312258024e-05,
+      "loss": 0.59,
+      "step": 18456
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2530536687889019e-05,
+      "loss": 0.5766,
+      "step": 18457
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2529847050661612e-05,
+      "loss": 0.6268,
+      "step": 18458
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2529157400579312e-05,
+      "loss": 0.5727,
+      "step": 18459
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2528467737645623e-05,
+      "loss": 0.6332,
+      "step": 18460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2527778061864051e-05,
+      "loss": 0.5945,
+      "step": 18461
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2527088373238097e-05,
+      "loss": 0.5963,
+      "step": 18462
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2526398671771267e-05,
+      "loss": 0.577,
+      "step": 18463
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2525708957467065e-05,
+      "loss": 0.5777,
+      "step": 18464
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2525019230328997e-05,
+      "loss": 0.6234,
+      "step": 18465
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2524329490360566e-05,
+      "loss": 0.5513,
+      "step": 18466
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2523639737565277e-05,
+      "loss": 0.6,
+      "step": 18467
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2522949971946635e-05,
+      "loss": 0.681,
+      "step": 18468
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2522260193508147e-05,
+      "loss": 0.615,
+      "step": 18469
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2521570402253316e-05,
+      "loss": 0.6075,
+      "step": 18470
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2520880598185648e-05,
+      "loss": 0.6371,
+      "step": 18471
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2520190781308646e-05,
+      "loss": 0.5766,
+      "step": 18472
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2519500951625818e-05,
+      "loss": 0.6758,
+      "step": 18473
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2518811109140666e-05,
+      "loss": 0.5932,
+      "step": 18474
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2518121253856696e-05,
+      "loss": 0.5741,
+      "step": 18475
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2517431385777417e-05,
+      "loss": 0.5884,
+      "step": 18476
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2516741504906332e-05,
+      "loss": 0.6045,
+      "step": 18477
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2516051611246942e-05,
+      "loss": 0.5888,
+      "step": 18478
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2515361704802765e-05,
+      "loss": 0.619,
+      "step": 18479
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.251467178557729e-05,
+      "loss": 0.66,
+      "step": 18480
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2513981853574036e-05,
+      "loss": 0.6393,
+      "step": 18481
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2513291908796504e-05,
+      "loss": 0.6133,
+      "step": 18482
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2512601951248197e-05,
+      "loss": 0.5763,
+      "step": 18483
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2511911980932627e-05,
+      "loss": 0.5988,
+      "step": 18484
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2511221997853293e-05,
+      "loss": 0.6261,
+      "step": 18485
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2510532002013706e-05,
+      "loss": 0.6533,
+      "step": 18486
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2509841993417373e-05,
+      "loss": 0.5341,
+      "step": 18487
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2509151972067797e-05,
+      "loss": 0.6344,
+      "step": 18488
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2508461937968484e-05,
+      "loss": 0.6128,
+      "step": 18489
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2507771891122941e-05,
+      "loss": 0.6357,
+      "step": 18490
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2507081831534676e-05,
+      "loss": 0.5686,
+      "step": 18491
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2506391759207192e-05,
+      "loss": 0.5742,
+      "step": 18492
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2505701674144e-05,
+      "loss": 0.6286,
+      "step": 18493
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.25050115763486e-05,
+      "loss": 0.5879,
+      "step": 18494
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2504321465824508e-05,
+      "loss": 0.5836,
+      "step": 18495
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.250363134257522e-05,
+      "loss": 0.6097,
+      "step": 18496
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2502941206604249e-05,
+      "loss": 0.6246,
+      "step": 18497
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2502251057915102e-05,
+      "loss": 0.5993,
+      "step": 18498
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2501560896511284e-05,
+      "loss": 0.5914,
+      "step": 18499
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2500870722396302e-05,
+      "loss": 0.593,
+      "step": 18500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2500180535573665e-05,
+      "loss": 0.6326,
+      "step": 18501
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2499490336046875e-05,
+      "loss": 0.5589,
+      "step": 18502
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2498800123819449e-05,
+      "loss": 0.647,
+      "step": 18503
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2498109898894882e-05,
+      "loss": 0.5922,
+      "step": 18504
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.249741966127669e-05,
+      "loss": 0.6294,
+      "step": 18505
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2496729410968376e-05,
+      "loss": 0.6152,
+      "step": 18506
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2496039147973446e-05,
+      "loss": 0.6196,
+      "step": 18507
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2495348872295417e-05,
+      "loss": 0.554,
+      "step": 18508
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2494658583937786e-05,
+      "loss": 0.5538,
+      "step": 18509
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2493968282904061e-05,
+      "loss": 0.6625,
+      "step": 18510
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2493277969197756e-05,
+      "loss": 0.6159,
+      "step": 18511
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2492587642822376e-05,
+      "loss": 0.5619,
+      "step": 18512
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2491897303781426e-05,
+      "loss": 0.6447,
+      "step": 18513
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2491206952078419e-05,
+      "loss": 0.6061,
+      "step": 18514
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2490516587716857e-05,
+      "loss": 0.6112,
+      "step": 18515
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2489826210700252e-05,
+      "loss": 0.6118,
+      "step": 18516
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.248913582103211e-05,
+      "loss": 0.6143,
+      "step": 18517
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.248844541871594e-05,
+      "loss": 0.5271,
+      "step": 18518
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2487755003755252e-05,
+      "loss": 0.5685,
+      "step": 18519
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.248706457615355e-05,
+      "loss": 0.5482,
+      "step": 18520
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2486374135914346e-05,
+      "loss": 0.6217,
+      "step": 18521
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2485683683041146e-05,
+      "loss": 0.6143,
+      "step": 18522
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.248499321753746e-05,
+      "loss": 0.6085,
+      "step": 18523
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2484302739406793e-05,
+      "loss": 0.5605,
+      "step": 18524
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.248361224865266e-05,
+      "loss": 0.644,
+      "step": 18525
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2482921745278564e-05,
+      "loss": 0.6017,
+      "step": 18526
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2482231229288018e-05,
+      "loss": 0.568,
+      "step": 18527
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2481540700684524e-05,
+      "loss": 0.6017,
+      "step": 18528
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2480850159471599e-05,
+      "loss": 0.5704,
+      "step": 18529
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2480159605652746e-05,
+      "loss": 0.656,
+      "step": 18530
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2479469039231473e-05,
+      "loss": 0.5847,
+      "step": 18531
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2478778460211294e-05,
+      "loss": 0.5542,
+      "step": 18532
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2478087868595717e-05,
+      "loss": 0.6455,
+      "step": 18533
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2477397264388246e-05,
+      "loss": 0.604,
+      "step": 18534
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2476706647592397e-05,
+      "loss": 0.5962,
+      "step": 18535
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2476016018211677e-05,
+      "loss": 0.5592,
+      "step": 18536
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2475325376249591e-05,
+      "loss": 0.6704,
+      "step": 18537
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2474634721709658e-05,
+      "loss": 0.6389,
+      "step": 18538
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2473944054595374e-05,
+      "loss": 0.5975,
+      "step": 18539
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.247325337491026e-05,
+      "loss": 0.6264,
+      "step": 18540
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.247256268265782e-05,
+      "loss": 0.596,
+      "step": 18541
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2471871977841566e-05,
+      "loss": 0.6149,
+      "step": 18542
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2471181260465004e-05,
+      "loss": 0.599,
+      "step": 18543
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2470490530531649e-05,
+      "loss": 0.565,
+      "step": 18544
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2469799788045005e-05,
+      "loss": 0.6733,
+      "step": 18545
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2469109033008588e-05,
+      "loss": 0.5579,
+      "step": 18546
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2468418265425904e-05,
+      "loss": 0.6408,
+      "step": 18547
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2467727485300464e-05,
+      "loss": 0.6708,
+      "step": 18548
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.246703669263578e-05,
+      "loss": 0.6225,
+      "step": 18549
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2466345887435356e-05,
+      "loss": 0.5613,
+      "step": 18550
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2465655069702707e-05,
+      "loss": 0.5426,
+      "step": 18551
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2464964239441345e-05,
+      "loss": 0.6223,
+      "step": 18552
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2464273396654775e-05,
+      "loss": 0.585,
+      "step": 18553
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.246358254134651e-05,
+      "loss": 0.6357,
+      "step": 18554
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2462891673520065e-05,
+      "loss": 0.5774,
+      "step": 18555
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2462200793178942e-05,
+      "loss": 0.6364,
+      "step": 18556
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2461509900326658e-05,
+      "loss": 0.5906,
+      "step": 18557
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.246081899496672e-05,
+      "loss": 0.6461,
+      "step": 18558
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.246012807710264e-05,
+      "loss": 0.5612,
+      "step": 18559
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2459437146737929e-05,
+      "loss": 0.6474,
+      "step": 18560
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2458746203876098e-05,
+      "loss": 0.6183,
+      "step": 18561
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2458055248520655e-05,
+      "loss": 0.6281,
+      "step": 18562
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2457364280675118e-05,
+      "loss": 0.643,
+      "step": 18563
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2456673300342987e-05,
+      "loss": 0.7398,
+      "step": 18564
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2455982307527785e-05,
+      "loss": 0.6062,
+      "step": 18565
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2455291302233017e-05,
+      "loss": 0.6089,
+      "step": 18566
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2454600284462194e-05,
+      "loss": 0.5764,
+      "step": 18567
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.245390925421883e-05,
+      "loss": 0.5701,
+      "step": 18568
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.245321821150643e-05,
+      "loss": 0.6567,
+      "step": 18569
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2452527156328511e-05,
+      "loss": 0.594,
+      "step": 18570
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2451836088688587e-05,
+      "loss": 0.6136,
+      "step": 18571
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2451145008590162e-05,
+      "loss": 0.6097,
+      "step": 18572
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.245045391603675e-05,
+      "loss": 0.5547,
+      "step": 18573
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2449762811031868e-05,
+      "loss": 0.6062,
+      "step": 18574
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2449071693579022e-05,
+      "loss": 0.6065,
+      "step": 18575
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2448380563681729e-05,
+      "loss": 0.5635,
+      "step": 18576
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2447689421343491e-05,
+      "loss": 0.6106,
+      "step": 18577
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2446998266567829e-05,
+      "loss": 0.6241,
+      "step": 18578
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2446307099358253e-05,
+      "loss": 0.609,
+      "step": 18579
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2445615919718272e-05,
+      "loss": 0.5297,
+      "step": 18580
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2444924727651401e-05,
+      "loss": 0.6247,
+      "step": 18581
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2444233523161156e-05,
+      "loss": 0.6188,
+      "step": 18582
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2443542306251039e-05,
+      "loss": 0.6377,
+      "step": 18583
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2442851076924571e-05,
+      "loss": 0.653,
+      "step": 18584
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2442159835185259e-05,
+      "loss": 0.6095,
+      "step": 18585
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.244146858103662e-05,
+      "loss": 0.5762,
+      "step": 18586
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.2440777314482163e-05,
+      "loss": 0.632,
+      "step": 18587
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.24400860355254e-05,
+      "loss": 0.6404,
+      "step": 18588
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2439394744169847e-05,
+      "loss": 0.5933,
+      "step": 18589
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2438703440419016e-05,
+      "loss": 0.5816,
+      "step": 18590
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2438012124276418e-05,
+      "loss": 0.6208,
+      "step": 18591
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2437320795745565e-05,
+      "loss": 0.6207,
+      "step": 18592
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2436629454829972e-05,
+      "loss": 0.5676,
+      "step": 18593
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.243593810153315e-05,
+      "loss": 0.6053,
+      "step": 18594
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2435246735858614e-05,
+      "loss": 0.6186,
+      "step": 18595
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2434555357809876e-05,
+      "loss": 0.5602,
+      "step": 18596
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2433863967390449e-05,
+      "loss": 0.6123,
+      "step": 18597
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2433172564603848e-05,
+      "loss": 0.5779,
+      "step": 18598
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2432481149453583e-05,
+      "loss": 0.6031,
+      "step": 18599
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2431789721943169e-05,
+      "loss": 0.6085,
+      "step": 18600
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2431098282076122e-05,
+      "loss": 0.5599,
+      "step": 18601
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2430406829855948e-05,
+      "loss": 0.5904,
+      "step": 18602
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2429715365286169e-05,
+      "loss": 0.6326,
+      "step": 18603
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2429023888370293e-05,
+      "loss": 0.5729,
+      "step": 18604
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2428332399111835e-05,
+      "loss": 0.5807,
+      "step": 18605
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2427640897514309e-05,
+      "loss": 0.559,
+      "step": 18606
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.242694938358123e-05,
+      "loss": 0.5703,
+      "step": 18607
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2426257857316109e-05,
+      "loss": 0.6341,
+      "step": 18608
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2425566318722462e-05,
+      "loss": 0.6366,
+      "step": 18609
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.24248747678038e-05,
+      "loss": 0.6141,
+      "step": 18610
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2424183204563645e-05,
+      "loss": 0.5626,
+      "step": 18611
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.24234916290055e-05,
+      "loss": 0.6274,
+      "step": 18612
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2422800041132883e-05,
+      "loss": 0.5723,
+      "step": 18613
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2422108440949314e-05,
+      "loss": 0.5968,
+      "step": 18614
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2421416828458299e-05,
+      "loss": 0.6158,
+      "step": 18615
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2420725203663356e-05,
+      "loss": 0.5643,
+      "step": 18616
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2420033566568002e-05,
+      "loss": 0.6807,
+      "step": 18617
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2419341917175746e-05,
+      "loss": 0.6051,
+      "step": 18618
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2418650255490106e-05,
+      "loss": 0.5768,
+      "step": 18619
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2417958581514597e-05,
+      "loss": 0.6144,
+      "step": 18620
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2417266895252731e-05,
+      "loss": 0.6448,
+      "step": 18621
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2416575196708027e-05,
+      "loss": 0.6596,
+      "step": 18622
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2415883485883993e-05,
+      "loss": 0.6181,
+      "step": 18623
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.241519176278415e-05,
+      "loss": 0.6103,
+      "step": 18624
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2414500027412008e-05,
+      "loss": 0.6669,
+      "step": 18625
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2413808279771086e-05,
+      "loss": 0.6307,
+      "step": 18626
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2413116519864896e-05,
+      "loss": 0.6306,
+      "step": 18627
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2412424747696956e-05,
+      "loss": 0.5485,
+      "step": 18628
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2411732963270776e-05,
+      "loss": 0.6067,
+      "step": 18629
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2411041166589879e-05,
+      "loss": 0.6196,
+      "step": 18630
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2410349357657774e-05,
+      "loss": 0.5901,
+      "step": 18631
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2409657536477975e-05,
+      "loss": 0.6196,
+      "step": 18632
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2408965703054006e-05,
+      "loss": 0.6136,
+      "step": 18633
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2408273857389372e-05,
+      "loss": 0.5792,
+      "step": 18634
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2407581999487593e-05,
+      "loss": 0.6542,
+      "step": 18635
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2406890129352188e-05,
+      "loss": 0.6069,
+      "step": 18636
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2406198246986667e-05,
+      "loss": 0.645,
+      "step": 18637
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2405506352394547e-05,
+      "loss": 0.6248,
+      "step": 18638
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.240481444557935e-05,
+      "loss": 0.5975,
+      "step": 18639
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2404122526544581e-05,
+      "loss": 0.636,
+      "step": 18640
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2403430595293764e-05,
+      "loss": 0.6103,
+      "step": 18641
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2402738651830413e-05,
+      "loss": 0.5668,
+      "step": 18642
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2402046696158041e-05,
+      "loss": 0.5201,
+      "step": 18643
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2401354728280168e-05,
+      "loss": 0.5911,
+      "step": 18644
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.240066274820031e-05,
+      "loss": 0.5818,
+      "step": 18645
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2399970755921976e-05,
+      "loss": 0.6124,
+      "step": 18646
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2399278751448694e-05,
+      "loss": 0.6459,
+      "step": 18647
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2398586734783969e-05,
+      "loss": 0.6068,
+      "step": 18648
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2397894705931326e-05,
+      "loss": 0.6071,
+      "step": 18649
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2397202664894275e-05,
+      "loss": 0.6101,
+      "step": 18650
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2396510611676335e-05,
+      "loss": 0.6009,
+      "step": 18651
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2395818546281024e-05,
+      "loss": 0.5902,
+      "step": 18652
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2395126468711858e-05,
+      "loss": 0.6486,
+      "step": 18653
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2394434378972352e-05,
+      "loss": 0.6222,
+      "step": 18654
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2393742277066023e-05,
+      "loss": 0.6252,
+      "step": 18655
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.239305016299639e-05,
+      "loss": 0.64,
+      "step": 18656
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2392358036766968e-05,
+      "loss": 0.5795,
+      "step": 18657
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2391665898381273e-05,
+      "loss": 0.5625,
+      "step": 18658
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2390973747842823e-05,
+      "loss": 0.5996,
+      "step": 18659
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2390281585155138e-05,
+      "loss": 0.5874,
+      "step": 18660
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2389589410321729e-05,
+      "loss": 0.5687,
+      "step": 18661
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2388897223346116e-05,
+      "loss": 0.606,
+      "step": 18662
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.238820502423182e-05,
+      "loss": 0.547,
+      "step": 18663
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2387512812982354e-05,
+      "loss": 0.5837,
+      "step": 18664
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2386820589601233e-05,
+      "loss": 0.6128,
+      "step": 18665
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2386128354091984e-05,
+      "loss": 0.6634,
+      "step": 18666
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2385436106458113e-05,
+      "loss": 0.5568,
+      "step": 18667
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2384743846703144e-05,
+      "loss": 0.6344,
+      "step": 18668
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2384051574830592e-05,
+      "loss": 0.6145,
+      "step": 18669
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2383359290843975e-05,
+      "loss": 0.6239,
+      "step": 18670
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2382666994746814e-05,
+      "loss": 0.6442,
+      "step": 18671
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2381974686542621e-05,
+      "loss": 0.6278,
+      "step": 18672
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2381282366234919e-05,
+      "loss": 0.5898,
+      "step": 18673
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2380590033827224e-05,
+      "loss": 0.5708,
+      "step": 18674
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2379897689323054e-05,
+      "loss": 0.5944,
+      "step": 18675
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2379205332725928e-05,
+      "loss": 0.5961,
+      "step": 18676
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2378512964039363e-05,
+      "loss": 0.5384,
+      "step": 18677
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2377820583266872e-05,
+      "loss": 0.5614,
+      "step": 18678
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2377128190411984e-05,
+      "loss": 0.5971,
+      "step": 18679
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2376435785478207e-05,
+      "loss": 0.5947,
+      "step": 18680
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2375743368469066e-05,
+      "loss": 0.598,
+      "step": 18681
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2375050939388078e-05,
+      "loss": 0.5826,
+      "step": 18682
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2374358498238759e-05,
+      "loss": 0.6474,
+      "step": 18683
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2373666045024633e-05,
+      "loss": 0.671,
+      "step": 18684
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2372973579749212e-05,
+      "loss": 0.6063,
+      "step": 18685
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2372281102416016e-05,
+      "loss": 0.6133,
+      "step": 18686
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2371588613028567e-05,
+      "loss": 0.6424,
+      "step": 18687
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.237089611159038e-05,
+      "loss": 0.5954,
+      "step": 18688
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2370203598104976e-05,
+      "loss": 0.5943,
+      "step": 18689
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2369511072575873e-05,
+      "loss": 0.5964,
+      "step": 18690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.236881853500659e-05,
+      "loss": 0.6075,
+      "step": 18691
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2368125985400647e-05,
+      "loss": 0.5974,
+      "step": 18692
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2367433423761564e-05,
+      "loss": 0.5773,
+      "step": 18693
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2366740850092855e-05,
+      "loss": 0.6216,
+      "step": 18694
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2366048264398046e-05,
+      "loss": 0.5862,
+      "step": 18695
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.236535566668065e-05,
+      "loss": 0.6396,
+      "step": 18696
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2364663056944191e-05,
+      "loss": 0.6144,
+      "step": 18697
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2363970435192188e-05,
+      "loss": 0.5447,
+      "step": 18698
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2363277801428156e-05,
+      "loss": 0.6122,
+      "step": 18699
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2362585155655616e-05,
+      "loss": 0.6052,
+      "step": 18700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2361892497878091e-05,
+      "loss": 0.5856,
+      "step": 18701
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.23611998280991e-05,
+      "loss": 0.5902,
+      "step": 18702
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.236050714632216e-05,
+      "loss": 0.6101,
+      "step": 18703
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2359814452550794e-05,
+      "loss": 0.6586,
+      "step": 18704
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2359121746788513e-05,
+      "loss": 0.6455,
+      "step": 18705
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.235842902903885e-05,
+      "loss": 0.5969,
+      "step": 18706
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2357736299305316e-05,
+      "loss": 0.6088,
+      "step": 18707
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2357043557591433e-05,
+      "loss": 0.608,
+      "step": 18708
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2356350803900722e-05,
+      "loss": 0.6376,
+      "step": 18709
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2355658038236701e-05,
+      "loss": 0.5718,
+      "step": 18710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2354965260602894e-05,
+      "loss": 0.5551,
+      "step": 18711
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2354272471002819e-05,
+      "loss": 0.6139,
+      "step": 18712
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2353579669439995e-05,
+      "loss": 0.5761,
+      "step": 18713
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2352886855917943e-05,
+      "loss": 0.5972,
+      "step": 18714
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2352194030440183e-05,
+      "loss": 0.5729,
+      "step": 18715
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2351501193010238e-05,
+      "loss": 0.6244,
+      "step": 18716
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2350808343631626e-05,
+      "loss": 0.6057,
+      "step": 18717
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.235011548230787e-05,
+      "loss": 0.6367,
+      "step": 18718
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2349422609042486e-05,
+      "loss": 0.6241,
+      "step": 18719
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2348729723838998e-05,
+      "loss": 0.6106,
+      "step": 18720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2348036826700929e-05,
+      "loss": 0.5788,
+      "step": 18721
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2347343917631797e-05,
+      "loss": 0.5537,
+      "step": 18722
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2346650996635122e-05,
+      "loss": 0.6068,
+      "step": 18723
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2345958063714427e-05,
+      "loss": 0.6054,
+      "step": 18724
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.234526511887323e-05,
+      "loss": 0.6093,
+      "step": 18725
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2344572162115056e-05,
+      "loss": 0.6502,
+      "step": 18726
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2343879193443422e-05,
+      "loss": 0.5426,
+      "step": 18727
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2343186212861855e-05,
+      "loss": 0.5832,
+      "step": 18728
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2342493220373868e-05,
+      "loss": 0.6351,
+      "step": 18729
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2341800215982989e-05,
+      "loss": 0.5808,
+      "step": 18730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2341107199692739e-05,
+      "loss": 0.5902,
+      "step": 18731
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2340414171506636e-05,
+      "loss": 0.593,
+      "step": 18732
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2339721131428204e-05,
+      "loss": 0.5728,
+      "step": 18733
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2339028079460962e-05,
+      "loss": 0.6038,
+      "step": 18734
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.2338335015608433e-05,
+      "loss": 0.5672,
+      "step": 18735
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2337641939874144e-05,
+      "loss": 0.6444,
+      "step": 18736
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2336948852261605e-05,
+      "loss": 0.568,
+      "step": 18737
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2336255752774346e-05,
+      "loss": 0.5749,
+      "step": 18738
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2335562641415893e-05,
+      "loss": 0.5389,
+      "step": 18739
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2334869518189756e-05,
+      "loss": 0.5917,
+      "step": 18740
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2334176383099465e-05,
+      "loss": 0.641,
+      "step": 18741
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.233348323614854e-05,
+      "loss": 0.6307,
+      "step": 18742
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2332790077340506e-05,
+      "loss": 0.6196,
+      "step": 18743
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2332096906678879e-05,
+      "loss": 0.5992,
+      "step": 18744
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2331403724167188e-05,
+      "loss": 0.579,
+      "step": 18745
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2330710529808947e-05,
+      "loss": 0.6234,
+      "step": 18746
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2330017323607688e-05,
+      "loss": 0.5693,
+      "step": 18747
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2329324105566927e-05,
+      "loss": 0.5621,
+      "step": 18748
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2328630875690189e-05,
+      "loss": 0.641,
+      "step": 18749
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.232793763398099e-05,
+      "loss": 0.6239,
+      "step": 18750
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2327244380442864e-05,
+      "loss": 0.592,
+      "step": 18751
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2326551115079327e-05,
+      "loss": 0.5838,
+      "step": 18752
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.23258578378939e-05,
+      "loss": 0.6368,
+      "step": 18753
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.232516454889011e-05,
+      "loss": 0.5937,
+      "step": 18754
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2324471248071477e-05,
+      "loss": 0.5408,
+      "step": 18755
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2323777935441526e-05,
+      "loss": 0.5746,
+      "step": 18756
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2323084611003778e-05,
+      "loss": 0.644,
+      "step": 18757
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2322391274761756e-05,
+      "loss": 0.5923,
+      "step": 18758
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2321697926718982e-05,
+      "loss": 0.5893,
+      "step": 18759
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2321004566878987e-05,
+      "loss": 0.6162,
+      "step": 18760
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2320311195245282e-05,
+      "loss": 0.5478,
+      "step": 18761
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2319617811821401e-05,
+      "loss": 0.6009,
+      "step": 18762
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.231892441661086e-05,
+      "loss": 0.6388,
+      "step": 18763
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2318231009617188e-05,
+      "loss": 0.6411,
+      "step": 18764
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.23175375908439e-05,
+      "loss": 0.6265,
+      "step": 18765
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.231684416029453e-05,
+      "loss": 0.5572,
+      "step": 18766
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2316150717972593e-05,
+      "loss": 0.637,
+      "step": 18767
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2315457263881619e-05,
+      "loss": 0.5727,
+      "step": 18768
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2314763798025125e-05,
+      "loss": 0.497,
+      "step": 18769
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2314070320406637e-05,
+      "loss": 0.5949,
+      "step": 18770
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2313376831029683e-05,
+      "loss": 0.6338,
+      "step": 18771
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2312683329897783e-05,
+      "loss": 0.6654,
+      "step": 18772
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.231198981701446e-05,
+      "loss": 0.5474,
+      "step": 18773
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2311296292383243e-05,
+      "loss": 0.6613,
+      "step": 18774
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2310602756007652e-05,
+      "loss": 0.6127,
+      "step": 18775
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2309909207891208e-05,
+      "loss": 0.5656,
+      "step": 18776
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2309215648037445e-05,
+      "loss": 0.5721,
+      "step": 18777
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2308522076449876e-05,
+      "loss": 0.5649,
+      "step": 18778
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2307828493132034e-05,
+      "loss": 0.6123,
+      "step": 18779
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2307134898087435e-05,
+      "loss": 0.568,
+      "step": 18780
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2306441291319613e-05,
+      "loss": 0.6384,
+      "step": 18781
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2305747672832085e-05,
+      "loss": 0.5958,
+      "step": 18782
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2305054042628377e-05,
+      "loss": 0.574,
+      "step": 18783
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2304360400712015e-05,
+      "loss": 0.5938,
+      "step": 18784
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2303666747086525e-05,
+      "loss": 0.5969,
+      "step": 18785
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2302973081755425e-05,
+      "loss": 0.6089,
+      "step": 18786
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2302279404722249e-05,
+      "loss": 0.6437,
+      "step": 18787
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2301585715990514e-05,
+      "loss": 0.5577,
+      "step": 18788
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2300892015563748e-05,
+      "loss": 0.6086,
+      "step": 18789
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2300198303445477e-05,
+      "loss": 0.642,
+      "step": 18790
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2299504579639224e-05,
+      "loss": 0.5627,
+      "step": 18791
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2298810844148514e-05,
+      "loss": 0.5744,
+      "step": 18792
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2298117096976877e-05,
+      "loss": 0.6063,
+      "step": 18793
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2297423338127827e-05,
+      "loss": 0.5969,
+      "step": 18794
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2296729567604904e-05,
+      "loss": 0.6481,
+      "step": 18795
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.229603578541162e-05,
+      "loss": 0.5396,
+      "step": 18796
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2295341991551506e-05,
+      "loss": 0.5748,
+      "step": 18797
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.229464818602809e-05,
+      "loss": 0.6069,
+      "step": 18798
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2293954368844892e-05,
+      "loss": 0.588,
+      "step": 18799
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2293260540005442e-05,
+      "loss": 0.5801,
+      "step": 18800
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2292566699513262e-05,
+      "loss": 0.635,
+      "step": 18801
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2291872847371882e-05,
+      "loss": 0.5699,
+      "step": 18802
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2291178983584821e-05,
+      "loss": 0.6329,
+      "step": 18803
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2290485108155612e-05,
+      "loss": 0.5346,
+      "step": 18804
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2289791221087775e-05,
+      "loss": 0.612,
+      "step": 18805
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.228909732238484e-05,
+      "loss": 0.6695,
+      "step": 18806
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.228840341205033e-05,
+      "loss": 0.6836,
+      "step": 18807
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2287709490087772e-05,
+      "loss": 0.5881,
+      "step": 18808
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2287015556500693e-05,
+      "loss": 0.5401,
+      "step": 18809
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2286321611292616e-05,
+      "loss": 0.5805,
+      "step": 18810
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.228562765446707e-05,
+      "loss": 0.6524,
+      "step": 18811
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2284933686027583e-05,
+      "loss": 0.6091,
+      "step": 18812
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2284239705977674e-05,
+      "loss": 0.6662,
+      "step": 18813
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2283545714320879e-05,
+      "loss": 0.6391,
+      "step": 18814
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2282851711060717e-05,
+      "loss": 0.6445,
+      "step": 18815
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2282157696200716e-05,
+      "loss": 0.6023,
+      "step": 18816
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2281463669744406e-05,
+      "loss": 0.589,
+      "step": 18817
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2280769631695308e-05,
+      "loss": 0.5772,
+      "step": 18818
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2280075582056955e-05,
+      "loss": 0.5801,
+      "step": 18819
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.227938152083287e-05,
+      "loss": 0.6163,
+      "step": 18820
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2278687448026579e-05,
+      "loss": 0.6293,
+      "step": 18821
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2277993363641608e-05,
+      "loss": 0.5896,
+      "step": 18822
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2277299267681486e-05,
+      "loss": 0.646,
+      "step": 18823
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2276605160149741e-05,
+      "loss": 0.6472,
+      "step": 18824
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2275911041049899e-05,
+      "loss": 0.57,
+      "step": 18825
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2275216910385483e-05,
+      "loss": 0.6034,
+      "step": 18826
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2274522768160027e-05,
+      "loss": 0.581,
+      "step": 18827
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2273828614377052e-05,
+      "loss": 0.5489,
+      "step": 18828
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.227313444904009e-05,
+      "loss": 0.5769,
+      "step": 18829
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2272440272152664e-05,
+      "loss": 0.6203,
+      "step": 18830
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2271746083718306e-05,
+      "loss": 0.6467,
+      "step": 18831
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2271051883740539e-05,
+      "loss": 0.6454,
+      "step": 18832
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2270357672222895e-05,
+      "loss": 0.5853,
+      "step": 18833
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2269663449168894e-05,
+      "loss": 0.6201,
+      "step": 18834
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2268969214582071e-05,
+      "loss": 0.5753,
+      "step": 18835
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2268274968465952e-05,
+      "loss": 0.5987,
+      "step": 18836
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.226758071082406e-05,
+      "loss": 0.6353,
+      "step": 18837
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2266886441659929e-05,
+      "loss": 0.6607,
+      "step": 18838
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2266192160977086e-05,
+      "loss": 0.6464,
+      "step": 18839
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2265497868779054e-05,
+      "loss": 0.6152,
+      "step": 18840
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2264803565069365e-05,
+      "loss": 0.6382,
+      "step": 18841
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2264109249851548e-05,
+      "loss": 0.6124,
+      "step": 18842
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2263414923129124e-05,
+      "loss": 0.5809,
+      "step": 18843
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2262720584905632e-05,
+      "loss": 0.586,
+      "step": 18844
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.226202623518459e-05,
+      "loss": 0.5985,
+      "step": 18845
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2261331873969529e-05,
+      "loss": 0.5599,
+      "step": 18846
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2260637501263982e-05,
+      "loss": 0.5819,
+      "step": 18847
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2259943117071468e-05,
+      "loss": 0.6474,
+      "step": 18848
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2259248721395527e-05,
+      "loss": 0.6305,
+      "step": 18849
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.225855431423968e-05,
+      "loss": 0.6596,
+      "step": 18850
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2257859895607455e-05,
+      "loss": 0.5883,
+      "step": 18851
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2257165465502385e-05,
+      "loss": 0.624,
+      "step": 18852
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2256471023927996e-05,
+      "loss": 0.6023,
+      "step": 18853
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2255776570887816e-05,
+      "loss": 0.5966,
+      "step": 18854
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2255082106385376e-05,
+      "loss": 0.5872,
+      "step": 18855
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.22543876304242e-05,
+      "loss": 0.6292,
+      "step": 18856
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2253693143007824e-05,
+      "loss": 0.6153,
+      "step": 18857
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2252998644139771e-05,
+      "loss": 0.6478,
+      "step": 18858
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2252304133823574e-05,
+      "loss": 0.6076,
+      "step": 18859
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2251609612062758e-05,
+      "loss": 0.5837,
+      "step": 18860
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2250915078860857e-05,
+      "loss": 0.6118,
+      "step": 18861
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2250220534221396e-05,
+      "loss": 0.6359,
+      "step": 18862
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2249525978147905e-05,
+      "loss": 0.5384,
+      "step": 18863
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2248831410643912e-05,
+      "loss": 0.5953,
+      "step": 18864
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.224813683171295e-05,
+      "loss": 0.6036,
+      "step": 18865
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2247442241358547e-05,
+      "loss": 0.603,
+      "step": 18866
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2246747639584229e-05,
+      "loss": 0.5768,
+      "step": 18867
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2246053026393529e-05,
+      "loss": 0.5666,
+      "step": 18868
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2245358401789977e-05,
+      "loss": 0.5893,
+      "step": 18869
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.22446637657771e-05,
+      "loss": 0.5811,
+      "step": 18870
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2243969118358431e-05,
+      "loss": 0.6231,
+      "step": 18871
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2243274459537497e-05,
+      "loss": 0.5924,
+      "step": 18872
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2242579789317828e-05,
+      "loss": 0.5763,
+      "step": 18873
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2241885107702956e-05,
+      "loss": 0.5654,
+      "step": 18874
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2241190414696406e-05,
+      "loss": 0.637,
+      "step": 18875
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2240495710301711e-05,
+      "loss": 0.5779,
+      "step": 18876
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2239800994522406e-05,
+      "loss": 0.6015,
+      "step": 18877
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2239106267362014e-05,
+      "loss": 0.6279,
+      "step": 18878
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2238411528824066e-05,
+      "loss": 0.5696,
+      "step": 18879
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2237716778912095e-05,
+      "loss": 0.5776,
+      "step": 18880
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.2237022017629629e-05,
+      "loss": 0.6445,
+      "step": 18881
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.22363272449802e-05,
+      "loss": 0.6794,
+      "step": 18882
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2235632460967335e-05,
+      "loss": 0.5899,
+      "step": 18883
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2234937665594566e-05,
+      "loss": 0.5518,
+      "step": 18884
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2234242858865429e-05,
+      "loss": 0.5423,
+      "step": 18885
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2233548040783446e-05,
+      "loss": 0.6442,
+      "step": 18886
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2232853211352151e-05,
+      "loss": 0.5877,
+      "step": 18887
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2232158370575076e-05,
+      "loss": 0.6237,
+      "step": 18888
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.223146351845575e-05,
+      "loss": 0.6247,
+      "step": 18889
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2230768654997707e-05,
+      "loss": 0.6185,
+      "step": 18890
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.223007378020447e-05,
+      "loss": 0.5814,
+      "step": 18891
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2229378894079578e-05,
+      "loss": 0.5833,
+      "step": 18892
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.222868399662656e-05,
+      "loss": 0.5514,
+      "step": 18893
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2227989087848943e-05,
+      "loss": 0.6281,
+      "step": 18894
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2227294167750263e-05,
+      "loss": 0.5666,
+      "step": 18895
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2226599236334052e-05,
+      "loss": 0.6025,
+      "step": 18896
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2225904293603832e-05,
+      "loss": 0.59,
+      "step": 18897
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2225209339563144e-05,
+      "loss": 0.6518,
+      "step": 18898
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2224514374215515e-05,
+      "loss": 0.7116,
+      "step": 18899
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2223819397564479e-05,
+      "loss": 0.5764,
+      "step": 18900
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2223124409613565e-05,
+      "loss": 0.5748,
+      "step": 18901
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2222429410366302e-05,
+      "loss": 0.5941,
+      "step": 18902
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2221734399826226e-05,
+      "loss": 0.5685,
+      "step": 18903
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2221039377996866e-05,
+      "loss": 0.5959,
+      "step": 18904
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2220344344881754e-05,
+      "loss": 0.6919,
+      "step": 18905
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2219649300484424e-05,
+      "loss": 0.6033,
+      "step": 18906
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2218954244808406e-05,
+      "loss": 0.6342,
+      "step": 18907
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2218259177857228e-05,
+      "loss": 0.6143,
+      "step": 18908
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2217564099634429e-05,
+      "loss": 0.6284,
+      "step": 18909
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2216869010143535e-05,
+      "loss": 0.6197,
+      "step": 18910
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2216173909388082e-05,
+      "loss": 0.5621,
+      "step": 18911
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.22154787973716e-05,
+      "loss": 0.6589,
+      "step": 18912
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.221478367409762e-05,
+      "loss": 0.5796,
+      "step": 18913
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2214088539569677e-05,
+      "loss": 0.6336,
+      "step": 18914
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2213393393791304e-05,
+      "loss": 0.5867,
+      "step": 18915
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2212698236766028e-05,
+      "loss": 0.6455,
+      "step": 18916
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2212003068497385e-05,
+      "loss": 0.6117,
+      "step": 18917
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2211307888988905e-05,
+      "loss": 0.6098,
+      "step": 18918
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2210612698244125e-05,
+      "loss": 0.5438,
+      "step": 18919
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2209917496266573e-05,
+      "loss": 0.6723,
+      "step": 18920
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2209222283059783e-05,
+      "loss": 0.5941,
+      "step": 18921
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2208527058627286e-05,
+      "loss": 0.6284,
+      "step": 18922
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.220783182297262e-05,
+      "loss": 0.6633,
+      "step": 18923
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.220713657609931e-05,
+      "loss": 0.61,
+      "step": 18924
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2206441318010896e-05,
+      "loss": 0.6531,
+      "step": 18925
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2205746048710905e-05,
+      "loss": 0.5173,
+      "step": 18926
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2205050768202873e-05,
+      "loss": 0.5443,
+      "step": 18927
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2204355476490332e-05,
+      "loss": 0.6378,
+      "step": 18928
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2203660173576814e-05,
+      "loss": 0.5798,
+      "step": 18929
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2202964859465856e-05,
+      "loss": 0.6349,
+      "step": 18930
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.220226953416099e-05,
+      "loss": 0.6024,
+      "step": 18931
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2201574197665742e-05,
+      "loss": 0.604,
+      "step": 18932
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2200878849983654e-05,
+      "loss": 0.57,
+      "step": 18933
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2200183491118253e-05,
+      "loss": 0.5855,
+      "step": 18934
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2199488121073078e-05,
+      "loss": 0.5178,
+      "step": 18935
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2198792739851662e-05,
+      "loss": 0.6177,
+      "step": 18936
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.219809734745753e-05,
+      "loss": 0.6309,
+      "step": 18937
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.219740194389423e-05,
+      "loss": 0.5739,
+      "step": 18938
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2196706529165281e-05,
+      "loss": 0.6109,
+      "step": 18939
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2196011103274226e-05,
+      "loss": 0.5933,
+      "step": 18940
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2195315666224592e-05,
+      "loss": 0.6344,
+      "step": 18941
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.219462021801992e-05,
+      "loss": 0.6635,
+      "step": 18942
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2193924758663736e-05,
+      "loss": 0.5853,
+      "step": 18943
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.219322928815958e-05,
+      "loss": 0.6015,
+      "step": 18944
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2192533806510985e-05,
+      "loss": 0.5942,
+      "step": 18945
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2191838313721481e-05,
+      "loss": 0.6238,
+      "step": 18946
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2191142809794605e-05,
+      "loss": 0.6576,
+      "step": 18947
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2190447294733892e-05,
+      "loss": 0.5805,
+      "step": 18948
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2189751768542872e-05,
+      "loss": 0.5828,
+      "step": 18949
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2189056231225085e-05,
+      "loss": 0.6263,
+      "step": 18950
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2188360682784059e-05,
+      "loss": 0.5929,
+      "step": 18951
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2187665123223335e-05,
+      "loss": 0.5606,
+      "step": 18952
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2186969552546443e-05,
+      "loss": 0.5971,
+      "step": 18953
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2186273970756917e-05,
+      "loss": 0.6129,
+      "step": 18954
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2185578377858293e-05,
+      "loss": 0.5919,
+      "step": 18955
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2184882773854105e-05,
+      "loss": 0.6679,
+      "step": 18956
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2184187158747892e-05,
+      "loss": 0.5652,
+      "step": 18957
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.218349153254318e-05,
+      "loss": 0.5836,
+      "step": 18958
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2182795895243509e-05,
+      "loss": 0.665,
+      "step": 18959
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.218210024685241e-05,
+      "loss": 0.6077,
+      "step": 18960
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2181404587373424e-05,
+      "loss": 0.5948,
+      "step": 18961
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2180708916810082e-05,
+      "loss": 0.6169,
+      "step": 18962
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.218001323516592e-05,
+      "loss": 0.5707,
+      "step": 18963
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.217931754244447e-05,
+      "loss": 0.6179,
+      "step": 18964
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2178621838649269e-05,
+      "loss": 0.5803,
+      "step": 18965
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2177926123783857e-05,
+      "loss": 0.6304,
+      "step": 18966
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.217723039785176e-05,
+      "loss": 0.6099,
+      "step": 18967
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2176534660856519e-05,
+      "loss": 0.6268,
+      "step": 18968
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.217583891280167e-05,
+      "loss": 0.6144,
+      "step": 18969
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2175143153690743e-05,
+      "loss": 0.5877,
+      "step": 18970
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.217444738352728e-05,
+      "loss": 0.5649,
+      "step": 18971
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.217375160231481e-05,
+      "loss": 0.592,
+      "step": 18972
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2173055810056871e-05,
+      "loss": 0.6144,
+      "step": 18973
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2172360006757003e-05,
+      "loss": 0.6426,
+      "step": 18974
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2171664192418734e-05,
+      "loss": 0.6128,
+      "step": 18975
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2170968367045602e-05,
+      "loss": 0.6148,
+      "step": 18976
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2170272530641147e-05,
+      "loss": 0.6154,
+      "step": 18977
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2169576683208903e-05,
+      "loss": 0.6388,
+      "step": 18978
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2168880824752401e-05,
+      "loss": 0.5643,
+      "step": 18979
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2168184955275184e-05,
+      "loss": 0.6421,
+      "step": 18980
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2167489074780781e-05,
+      "loss": 0.5918,
+      "step": 18981
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2166793183272733e-05,
+      "loss": 0.6233,
+      "step": 18982
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2166097280754575e-05,
+      "loss": 0.6009,
+      "step": 18983
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2165401367229838e-05,
+      "loss": 0.5466,
+      "step": 18984
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2164705442702067e-05,
+      "loss": 0.5873,
+      "step": 18985
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2164009507174793e-05,
+      "loss": 0.5976,
+      "step": 18986
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2163313560651548e-05,
+      "loss": 0.6258,
+      "step": 18987
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2162617603135881e-05,
+      "loss": 0.6045,
+      "step": 18988
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2161921634631315e-05,
+      "loss": 0.6376,
+      "step": 18989
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2161225655141395e-05,
+      "loss": 0.6461,
+      "step": 18990
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2160529664669653e-05,
+      "loss": 0.6437,
+      "step": 18991
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2159833663219627e-05,
+      "loss": 0.5737,
+      "step": 18992
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2159137650794854e-05,
+      "loss": 0.6296,
+      "step": 18993
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2158441627398868e-05,
+      "loss": 0.604,
+      "step": 18994
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.215774559303521e-05,
+      "loss": 0.5779,
+      "step": 18995
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2157049547707417e-05,
+      "loss": 0.5809,
+      "step": 18996
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2156353491419025e-05,
+      "loss": 0.6304,
+      "step": 18997
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2155657424173563e-05,
+      "loss": 0.6266,
+      "step": 18998
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2154961345974578e-05,
+      "loss": 0.6166,
+      "step": 18999
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2154265256825601e-05,
+      "loss": 0.6224,
+      "step": 19000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2153569156730176e-05,
+      "loss": 0.5857,
+      "step": 19001
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2152873045691832e-05,
+      "loss": 0.5839,
+      "step": 19002
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2152176923714108e-05,
+      "loss": 0.5719,
+      "step": 19003
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2151480790800547e-05,
+      "loss": 0.6075,
+      "step": 19004
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2150784646954678e-05,
+      "loss": 0.6285,
+      "step": 19005
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2150088492180044e-05,
+      "loss": 0.6628,
+      "step": 19006
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.214939232648018e-05,
+      "loss": 0.6015,
+      "step": 19007
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2148696149858626e-05,
+      "loss": 0.5488,
+      "step": 19008
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2147999962318916e-05,
+      "loss": 0.6372,
+      "step": 19009
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.214730376386459e-05,
+      "loss": 0.6237,
+      "step": 19010
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2146607554499181e-05,
+      "loss": 0.6506,
+      "step": 19011
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2145911334226235e-05,
+      "loss": 0.5514,
+      "step": 19012
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2145215103049284e-05,
+      "loss": 0.6105,
+      "step": 19013
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2144518860971865e-05,
+      "loss": 0.622,
+      "step": 19014
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.214382260799752e-05,
+      "loss": 0.6064,
+      "step": 19015
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2143126344129783e-05,
+      "loss": 0.5647,
+      "step": 19016
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2142430069372194e-05,
+      "loss": 0.629,
+      "step": 19017
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2141733783728293e-05,
+      "loss": 0.6009,
+      "step": 19018
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.214103748720161e-05,
+      "loss": 0.6219,
+      "step": 19019
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2140341179795694e-05,
+      "loss": 0.6775,
+      "step": 19020
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2139644861514073e-05,
+      "loss": 0.5479,
+      "step": 19021
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.213894853236029e-05,
+      "loss": 0.6531,
+      "step": 19022
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2138252192337888e-05,
+      "loss": 0.6293,
+      "step": 19023
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2137555841450394e-05,
+      "loss": 0.6156,
+      "step": 19024
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2136859479701357e-05,
+      "loss": 0.6457,
+      "step": 19025
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.213616310709431e-05,
+      "loss": 0.6153,
+      "step": 19026
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2135466723632791e-05,
+      "loss": 0.6068,
+      "step": 19027
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.2134770329320344e-05,
+      "loss": 0.5238,
+      "step": 19028
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.21340739241605e-05,
+      "loss": 0.6784,
+      "step": 19029
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2133377508156803e-05,
+      "loss": 0.5783,
+      "step": 19030
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.213268108131279e-05,
+      "loss": 0.6013,
+      "step": 19031
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2131984643632e-05,
+      "loss": 0.6288,
+      "step": 19032
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2131288195117972e-05,
+      "loss": 0.5958,
+      "step": 19033
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2130591735774244e-05,
+      "loss": 0.6172,
+      "step": 19034
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2129895265604355e-05,
+      "loss": 0.596,
+      "step": 19035
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2129198784611847e-05,
+      "loss": 0.556,
+      "step": 19036
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2128502292800255e-05,
+      "loss": 0.5936,
+      "step": 19037
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2127805790173118e-05,
+      "loss": 0.6033,
+      "step": 19038
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2127109276733977e-05,
+      "loss": 0.6399,
+      "step": 19039
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2126412752486372e-05,
+      "loss": 0.6019,
+      "step": 19040
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2125716217433839e-05,
+      "loss": 0.6034,
+      "step": 19041
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2125019671579923e-05,
+      "loss": 0.642,
+      "step": 19042
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2124323114928156e-05,
+      "loss": 0.6448,
+      "step": 19043
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2123626547482083e-05,
+      "loss": 0.6073,
+      "step": 19044
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.212292996924524e-05,
+      "loss": 0.6245,
+      "step": 19045
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.212223338022117e-05,
+      "loss": 0.5704,
+      "step": 19046
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.212153678041341e-05,
+      "loss": 0.5424,
+      "step": 19047
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2120840169825498e-05,
+      "loss": 0.5653,
+      "step": 19048
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.212014354846098e-05,
+      "loss": 0.5944,
+      "step": 19049
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.211944691632339e-05,
+      "loss": 0.5736,
+      "step": 19050
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2118750273416267e-05,
+      "loss": 0.6225,
+      "step": 19051
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2118053619743155e-05,
+      "loss": 0.6509,
+      "step": 19052
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2117356955307595e-05,
+      "loss": 0.6122,
+      "step": 19053
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2116660280113122e-05,
+      "loss": 0.5933,
+      "step": 19054
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2115963594163278e-05,
+      "loss": 0.6087,
+      "step": 19055
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2115266897461604e-05,
+      "loss": 0.5447,
+      "step": 19056
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2114570190011639e-05,
+      "loss": 0.6372,
+      "step": 19057
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2113873471816925e-05,
+      "loss": 0.6407,
+      "step": 19058
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2113176742880999e-05,
+      "loss": 0.5903,
+      "step": 19059
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2112480003207404e-05,
+      "loss": 0.6047,
+      "step": 19060
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2111783252799678e-05,
+      "loss": 0.5997,
+      "step": 19061
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2111086491661364e-05,
+      "loss": 0.6328,
+      "step": 19062
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2110389719796003e-05,
+      "loss": 0.6238,
+      "step": 19063
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.210969293720713e-05,
+      "loss": 0.5316,
+      "step": 19064
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.210899614389829e-05,
+      "loss": 0.5838,
+      "step": 19065
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2108299339873027e-05,
+      "loss": 0.572,
+      "step": 19066
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2107602525134872e-05,
+      "loss": 0.609,
+      "step": 19067
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2106905699687375e-05,
+      "loss": 0.6326,
+      "step": 19068
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2106208863534072e-05,
+      "loss": 0.602,
+      "step": 19069
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2105512016678505e-05,
+      "loss": 0.5944,
+      "step": 19070
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2104815159124215e-05,
+      "loss": 0.5673,
+      "step": 19071
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2104118290874742e-05,
+      "loss": 0.5765,
+      "step": 19072
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.210342141193363e-05,
+      "loss": 0.5839,
+      "step": 19073
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2102724522304416e-05,
+      "loss": 0.5801,
+      "step": 19074
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2102027621990641e-05,
+      "loss": 0.5679,
+      "step": 19075
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2101330710995854e-05,
+      "loss": 0.6175,
+      "step": 19076
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2100633789323586e-05,
+      "loss": 0.5865,
+      "step": 19077
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2099936856977381e-05,
+      "loss": 0.5217,
+      "step": 19078
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2099239913960782e-05,
+      "loss": 0.6668,
+      "step": 19079
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2098542960277335e-05,
+      "loss": 0.5883,
+      "step": 19080
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2097845995930573e-05,
+      "loss": 0.5907,
+      "step": 19081
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2097149020924044e-05,
+      "loss": 0.6035,
+      "step": 19082
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.209645203526128e-05,
+      "loss": 0.6015,
+      "step": 19083
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2095755038945835e-05,
+      "loss": 0.5775,
+      "step": 19084
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2095058031981244e-05,
+      "loss": 0.5995,
+      "step": 19085
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2094361014371048e-05,
+      "loss": 0.6368,
+      "step": 19086
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.209366398611879e-05,
+      "loss": 0.6081,
+      "step": 19087
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2092966947228015e-05,
+      "loss": 0.579,
+      "step": 19088
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.209226989770226e-05,
+      "loss": 0.6614,
+      "step": 19089
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2091572837545072e-05,
+      "loss": 0.626,
+      "step": 19090
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2090875766759987e-05,
+      "loss": 0.5882,
+      "step": 19091
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.209017868535055e-05,
+      "loss": 0.5985,
+      "step": 19092
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2089481593320303e-05,
+      "loss": 0.6047,
+      "step": 19093
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2088784490672788e-05,
+      "loss": 0.5504,
+      "step": 19094
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2088087377411549e-05,
+      "loss": 0.6322,
+      "step": 19095
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2087390253540124e-05,
+      "loss": 0.5729,
+      "step": 19096
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2086693119062058e-05,
+      "loss": 0.6593,
+      "step": 19097
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2085995973980893e-05,
+      "loss": 0.6095,
+      "step": 19098
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2085298818300176e-05,
+      "loss": 0.5455,
+      "step": 19099
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.208460165202344e-05,
+      "loss": 0.6077,
+      "step": 19100
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2083904475154234e-05,
+      "loss": 0.5948,
+      "step": 19101
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.20832072876961e-05,
+      "loss": 0.6843,
+      "step": 19102
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2082510089652577e-05,
+      "loss": 0.6401,
+      "step": 19103
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2081812881027213e-05,
+      "loss": 0.6219,
+      "step": 19104
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2081115661823547e-05,
+      "loss": 0.6257,
+      "step": 19105
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.208041843204512e-05,
+      "loss": 0.6433,
+      "step": 19106
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2079721191695484e-05,
+      "loss": 0.6306,
+      "step": 19107
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.207902394077817e-05,
+      "loss": 0.6091,
+      "step": 19108
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.207832667929673e-05,
+      "loss": 0.5935,
+      "step": 19109
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.20776294072547e-05,
+      "loss": 0.5803,
+      "step": 19110
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2076932124655628e-05,
+      "loss": 0.6193,
+      "step": 19111
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2076234831503058e-05,
+      "loss": 0.5667,
+      "step": 19112
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.207553752780053e-05,
+      "loss": 0.6055,
+      "step": 19113
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2074840213551586e-05,
+      "loss": 0.6214,
+      "step": 19114
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2074142888759772e-05,
+      "loss": 0.6169,
+      "step": 19115
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2073445553428632e-05,
+      "loss": 0.6262,
+      "step": 19116
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2072748207561707e-05,
+      "loss": 0.6394,
+      "step": 19117
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2072050851162542e-05,
+      "loss": 0.6609,
+      "step": 19118
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2071353484234678e-05,
+      "loss": 0.5805,
+      "step": 19119
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2070656106781662e-05,
+      "loss": 0.6233,
+      "step": 19120
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2069958718807033e-05,
+      "loss": 0.5967,
+      "step": 19121
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.206926132031434e-05,
+      "loss": 0.5986,
+      "step": 19122
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2068563911307126e-05,
+      "loss": 0.6221,
+      "step": 19123
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2067866491788929e-05,
+      "loss": 0.5843,
+      "step": 19124
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2067169061763297e-05,
+      "loss": 0.5688,
+      "step": 19125
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2066471621233777e-05,
+      "loss": 0.6441,
+      "step": 19126
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2065774170203906e-05,
+      "loss": 0.5442,
+      "step": 19127
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2065076708677234e-05,
+      "loss": 0.6355,
+      "step": 19128
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2064379236657302e-05,
+      "loss": 0.586,
+      "step": 19129
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2063681754147652e-05,
+      "loss": 0.6008,
+      "step": 19130
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2062984261151834e-05,
+      "loss": 0.6138,
+      "step": 19131
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2062286757673387e-05,
+      "loss": 0.6264,
+      "step": 19132
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.206158924371586e-05,
+      "loss": 0.6477,
+      "step": 19133
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.206089171928279e-05,
+      "loss": 0.5936,
+      "step": 19134
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.206019418437773e-05,
+      "loss": 0.6044,
+      "step": 19135
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2059496639004213e-05,
+      "loss": 0.5165,
+      "step": 19136
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2058799083165797e-05,
+      "loss": 0.5836,
+      "step": 19137
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2058101516866016e-05,
+      "loss": 0.6319,
+      "step": 19138
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2057403940108423e-05,
+      "loss": 0.5789,
+      "step": 19139
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2056706352896554e-05,
+      "loss": 0.5948,
+      "step": 19140
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2056008755233959e-05,
+      "loss": 0.6168,
+      "step": 19141
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2055311147124183e-05,
+      "loss": 0.5782,
+      "step": 19142
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2054613528570765e-05,
+      "loss": 0.5947,
+      "step": 19143
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2053915899577255e-05,
+      "loss": 0.6124,
+      "step": 19144
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2053218260147199e-05,
+      "loss": 0.5508,
+      "step": 19145
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.205252061028414e-05,
+      "loss": 0.6405,
+      "step": 19146
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2051822949991622e-05,
+      "loss": 0.5272,
+      "step": 19147
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2051125279273189e-05,
+      "loss": 0.6015,
+      "step": 19148
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2050427598132388e-05,
+      "loss": 0.637,
+      "step": 19149
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2049729906572767e-05,
+      "loss": 0.5768,
+      "step": 19150
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2049032204597864e-05,
+      "loss": 0.6243,
+      "step": 19151
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.204833449221123e-05,
+      "loss": 0.6642,
+      "step": 19152
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.204763676941641e-05,
+      "loss": 0.5927,
+      "step": 19153
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2046939036216947e-05,
+      "loss": 0.6399,
+      "step": 19154
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2046241292616387e-05,
+      "loss": 0.6382,
+      "step": 19155
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2045543538618276e-05,
+      "loss": 0.5942,
+      "step": 19156
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.204484577422616e-05,
+      "loss": 0.5629,
+      "step": 19157
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2044147999443582e-05,
+      "loss": 0.5864,
+      "step": 19158
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2043450214274088e-05,
+      "loss": 0.6212,
+      "step": 19159
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2042752418721227e-05,
+      "loss": 0.6016,
+      "step": 19160
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2042054612788546e-05,
+      "loss": 0.5914,
+      "step": 19161
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2041356796479582e-05,
+      "loss": 0.5819,
+      "step": 19162
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.204065896979789e-05,
+      "loss": 0.5927,
+      "step": 19163
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2039961132747008e-05,
+      "loss": 0.6067,
+      "step": 19164
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.203926328533049e-05,
+      "loss": 0.6186,
+      "step": 19165
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2038565427551875e-05,
+      "loss": 0.5913,
+      "step": 19166
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2037867559414713e-05,
+      "loss": 0.6067,
+      "step": 19167
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2037169680922548e-05,
+      "loss": 0.5928,
+      "step": 19168
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2036471792078932e-05,
+      "loss": 0.6043,
+      "step": 19169
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2035773892887403e-05,
+      "loss": 0.6164,
+      "step": 19170
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.203507598335151e-05,
+      "loss": 0.6056,
+      "step": 19171
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2034378063474804e-05,
+      "loss": 0.6195,
+      "step": 19172
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2033680133260824e-05,
+      "loss": 0.6057,
+      "step": 19173
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.2032982192713118e-05,
+      "loss": 0.5949,
+      "step": 19174
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.203228424183524e-05,
+      "loss": 0.5806,
+      "step": 19175
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2031586280630725e-05,
+      "loss": 0.627,
+      "step": 19176
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2030888309103128e-05,
+      "loss": 0.5845,
+      "step": 19177
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.203019032725599e-05,
+      "loss": 0.6556,
+      "step": 19178
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.202949233509286e-05,
+      "loss": 0.6582,
+      "step": 19179
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2028794332617288e-05,
+      "loss": 0.6603,
+      "step": 19180
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2028096319832815e-05,
+      "loss": 0.6053,
+      "step": 19181
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2027398296742996e-05,
+      "loss": 0.6728,
+      "step": 19182
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2026700263351367e-05,
+      "loss": 0.6072,
+      "step": 19183
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2026002219661481e-05,
+      "loss": 0.5969,
+      "step": 19184
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2025304165676887e-05,
+      "loss": 0.5457,
+      "step": 19185
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2024606101401128e-05,
+      "loss": 0.6203,
+      "step": 19186
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2023908026837751e-05,
+      "loss": 0.6022,
+      "step": 19187
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2023209941990308e-05,
+      "loss": 0.6323,
+      "step": 19188
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2022511846862342e-05,
+      "loss": 0.6317,
+      "step": 19189
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2021813741457398e-05,
+      "loss": 0.637,
+      "step": 19190
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2021115625779029e-05,
+      "loss": 0.5659,
+      "step": 19191
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2020417499830779e-05,
+      "loss": 0.5899,
+      "step": 19192
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2019719363616196e-05,
+      "loss": 0.6577,
+      "step": 19193
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.201902121713883e-05,
+      "loss": 0.5788,
+      "step": 19194
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2018323060402222e-05,
+      "loss": 0.6582,
+      "step": 19195
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2017624893409924e-05,
+      "loss": 0.5815,
+      "step": 19196
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2016926716165482e-05,
+      "loss": 0.6309,
+      "step": 19197
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2016228528672446e-05,
+      "loss": 0.5596,
+      "step": 19198
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2015530330934364e-05,
+      "loss": 0.6485,
+      "step": 19199
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.201483212295478e-05,
+      "loss": 0.5473,
+      "step": 19200
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2014133904737246e-05,
+      "loss": 0.6398,
+      "step": 19201
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2013435676285305e-05,
+      "loss": 0.5853,
+      "step": 19202
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2012737437602508e-05,
+      "loss": 0.6013,
+      "step": 19203
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2012039188692406e-05,
+      "loss": 0.5896,
+      "step": 19204
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.201134092955854e-05,
+      "loss": 0.6037,
+      "step": 19205
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2010642660204462e-05,
+      "loss": 0.5965,
+      "step": 19206
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2009944380633722e-05,
+      "loss": 0.6243,
+      "step": 19207
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2009246090849865e-05,
+      "loss": 0.6999,
+      "step": 19208
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.200854779085644e-05,
+      "loss": 0.5397,
+      "step": 19209
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2007849480656994e-05,
+      "loss": 0.5319,
+      "step": 19210
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2007151160255078e-05,
+      "loss": 0.6384,
+      "step": 19211
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.200645282965424e-05,
+      "loss": 0.5952,
+      "step": 19212
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2005754488858025e-05,
+      "loss": 0.5863,
+      "step": 19213
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2005056137869988e-05,
+      "loss": 0.5916,
+      "step": 19214
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2004357776693673e-05,
+      "loss": 0.5534,
+      "step": 19215
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2003659405332626e-05,
+      "loss": 0.6075,
+      "step": 19216
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2002961023790401e-05,
+      "loss": 0.6009,
+      "step": 19217
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2002262632070542e-05,
+      "loss": 0.6061,
+      "step": 19218
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.2001564230176603e-05,
+      "loss": 0.628,
+      "step": 19219
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.200086581811213e-05,
+      "loss": 0.6351,
+      "step": 19220
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.200016739588067e-05,
+      "loss": 0.6128,
+      "step": 19221
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1999468963485774e-05,
+      "loss": 0.5283,
+      "step": 19222
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1998770520930991e-05,
+      "loss": 0.5875,
+      "step": 19223
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1998072068219869e-05,
+      "loss": 0.6473,
+      "step": 19224
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1997373605355958e-05,
+      "loss": 0.5554,
+      "step": 19225
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1996675132342808e-05,
+      "loss": 0.505,
+      "step": 19226
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1995976649183965e-05,
+      "loss": 0.5582,
+      "step": 19227
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1995278155882982e-05,
+      "loss": 0.5659,
+      "step": 19228
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1994579652443404e-05,
+      "loss": 0.6478,
+      "step": 19229
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1993881138868784e-05,
+      "loss": 0.5786,
+      "step": 19230
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.199318261516267e-05,
+      "loss": 0.5847,
+      "step": 19231
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.199248408132861e-05,
+      "loss": 0.6526,
+      "step": 19232
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1991785537370158e-05,
+      "loss": 0.6405,
+      "step": 19233
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1991086983290861e-05,
+      "loss": 0.6274,
+      "step": 19234
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1990388419094264e-05,
+      "loss": 0.5904,
+      "step": 19235
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.198968984478392e-05,
+      "loss": 0.6409,
+      "step": 19236
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1988991260363383e-05,
+      "loss": 0.5947,
+      "step": 19237
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1988292665836196e-05,
+      "loss": 0.5308,
+      "step": 19238
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1987594061205917e-05,
+      "loss": 0.632,
+      "step": 19239
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1986895446476083e-05,
+      "loss": 0.5935,
+      "step": 19240
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1986196821650253e-05,
+      "loss": 0.6482,
+      "step": 19241
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.198549818673198e-05,
+      "loss": 0.5982,
+      "step": 19242
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1984799541724805e-05,
+      "loss": 0.58,
+      "step": 19243
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1984100886632282e-05,
+      "loss": 0.6159,
+      "step": 19244
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1983402221457965e-05,
+      "loss": 0.5873,
+      "step": 19245
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1982703546205396e-05,
+      "loss": 0.6281,
+      "step": 19246
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1982004860878134e-05,
+      "loss": 0.5749,
+      "step": 19247
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1981306165479722e-05,
+      "loss": 0.6304,
+      "step": 19248
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1980607460013711e-05,
+      "loss": 0.6856,
+      "step": 19249
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.197990874448366e-05,
+      "loss": 0.5284,
+      "step": 19250
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1979210018893107e-05,
+      "loss": 0.6606,
+      "step": 19251
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1978511283245611e-05,
+      "loss": 0.6144,
+      "step": 19252
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1977812537544717e-05,
+      "loss": 0.616,
+      "step": 19253
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1977113781793983e-05,
+      "loss": 0.6325,
+      "step": 19254
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1976415015996955e-05,
+      "loss": 0.6426,
+      "step": 19255
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1975716240157178e-05,
+      "loss": 0.6477,
+      "step": 19256
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1975017454278211e-05,
+      "loss": 0.6441,
+      "step": 19257
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1974318658363601e-05,
+      "loss": 0.5913,
+      "step": 19258
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.19736198524169e-05,
+      "loss": 0.5556,
+      "step": 19259
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1972921036441659e-05,
+      "loss": 0.5795,
+      "step": 19260
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1972222210441428e-05,
+      "loss": 0.609,
+      "step": 19261
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1971523374419757e-05,
+      "loss": 0.5888,
+      "step": 19262
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.19708245283802e-05,
+      "loss": 0.6249,
+      "step": 19263
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.197012567232631e-05,
+      "loss": 0.6193,
+      "step": 19264
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1969426806261628e-05,
+      "loss": 0.6204,
+      "step": 19265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1968727930189716e-05,
+      "loss": 0.6305,
+      "step": 19266
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.196802904411412e-05,
+      "loss": 0.5673,
+      "step": 19267
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.196733014803839e-05,
+      "loss": 0.5806,
+      "step": 19268
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1966631241966083e-05,
+      "loss": 0.5615,
+      "step": 19269
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1965932325900742e-05,
+      "loss": 0.6063,
+      "step": 19270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1965233399845925e-05,
+      "loss": 0.6351,
+      "step": 19271
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1964534463805184e-05,
+      "loss": 0.6701,
+      "step": 19272
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1963835517782067e-05,
+      "loss": 0.5684,
+      "step": 19273
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1963136561780127e-05,
+      "loss": 0.6026,
+      "step": 19274
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1962437595802911e-05,
+      "loss": 0.6007,
+      "step": 19275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1961738619853978e-05,
+      "loss": 0.595,
+      "step": 19276
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1961039633936877e-05,
+      "loss": 0.5301,
+      "step": 19277
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.196034063805516e-05,
+      "loss": 0.5837,
+      "step": 19278
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1959641632212375e-05,
+      "loss": 0.6581,
+      "step": 19279
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.195894261641208e-05,
+      "loss": 0.5717,
+      "step": 19280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.195824359065782e-05,
+      "loss": 0.6271,
+      "step": 19281
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1957544554953153e-05,
+      "loss": 0.5661,
+      "step": 19282
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1956845509301631e-05,
+      "loss": 0.6309,
+      "step": 19283
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1956146453706801e-05,
+      "loss": 0.6165,
+      "step": 19284
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.195544738817222e-05,
+      "loss": 0.5703,
+      "step": 19285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1954748312701439e-05,
+      "loss": 0.5256,
+      "step": 19286
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1954049227298005e-05,
+      "loss": 0.6053,
+      "step": 19287
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1953350131965478e-05,
+      "loss": 0.6549,
+      "step": 19288
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1952651026707405e-05,
+      "loss": 0.5215,
+      "step": 19289
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1951951911527345e-05,
+      "loss": 0.61,
+      "step": 19290
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.195125278642884e-05,
+      "loss": 0.566,
+      "step": 19291
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.195055365141545e-05,
+      "loss": 0.5712,
+      "step": 19292
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1949854506490726e-05,
+      "loss": 0.5837,
+      "step": 19293
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1949155351658221e-05,
+      "loss": 0.5596,
+      "step": 19294
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1948456186921486e-05,
+      "loss": 0.5871,
+      "step": 19295
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1947757012284075e-05,
+      "loss": 0.5885,
+      "step": 19296
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1947057827749538e-05,
+      "loss": 0.6044,
+      "step": 19297
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.194635863332143e-05,
+      "loss": 0.5744,
+      "step": 19298
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1945659429003307e-05,
+      "loss": 0.6401,
+      "step": 19299
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1944960214798718e-05,
+      "loss": 0.6764,
+      "step": 19300
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1944260990711217e-05,
+      "loss": 0.6106,
+      "step": 19301
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1943561756744353e-05,
+      "loss": 0.5894,
+      "step": 19302
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1942862512901684e-05,
+      "loss": 0.6215,
+      "step": 19303
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1942163259186764e-05,
+      "loss": 0.6171,
+      "step": 19304
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1941463995603139e-05,
+      "loss": 0.5855,
+      "step": 19305
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.194076472215437e-05,
+      "loss": 0.562,
+      "step": 19306
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1940065438844006e-05,
+      "loss": 0.5871,
+      "step": 19307
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1939366145675601e-05,
+      "loss": 0.5835,
+      "step": 19308
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1938666842652711e-05,
+      "loss": 0.5699,
+      "step": 19309
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1937967529778886e-05,
+      "loss": 0.5688,
+      "step": 19310
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1937268207057683e-05,
+      "loss": 0.5705,
+      "step": 19311
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1936568874492647e-05,
+      "loss": 0.5583,
+      "step": 19312
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1935869532087341e-05,
+      "loss": 0.5655,
+      "step": 19313
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1935170179845315e-05,
+      "loss": 0.5824,
+      "step": 19314
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1934470817770123e-05,
+      "loss": 0.5448,
+      "step": 19315
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1933771445865316e-05,
+      "loss": 0.5619,
+      "step": 19316
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1933072064134449e-05,
+      "loss": 0.5751,
+      "step": 19317
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.193237267258108e-05,
+      "loss": 0.56,
+      "step": 19318
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.193167327120876e-05,
+      "loss": 0.5473,
+      "step": 19319
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.193097386002104e-05,
+      "loss": 0.5655,
+      "step": 19320
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1930274439021477e-05,
+      "loss": 0.6561,
+      "step": 19321
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1929575008213623e-05,
+      "loss": 0.6131,
+      "step": 19322
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1928875567601035e-05,
+      "loss": 0.5852,
+      "step": 19323
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1928176117187264e-05,
+      "loss": 0.5755,
+      "step": 19324
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1927476656975867e-05,
+      "loss": 0.5724,
+      "step": 19325
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1926777186970395e-05,
+      "loss": 0.6108,
+      "step": 19326
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1926077707174406e-05,
+      "loss": 0.6412,
+      "step": 19327
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.192537821759145e-05,
+      "loss": 0.6236,
+      "step": 19328
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1924678718225085e-05,
+      "loss": 0.5696,
+      "step": 19329
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1923979209078864e-05,
+      "loss": 0.6358,
+      "step": 19330
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1923279690156339e-05,
+      "loss": 0.5934,
+      "step": 19331
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.192258016146107e-05,
+      "loss": 0.5884,
+      "step": 19332
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1921880622996604e-05,
+      "loss": 0.6124,
+      "step": 19333
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1921181074766504e-05,
+      "loss": 0.546,
+      "step": 19334
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1920481516774318e-05,
+      "loss": 0.5924,
+      "step": 19335
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.19197819490236e-05,
+      "loss": 0.6359,
+      "step": 19336
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1919082371517914e-05,
+      "loss": 0.6669,
+      "step": 19337
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1918382784260803e-05,
+      "loss": 0.5998,
+      "step": 19338
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1917683187255831e-05,
+      "loss": 0.6107,
+      "step": 19339
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1916983580506547e-05,
+      "loss": 0.572,
+      "step": 19340
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1916283964016508e-05,
+      "loss": 0.6081,
+      "step": 19341
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.191558433778927e-05,
+      "loss": 0.5689,
+      "step": 19342
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1914884701828386e-05,
+      "loss": 0.5899,
+      "step": 19343
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1914185056137411e-05,
+      "loss": 0.5912,
+      "step": 19344
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1913485400719904e-05,
+      "loss": 0.5809,
+      "step": 19345
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1912785735579415e-05,
+      "loss": 0.6536,
+      "step": 19346
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1912086060719503e-05,
+      "loss": 0.6505,
+      "step": 19347
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1911386376143721e-05,
+      "loss": 0.5276,
+      "step": 19348
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1910686681855625e-05,
+      "loss": 0.5852,
+      "step": 19349
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1909986977858771e-05,
+      "loss": 0.6229,
+      "step": 19350
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1909287264156716e-05,
+      "loss": 0.5779,
+      "step": 19351
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1908587540753009e-05,
+      "loss": 0.4933,
+      "step": 19352
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1907887807651212e-05,
+      "loss": 0.6058,
+      "step": 19353
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1907188064854877e-05,
+      "loss": 0.5807,
+      "step": 19354
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1906488312367561e-05,
+      "loss": 0.5592,
+      "step": 19355
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.190578855019282e-05,
+      "loss": 0.6082,
+      "step": 19356
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.190508877833421e-05,
+      "loss": 0.6179,
+      "step": 19357
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1904388996795287e-05,
+      "loss": 0.5435,
+      "step": 19358
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1903689205579603e-05,
+      "loss": 0.5995,
+      "step": 19359
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1902989404690717e-05,
+      "loss": 0.5782,
+      "step": 19360
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1902289594132185e-05,
+      "loss": 0.6113,
+      "step": 19361
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1901589773907563e-05,
+      "loss": 0.614,
+      "step": 19362
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1900889944020407e-05,
+      "loss": 0.6676,
+      "step": 19363
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1900190104474271e-05,
+      "loss": 0.5897,
+      "step": 19364
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1899490255272712e-05,
+      "loss": 0.662,
+      "step": 19365
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1898790396419291e-05,
+      "loss": 0.5889,
+      "step": 19366
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1898090527917553e-05,
+      "loss": 0.603,
+      "step": 19367
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1897390649771066e-05,
+      "loss": 0.5443,
+      "step": 19368
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1896690761983382e-05,
+      "loss": 0.5624,
+      "step": 19369
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1895990864558056e-05,
+      "loss": 0.5697,
+      "step": 19370
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1895290957498644e-05,
+      "loss": 0.6309,
+      "step": 19371
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1894591040808706e-05,
+      "loss": 0.5855,
+      "step": 19372
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1893891114491792e-05,
+      "loss": 0.6621,
+      "step": 19373
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1893191178551464e-05,
+      "loss": 0.5957,
+      "step": 19374
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.189249123299128e-05,
+      "loss": 0.5768,
+      "step": 19375
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.189179127781479e-05,
+      "loss": 0.6744,
+      "step": 19376
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1891091313025558e-05,
+      "loss": 0.5608,
+      "step": 19377
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1890391338627134e-05,
+      "loss": 0.5808,
+      "step": 19378
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1889691354623079e-05,
+      "loss": 0.6606,
+      "step": 19379
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1888991361016951e-05,
+      "loss": 0.6417,
+      "step": 19380
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.18882913578123e-05,
+      "loss": 0.5981,
+      "step": 19381
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.188759134501269e-05,
+      "loss": 0.6031,
+      "step": 19382
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1886891322621677e-05,
+      "loss": 0.6233,
+      "step": 19383
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1886191290642815e-05,
+      "loss": 0.64,
+      "step": 19384
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1885491249079664e-05,
+      "loss": 0.6489,
+      "step": 19385
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.188479119793578e-05,
+      "loss": 0.6123,
+      "step": 19386
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1884091137214717e-05,
+      "loss": 0.6061,
+      "step": 19387
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1883391066920039e-05,
+      "loss": 0.6751,
+      "step": 19388
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1882690987055296e-05,
+      "loss": 0.6682,
+      "step": 19389
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1881990897624052e-05,
+      "loss": 0.5962,
+      "step": 19390
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1881290798629858e-05,
+      "loss": 0.5903,
+      "step": 19391
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1880590690076277e-05,
+      "loss": 0.6456,
+      "step": 19392
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1879890571966864e-05,
+      "loss": 0.5715,
+      "step": 19393
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1879190444305173e-05,
+      "loss": 0.5783,
+      "step": 19394
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1878490307094767e-05,
+      "loss": 0.6289,
+      "step": 19395
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1877790160339203e-05,
+      "loss": 0.5731,
+      "step": 19396
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1877090004042036e-05,
+      "loss": 0.5767,
+      "step": 19397
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1876389838206824e-05,
+      "loss": 0.6619,
+      "step": 19398
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1875689662837127e-05,
+      "loss": 0.6908,
+      "step": 19399
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1874989477936501e-05,
+      "loss": 0.6158,
+      "step": 19400
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1874289283508503e-05,
+      "loss": 0.5584,
+      "step": 19401
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1873589079556696e-05,
+      "loss": 0.5212,
+      "step": 19402
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.187288886608463e-05,
+      "loss": 0.5915,
+      "step": 19403
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.187218864309587e-05,
+      "loss": 0.5988,
+      "step": 19404
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.187148841059397e-05,
+      "loss": 0.6116,
+      "step": 19405
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1870788168582488e-05,
+      "loss": 0.6208,
+      "step": 19406
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1870087917064987e-05,
+      "loss": 0.598,
+      "step": 19407
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.186938765604502e-05,
+      "loss": 0.592,
+      "step": 19408
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1868687385526148e-05,
+      "loss": 0.574,
+      "step": 19409
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1867987105511926e-05,
+      "loss": 0.6894,
+      "step": 19410
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1867286816005914e-05,
+      "loss": 0.6035,
+      "step": 19411
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1866586517011673e-05,
+      "loss": 0.5843,
+      "step": 19412
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1865886208532757e-05,
+      "loss": 0.613,
+      "step": 19413
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1865185890572728e-05,
+      "loss": 0.6367,
+      "step": 19414
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1864485563135144e-05,
+      "loss": 0.6184,
+      "step": 19415
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1863785226223561e-05,
+      "loss": 0.5208,
+      "step": 19416
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.186308487984154e-05,
+      "loss": 0.6206,
+      "step": 19417
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1862384523992641e-05,
+      "loss": 0.6011,
+      "step": 19418
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1861684158680418e-05,
+      "loss": 0.5575,
+      "step": 19419
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1860983783908432e-05,
+      "loss": 0.591,
+      "step": 19420
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1860283399680246e-05,
+      "loss": 0.5669,
+      "step": 19421
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1859583005999413e-05,
+      "loss": 0.6024,
+      "step": 19422
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1858882602869497e-05,
+      "loss": 0.5962,
+      "step": 19423
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.185818219029405e-05,
+      "loss": 0.6024,
+      "step": 19424
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1857481768276635e-05,
+      "loss": 0.569,
+      "step": 19425
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1856781336820815e-05,
+      "loss": 0.6859,
+      "step": 19426
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1856080895930142e-05,
+      "loss": 0.5929,
+      "step": 19427
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1855380445608182e-05,
+      "loss": 0.5874,
+      "step": 19428
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1854679985858486e-05,
+      "loss": 0.5933,
+      "step": 19429
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1853979516684622e-05,
+      "loss": 0.5747,
+      "step": 19430
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1853279038090145e-05,
+      "loss": 0.5487,
+      "step": 19431
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.185257855007861e-05,
+      "loss": 0.5532,
+      "step": 19432
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1851878052653583e-05,
+      "loss": 0.6505,
+      "step": 19433
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1851177545818624e-05,
+      "loss": 0.6158,
+      "step": 19434
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1850477029577286e-05,
+      "loss": 0.5837,
+      "step": 19435
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1849776503933132e-05,
+      "loss": 0.5964,
+      "step": 19436
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1849075968889726e-05,
+      "loss": 0.6275,
+      "step": 19437
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.184837542445062e-05,
+      "loss": 0.5652,
+      "step": 19438
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1847674870619378e-05,
+      "loss": 0.5866,
+      "step": 19439
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1846974307399558e-05,
+      "loss": 0.5596,
+      "step": 19440
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1846273734794722e-05,
+      "loss": 0.6342,
+      "step": 19441
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.184557315280843e-05,
+      "loss": 0.5865,
+      "step": 19442
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1844872561444236e-05,
+      "loss": 0.6469,
+      "step": 19443
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1844171960705705e-05,
+      "loss": 0.6157,
+      "step": 19444
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.18434713505964e-05,
+      "loss": 0.6351,
+      "step": 19445
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1842770731119874e-05,
+      "loss": 0.6196,
+      "step": 19446
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1842070102279692e-05,
+      "loss": 0.581,
+      "step": 19447
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1841369464079413e-05,
+      "loss": 0.5894,
+      "step": 19448
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1840668816522596e-05,
+      "loss": 0.6313,
+      "step": 19449
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.18399681596128e-05,
+      "loss": 0.6204,
+      "step": 19450
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1839267493353588e-05,
+      "loss": 0.5868,
+      "step": 19451
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1838566817748519e-05,
+      "loss": 0.5112,
+      "step": 19452
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1837866132801155e-05,
+      "loss": 0.5885,
+      "step": 19453
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.183716543851505e-05,
+      "loss": 0.6524,
+      "step": 19454
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1836464734893773e-05,
+      "loss": 0.5948,
+      "step": 19455
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1835764021940883e-05,
+      "loss": 0.6314,
+      "step": 19456
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1835063299659935e-05,
+      "loss": 0.6335,
+      "step": 19457
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1834362568054493e-05,
+      "loss": 0.6517,
+      "step": 19458
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1833661827128118e-05,
+      "loss": 0.6269,
+      "step": 19459
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1832961076884371e-05,
+      "loss": 0.5283,
+      "step": 19460
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1832260317326813e-05,
+      "loss": 0.5745,
+      "step": 19461
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1831559548459002e-05,
+      "loss": 0.6226,
+      "step": 19462
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1830858770284499e-05,
+      "loss": 0.6689,
+      "step": 19463
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1830157982806868e-05,
+      "loss": 0.56,
+      "step": 19464
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1829457186029668e-05,
+      "loss": 0.5996,
+      "step": 19465
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1828756379956463e-05,
+      "loss": 0.5451,
+      "step": 19466
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1828055564590808e-05,
+      "loss": 0.6389,
+      "step": 19467
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.182735473993627e-05,
+      "loss": 0.601,
+      "step": 19468
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1826653905996406e-05,
+      "loss": 0.608,
+      "step": 19469
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.182595306277478e-05,
+      "loss": 0.5475,
+      "step": 19470
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1825252210274949e-05,
+      "loss": 0.6489,
+      "step": 19471
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1824551348500479e-05,
+      "loss": 0.6353,
+      "step": 19472
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1823850477454927e-05,
+      "loss": 0.5569,
+      "step": 19473
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1823149597141856e-05,
+      "loss": 0.5947,
+      "step": 19474
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1822448707564832e-05,
+      "loss": 0.5759,
+      "step": 19475
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1821747808727411e-05,
+      "loss": 0.6441,
+      "step": 19476
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1821046900633155e-05,
+      "loss": 0.5409,
+      "step": 19477
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1820345983285626e-05,
+      "loss": 0.6153,
+      "step": 19478
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1819645056688384e-05,
+      "loss": 0.6195,
+      "step": 19479
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1818944120844998e-05,
+      "loss": 0.6301,
+      "step": 19480
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1818243175759019e-05,
+      "loss": 0.5705,
+      "step": 19481
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1817542221434018e-05,
+      "loss": 0.5844,
+      "step": 19482
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.181684125787355e-05,
+      "loss": 0.5689,
+      "step": 19483
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.181614028508118e-05,
+      "loss": 0.6522,
+      "step": 19484
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.181543930306047e-05,
+      "loss": 0.579,
+      "step": 19485
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1814738311814982e-05,
+      "loss": 0.5815,
+      "step": 19486
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1814037311348276e-05,
+      "loss": 0.6257,
+      "step": 19487
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1813336301663914e-05,
+      "loss": 0.5733,
+      "step": 19488
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1812635282765461e-05,
+      "loss": 0.6441,
+      "step": 19489
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1811934254656477e-05,
+      "loss": 0.6025,
+      "step": 19490
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1811233217340526e-05,
+      "loss": 0.6164,
+      "step": 19491
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1810532170821166e-05,
+      "loss": 0.6519,
+      "step": 19492
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1809831115101962e-05,
+      "loss": 0.6031,
+      "step": 19493
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1809130050186477e-05,
+      "loss": 0.6392,
+      "step": 19494
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.180842897607827e-05,
+      "loss": 0.6211,
+      "step": 19495
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.180772789278091e-05,
+      "loss": 0.5947,
+      "step": 19496
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1807026800297952e-05,
+      "loss": 0.6107,
+      "step": 19497
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.180632569863296e-05,
+      "loss": 0.6269,
+      "step": 19498
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1805624587789501e-05,
+      "loss": 0.585,
+      "step": 19499
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1804923467771134e-05,
+      "loss": 0.6091,
+      "step": 19500
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1804222338581424e-05,
+      "loss": 0.5741,
+      "step": 19501
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.180352120022393e-05,
+      "loss": 0.6115,
+      "step": 19502
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1802820052702217e-05,
+      "loss": 0.6061,
+      "step": 19503
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1802118896019847e-05,
+      "loss": 0.557,
+      "step": 19504
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1801417730180385e-05,
+      "loss": 0.6,
+      "step": 19505
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1800716555187389e-05,
+      "loss": 0.6106,
+      "step": 19506
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1800015371044426e-05,
+      "loss": 0.572,
+      "step": 19507
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1799314177755059e-05,
+      "loss": 0.6328,
+      "step": 19508
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.179861297532285e-05,
+      "loss": 0.6038,
+      "step": 19509
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.179791176375136e-05,
+      "loss": 0.6077,
+      "step": 19510
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1797210543044153e-05,
+      "loss": 0.6119,
+      "step": 19511
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1796509313204794e-05,
+      "loss": 0.58,
+      "step": 19512
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1795808074236846e-05,
+      "loss": 0.6372,
+      "step": 19513
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.179510682614387e-05,
+      "loss": 0.5315,
+      "step": 19514
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1794405568929432e-05,
+      "loss": 0.5637,
+      "step": 19515
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.179370430259709e-05,
+      "loss": 0.6845,
+      "step": 19516
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1793003027150413e-05,
+      "loss": 0.64,
+      "step": 19517
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1792301742592967e-05,
+      "loss": 0.6498,
+      "step": 19518
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1791600448928305e-05,
+      "loss": 0.5783,
+      "step": 19519
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1790899146159997e-05,
+      "loss": 0.6354,
+      "step": 19520
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.179019783429161e-05,
+      "loss": 0.6292,
+      "step": 19521
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.17894965133267e-05,
+      "loss": 0.6095,
+      "step": 19522
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1788795183268835e-05,
+      "loss": 0.5497,
+      "step": 19523
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.178809384412158e-05,
+      "loss": 0.5403,
+      "step": 19524
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1787392495888491e-05,
+      "loss": 0.6017,
+      "step": 19525
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1786691138573142e-05,
+      "loss": 0.5923,
+      "step": 19526
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1785989772179094e-05,
+      "loss": 0.6069,
+      "step": 19527
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1785288396709905e-05,
+      "loss": 0.5705,
+      "step": 19528
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1784587012169145e-05,
+      "loss": 0.628,
+      "step": 19529
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1783885618560373e-05,
+      "loss": 0.582,
+      "step": 19530
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1783184215887159e-05,
+      "loss": 0.594,
+      "step": 19531
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.178248280415306e-05,
+      "loss": 0.6016,
+      "step": 19532
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1781781383361646e-05,
+      "loss": 0.6424,
+      "step": 19533
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1781079953516479e-05,
+      "loss": 0.5753,
+      "step": 19534
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1780378514621123e-05,
+      "loss": 0.6085,
+      "step": 19535
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.177967706667914e-05,
+      "loss": 0.6071,
+      "step": 19536
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.17789756096941e-05,
+      "loss": 0.5915,
+      "step": 19537
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1778274143669562e-05,
+      "loss": 0.5476,
+      "step": 19538
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1777572668609091e-05,
+      "loss": 0.6695,
+      "step": 19539
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1776871184516256e-05,
+      "loss": 0.5786,
+      "step": 19540
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1776169691394617e-05,
+      "loss": 0.6295,
+      "step": 19541
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.177546818924774e-05,
+      "loss": 0.6315,
+      "step": 19542
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1774766678079188e-05,
+      "loss": 0.6217,
+      "step": 19543
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1774065157892527e-05,
+      "loss": 0.5815,
+      "step": 19544
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1773363628691323e-05,
+      "loss": 0.6385,
+      "step": 19545
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1772662090479137e-05,
+      "loss": 0.5866,
+      "step": 19546
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1771960543259537e-05,
+      "loss": 0.6036,
+      "step": 19547
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.177125898703609e-05,
+      "loss": 0.5906,
+      "step": 19548
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1770557421812351e-05,
+      "loss": 0.5973,
+      "step": 19549
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1769855847591895e-05,
+      "loss": 0.594,
+      "step": 19550
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.176915426437828e-05,
+      "loss": 0.6035,
+      "step": 19551
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1768452672175075e-05,
+      "loss": 0.6201,
+      "step": 19552
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1767751070985847e-05,
+      "loss": 0.5905,
+      "step": 19553
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1767049460814155e-05,
+      "loss": 0.5864,
+      "step": 19554
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1766347841663568e-05,
+      "loss": 0.5856,
+      "step": 19555
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1765646213537652e-05,
+      "loss": 0.6287,
+      "step": 19556
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1764944576439966e-05,
+      "loss": 0.6744,
+      "step": 19557
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1764242930374084e-05,
+      "loss": 0.5932,
+      "step": 19558
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1763541275343566e-05,
+      "loss": 0.6195,
+      "step": 19559
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1762839611351978e-05,
+      "loss": 0.5554,
+      "step": 19560
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1762137938402886e-05,
+      "loss": 0.5763,
+      "step": 19561
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1761436256499854e-05,
+      "loss": 0.5956,
+      "step": 19562
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1760734565646448e-05,
+      "loss": 0.6095,
+      "step": 19563
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1760032865846239e-05,
+      "loss": 0.6077,
+      "step": 19564
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.175933115710278e-05,
+      "loss": 0.6412,
+      "step": 19565
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1758629439419652e-05,
+      "loss": 0.5657,
+      "step": 19566
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1757927712800407e-05,
+      "loss": 0.6665,
+      "step": 19567
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1757225977248618e-05,
+      "loss": 0.6063,
+      "step": 19568
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1756524232767851e-05,
+      "loss": 0.5798,
+      "step": 19569
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1755822479361668e-05,
+      "loss": 0.5526,
+      "step": 19570
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1755120717033637e-05,
+      "loss": 0.5635,
+      "step": 19571
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1754418945787324e-05,
+      "loss": 0.606,
+      "step": 19572
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1753717165626292e-05,
+      "loss": 0.5624,
+      "step": 19573
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1753015376554111e-05,
+      "loss": 0.5985,
+      "step": 19574
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1752313578574349e-05,
+      "loss": 0.6546,
+      "step": 19575
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1751611771690566e-05,
+      "loss": 0.6333,
+      "step": 19576
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1750909955906331e-05,
+      "loss": 0.656,
+      "step": 19577
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1750208131225207e-05,
+      "loss": 0.6206,
+      "step": 19578
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1749506297650765e-05,
+      "loss": 0.5902,
+      "step": 19579
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.174880445518657e-05,
+      "loss": 0.6896,
+      "step": 19580
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1748102603836187e-05,
+      "loss": 0.558,
+      "step": 19581
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.174740074360318e-05,
+      "loss": 0.6048,
+      "step": 19582
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1746698874491123e-05,
+      "loss": 0.689,
+      "step": 19583
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1745996996503572e-05,
+      "loss": 0.5926,
+      "step": 19584
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1745295109644104e-05,
+      "loss": 0.598,
+      "step": 19585
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1744593213916277e-05,
+      "loss": 0.5821,
+      "step": 19586
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1743891309323662e-05,
+      "loss": 0.688,
+      "step": 19587
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1743189395869828e-05,
+      "loss": 0.6702,
+      "step": 19588
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1742487473558332e-05,
+      "loss": 0.6397,
+      "step": 19589
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1741785542392749e-05,
+      "loss": 0.634,
+      "step": 19590
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1741083602376646e-05,
+      "loss": 0.6503,
+      "step": 19591
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1740381653513585e-05,
+      "loss": 0.6082,
+      "step": 19592
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1739679695807134e-05,
+      "loss": 0.5959,
+      "step": 19593
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1738977729260866e-05,
+      "loss": 0.5787,
+      "step": 19594
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1738275753878336e-05,
+      "loss": 0.5699,
+      "step": 19595
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1737573769663124e-05,
+      "loss": 0.7071,
+      "step": 19596
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1736871776618785e-05,
+      "loss": 0.5715,
+      "step": 19597
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1736169774748895e-05,
+      "loss": 0.6508,
+      "step": 19598
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.173546776405702e-05,
+      "loss": 0.6557,
+      "step": 19599
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1734765744546722e-05,
+      "loss": 0.5848,
+      "step": 19600
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.173406371622157e-05,
+      "loss": 0.5967,
+      "step": 19601
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1733361679085138e-05,
+      "loss": 0.6248,
+      "step": 19602
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1732659633140982e-05,
+      "loss": 0.605,
+      "step": 19603
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1731957578392678e-05,
+      "loss": 0.6213,
+      "step": 19604
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1731255514843789e-05,
+      "loss": 0.5713,
+      "step": 19605
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1730553442497885e-05,
+      "loss": 0.6053,
+      "step": 19606
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1729851361358533e-05,
+      "loss": 0.5579,
+      "step": 19607
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1729149271429297e-05,
+      "loss": 0.5858,
+      "step": 19608
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1728447172713748e-05,
+      "loss": 0.5795,
+      "step": 19609
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1727745065215453e-05,
+      "loss": 0.6188,
+      "step": 19610
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1727042948937978e-05,
+      "loss": 0.5825,
+      "step": 19611
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1726340823884891e-05,
+      "loss": 0.5787,
+      "step": 19612
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1725638690059767e-05,
+      "loss": 0.6301,
+      "step": 19613
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1724936547466161e-05,
+      "loss": 0.6046,
+      "step": 19614
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.172423439610765e-05,
+      "loss": 0.5904,
+      "step": 19615
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1723532235987799e-05,
+      "loss": 0.5855,
+      "step": 19616
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1722830067110174e-05,
+      "loss": 0.6037,
+      "step": 19617
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1722127889478347e-05,
+      "loss": 0.5525,
+      "step": 19618
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1721425703095884e-05,
+      "loss": 0.6187,
+      "step": 19619
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1720723507966349e-05,
+      "loss": 0.5959,
+      "step": 19620
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1720021304093319e-05,
+      "loss": 0.6574,
+      "step": 19621
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1719319091480355e-05,
+      "loss": 0.5873,
+      "step": 19622
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1718616870131028e-05,
+      "loss": 0.5777,
+      "step": 19623
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1717914640048903e-05,
+      "loss": 0.5467,
+      "step": 19624
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1717212401237554e-05,
+      "loss": 0.5993,
+      "step": 19625
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1716510153700545e-05,
+      "loss": 0.5576,
+      "step": 19626
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1715807897441445e-05,
+      "loss": 0.6242,
+      "step": 19627
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1715105632463819e-05,
+      "loss": 0.5803,
+      "step": 19628
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1714403358771244e-05,
+      "loss": 0.6178,
+      "step": 19629
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1713701076367281e-05,
+      "loss": 0.6016,
+      "step": 19630
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1712998785255501e-05,
+      "loss": 0.5855,
+      "step": 19631
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1712296485439472e-05,
+      "loss": 0.5407,
+      "step": 19632
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1711594176922762e-05,
+      "loss": 0.5995,
+      "step": 19633
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1710891859708946e-05,
+      "loss": 0.5931,
+      "step": 19634
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1710189533801581e-05,
+      "loss": 0.6297,
+      "step": 19635
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1709487199204246e-05,
+      "loss": 0.5688,
+      "step": 19636
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1708784855920505e-05,
+      "loss": 0.6024,
+      "step": 19637
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1708082503953925e-05,
+      "loss": 0.5893,
+      "step": 19638
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1707380143308082e-05,
+      "loss": 0.6708,
+      "step": 19639
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1706677773986538e-05,
+      "loss": 0.5944,
+      "step": 19640
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1705975395992866e-05,
+      "loss": 0.6319,
+      "step": 19641
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1705273009330633e-05,
+      "loss": 0.6125,
+      "step": 19642
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1704570614003407e-05,
+      "loss": 0.6052,
+      "step": 19643
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.170386821001476e-05,
+      "loss": 0.6324,
+      "step": 19644
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1703165797368259e-05,
+      "loss": 0.5921,
+      "step": 19645
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1702463376067474e-05,
+      "loss": 0.5658,
+      "step": 19646
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1701760946115974e-05,
+      "loss": 0.5939,
+      "step": 19647
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.170105850751733e-05,
+      "loss": 0.5758,
+      "step": 19648
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1700356060275107e-05,
+      "loss": 0.6037,
+      "step": 19649
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1699653604392876e-05,
+      "loss": 0.5499,
+      "step": 19650
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1698951139874208e-05,
+      "loss": 0.6204,
+      "step": 19651
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1698248666722674e-05,
+      "loss": 0.5887,
+      "step": 19652
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.169754618494184e-05,
+      "loss": 0.6212,
+      "step": 19653
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1696843694535276e-05,
+      "loss": 0.6182,
+      "step": 19654
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1696141195506551e-05,
+      "loss": 0.6224,
+      "step": 19655
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1695438687859238e-05,
+      "loss": 0.5751,
+      "step": 19656
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1694736171596903e-05,
+      "loss": 0.6067,
+      "step": 19657
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1694033646723117e-05,
+      "loss": 0.5885,
+      "step": 19658
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1693331113241453e-05,
+      "loss": 0.6007,
+      "step": 19659
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1692628571155476e-05,
+      "loss": 0.622,
+      "step": 19660
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1691926020468758e-05,
+      "loss": 0.6005,
+      "step": 19661
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1691223461184866e-05,
+      "loss": 0.5777,
+      "step": 19662
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1690520893307376e-05,
+      "loss": 0.5884,
+      "step": 19663
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1689818316839849e-05,
+      "loss": 0.5835,
+      "step": 19664
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1689115731785865e-05,
+      "loss": 0.6045,
+      "step": 19665
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1688413138148989e-05,
+      "loss": 0.6012,
+      "step": 19666
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.168771053593279e-05,
+      "loss": 0.5837,
+      "step": 19667
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1687007925140838e-05,
+      "loss": 0.6061,
+      "step": 19668
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.168630530577671e-05,
+      "loss": 0.6174,
+      "step": 19669
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1685602677843963e-05,
+      "loss": 0.6333,
+      "step": 19670
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.168490004134618e-05,
+      "loss": 0.5892,
+      "step": 19671
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1684197396286927e-05,
+      "loss": 0.6297,
+      "step": 19672
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1683494742669772e-05,
+      "loss": 0.6345,
+      "step": 19673
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1682792080498287e-05,
+      "loss": 0.6299,
+      "step": 19674
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1682089409776043e-05,
+      "loss": 0.6275,
+      "step": 19675
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1681386730506609e-05,
+      "loss": 0.6356,
+      "step": 19676
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1680684042693557e-05,
+      "loss": 0.6003,
+      "step": 19677
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.167998134634046e-05,
+      "loss": 0.6432,
+      "step": 19678
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1679278641450883e-05,
+      "loss": 0.6326,
+      "step": 19679
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1678575928028401e-05,
+      "loss": 0.5845,
+      "step": 19680
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1677873206076582e-05,
+      "loss": 0.5673,
+      "step": 19681
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1677170475598998e-05,
+      "loss": 0.6107,
+      "step": 19682
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1676467736599221e-05,
+      "loss": 0.592,
+      "step": 19683
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.167576498908082e-05,
+      "loss": 0.6639,
+      "step": 19684
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1675062233047365e-05,
+      "loss": 0.5906,
+      "step": 19685
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.167435946850243e-05,
+      "loss": 0.6777,
+      "step": 19686
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1673656695449582e-05,
+      "loss": 0.5973,
+      "step": 19687
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1672953913892398e-05,
+      "loss": 0.6159,
+      "step": 19688
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.167225112383444e-05,
+      "loss": 0.5464,
+      "step": 19689
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1671548325279288e-05,
+      "loss": 0.5395,
+      "step": 19690
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1670845518230508e-05,
+      "loss": 0.6506,
+      "step": 19691
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.167014270269167e-05,
+      "loss": 0.5597,
+      "step": 19692
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.166943987866635e-05,
+      "loss": 0.5947,
+      "step": 19693
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.166873704615812e-05,
+      "loss": 0.635,
+      "step": 19694
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1668034205170547e-05,
+      "loss": 0.5484,
+      "step": 19695
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1667331355707201e-05,
+      "loss": 0.593,
+      "step": 19696
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1666628497771659e-05,
+      "loss": 0.609,
+      "step": 19697
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1665925631367488e-05,
+      "loss": 0.6335,
+      "step": 19698
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1665222756498263e-05,
+      "loss": 0.6254,
+      "step": 19699
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1664519873167551e-05,
+      "loss": 0.5968,
+      "step": 19700
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1663816981378924e-05,
+      "loss": 0.6003,
+      "step": 19701
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.166311408113596e-05,
+      "loss": 0.5862,
+      "step": 19702
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1662411172442227e-05,
+      "loss": 0.5987,
+      "step": 19703
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1661708255301291e-05,
+      "loss": 0.6375,
+      "step": 19704
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1661005329716736e-05,
+      "loss": 0.5627,
+      "step": 19705
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1660302395692119e-05,
+      "loss": 0.5464,
+      "step": 19706
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1659599453231023e-05,
+      "loss": 0.5431,
+      "step": 19707
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1658896502337016e-05,
+      "loss": 0.6044,
+      "step": 19708
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1658193543013666e-05,
+      "loss": 0.6248,
+      "step": 19709
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1657490575264556e-05,
+      "loss": 0.5879,
+      "step": 19710
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1656787599093247e-05,
+      "loss": 0.5927,
+      "step": 19711
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1656084614503314e-05,
+      "loss": 0.6173,
+      "step": 19712
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1655381621498331e-05,
+      "loss": 0.6141,
+      "step": 19713
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.165467862008187e-05,
+      "loss": 0.6228,
+      "step": 19714
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1653975610257501e-05,
+      "loss": 0.5942,
+      "step": 19715
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1653272592028799e-05,
+      "loss": 0.5624,
+      "step": 19716
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1652569565399335e-05,
+      "loss": 0.5484,
+      "step": 19717
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.165186653037268e-05,
+      "loss": 0.5863,
+      "step": 19718
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1651163486952406e-05,
+      "loss": 0.5689,
+      "step": 19719
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1650460435142087e-05,
+      "loss": 0.6308,
+      "step": 19720
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.16497573749453e-05,
+      "loss": 0.6294,
+      "step": 19721
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1649054306365607e-05,
+      "loss": 0.5654,
+      "step": 19722
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.164835122940659e-05,
+      "loss": 0.5894,
+      "step": 19723
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1647648144071817e-05,
+      "loss": 0.5151,
+      "step": 19724
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.164694505036486e-05,
+      "loss": 0.595,
+      "step": 19725
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1646241948289294e-05,
+      "loss": 0.6253,
+      "step": 19726
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1645538837848689e-05,
+      "loss": 0.5566,
+      "step": 19727
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.164483571904662e-05,
+      "loss": 0.6487,
+      "step": 19728
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1644132591886659e-05,
+      "loss": 0.6424,
+      "step": 19729
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1643429456372377e-05,
+      "loss": 0.6329,
+      "step": 19730
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1642726312507352e-05,
+      "loss": 0.6021,
+      "step": 19731
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.164202316029515e-05,
+      "loss": 0.6563,
+      "step": 19732
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1641319999739349e-05,
+      "loss": 0.6402,
+      "step": 19733
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1640616830843522e-05,
+      "loss": 0.6275,
+      "step": 19734
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.163991365361124e-05,
+      "loss": 0.6584,
+      "step": 19735
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1639210468046074e-05,
+      "loss": 0.6288,
+      "step": 19736
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1638507274151603e-05,
+      "loss": 0.5634,
+      "step": 19737
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1637804071931393e-05,
+      "loss": 0.5932,
+      "step": 19738
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1637100861389021e-05,
+      "loss": 0.5673,
+      "step": 19739
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1636397642528063e-05,
+      "loss": 0.5935,
+      "step": 19740
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1635694415352087e-05,
+      "loss": 0.5707,
+      "step": 19741
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1634991179864671e-05,
+      "loss": 0.6752,
+      "step": 19742
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1634287936069386e-05,
+      "loss": 0.5753,
+      "step": 19743
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1633584683969802e-05,
+      "loss": 0.5893,
+      "step": 19744
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.16328814235695e-05,
+      "loss": 0.6011,
+      "step": 19745
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1632178154872046e-05,
+      "loss": 0.5863,
+      "step": 19746
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1631474877881017e-05,
+      "loss": 0.5681,
+      "step": 19747
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1630771592599989e-05,
+      "loss": 0.6552,
+      "step": 19748
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.163006829903253e-05,
+      "loss": 0.6438,
+      "step": 19749
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1629364997182218e-05,
+      "loss": 0.5902,
+      "step": 19750
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1628661687052624e-05,
+      "loss": 0.6624,
+      "step": 19751
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1627958368647326e-05,
+      "loss": 0.6121,
+      "step": 19752
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1627255041969895e-05,
+      "loss": 0.5482,
+      "step": 19753
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1626551707023899e-05,
+      "loss": 0.5703,
+      "step": 19754
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1625848363812919e-05,
+      "loss": 0.6595,
+      "step": 19755
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1625145012340531e-05,
+      "loss": 0.5693,
+      "step": 19756
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1624441652610302e-05,
+      "loss": 0.6431,
+      "step": 19757
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1623738284625809e-05,
+      "loss": 0.5435,
+      "step": 19758
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1623034908390632e-05,
+      "loss": 0.6864,
+      "step": 19759
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1622331523908332e-05,
+      "loss": 0.6254,
+      "step": 19760
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1621628131182497e-05,
+      "loss": 0.6255,
+      "step": 19761
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.162092473021669e-05,
+      "loss": 0.6518,
+      "step": 19762
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1620221321014491e-05,
+      "loss": 0.551,
+      "step": 19763
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1619517903579474e-05,
+      "loss": 0.6098,
+      "step": 19764
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.161881447791521e-05,
+      "loss": 0.59,
+      "step": 19765
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1618111044025276e-05,
+      "loss": 0.5766,
+      "step": 19766
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1617407601913246e-05,
+      "loss": 0.6533,
+      "step": 19767
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1616704151582696e-05,
+      "loss": 0.5939,
+      "step": 19768
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1616000693037194e-05,
+      "loss": 0.6605,
+      "step": 19769
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1615297226280326e-05,
+      "loss": 0.6471,
+      "step": 19770
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1614593751315654e-05,
+      "loss": 0.6031,
+      "step": 19771
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1613890268146763e-05,
+      "loss": 0.582,
+      "step": 19772
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1613186776777218e-05,
+      "loss": 0.6473,
+      "step": 19773
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1612483277210602e-05,
+      "loss": 0.613,
+      "step": 19774
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1611779769450486e-05,
+      "loss": 0.6398,
+      "step": 19775
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1611076253500445e-05,
+      "loss": 0.5479,
+      "step": 19776
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.161037272936405e-05,
+      "loss": 0.5649,
+      "step": 19777
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1609669197044887e-05,
+      "loss": 0.6292,
+      "step": 19778
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1608965656546516e-05,
+      "loss": 0.5923,
+      "step": 19779
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1608262107872524e-05,
+      "loss": 0.6263,
+      "step": 19780
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1607558551026478e-05,
+      "loss": 0.628,
+      "step": 19781
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.160685498601196e-05,
+      "loss": 0.5892,
+      "step": 19782
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1606151412832539e-05,
+      "loss": 0.6202,
+      "step": 19783
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1605447831491791e-05,
+      "loss": 0.5731,
+      "step": 19784
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1604744241993293e-05,
+      "loss": 0.5869,
+      "step": 19785
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1604040644340621e-05,
+      "loss": 0.6099,
+      "step": 19786
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1603337038537349e-05,
+      "loss": 0.6038,
+      "step": 19787
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1602633424587049e-05,
+      "loss": 0.6265,
+      "step": 19788
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1601929802493301e-05,
+      "loss": 0.6365,
+      "step": 19789
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1601226172259679e-05,
+      "loss": 0.6585,
+      "step": 19790
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.160052253388976e-05,
+      "loss": 0.5866,
+      "step": 19791
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1599818887387113e-05,
+      "loss": 0.6059,
+      "step": 19792
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1599115232755318e-05,
+      "loss": 0.6633,
+      "step": 19793
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1598411569997954e-05,
+      "loss": 0.5632,
+      "step": 19794
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1597707899118589e-05,
+      "loss": 0.5984,
+      "step": 19795
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1597004220120802e-05,
+      "loss": 0.6273,
+      "step": 19796
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1596300533008172e-05,
+      "loss": 0.6086,
+      "step": 19797
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.159559683778427e-05,
+      "loss": 0.5449,
+      "step": 19798
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1594893134452674e-05,
+      "loss": 0.6542,
+      "step": 19799
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1594189423016958e-05,
+      "loss": 0.6074,
+      "step": 19800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1593485703480698e-05,
+      "loss": 0.5827,
+      "step": 19801
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.159278197584747e-05,
+      "loss": 0.6176,
+      "step": 19802
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1592078240120852e-05,
+      "loss": 0.6403,
+      "step": 19803
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1591374496304418e-05,
+      "loss": 0.5948,
+      "step": 19804
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1590670744401745e-05,
+      "loss": 0.6092,
+      "step": 19805
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1589966984416405e-05,
+      "loss": 0.6103,
+      "step": 19806
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1589263216351981e-05,
+      "loss": 0.5701,
+      "step": 19807
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1588559440212041e-05,
+      "loss": 0.5933,
+      "step": 19808
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1587855656000167e-05,
+      "loss": 0.565,
+      "step": 19809
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1587151863719936e-05,
+      "loss": 0.6464,
+      "step": 19810
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1586448063374918e-05,
+      "loss": 0.609,
+      "step": 19811
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1585744254968695e-05,
+      "loss": 0.6034,
+      "step": 19812
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.158504043850484e-05,
+      "loss": 0.6347,
+      "step": 19813
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1584336613986928e-05,
+      "loss": 0.5958,
+      "step": 19814
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.158363278141854e-05,
+      "loss": 0.5716,
+      "step": 19815
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1582928940803252e-05,
+      "loss": 0.6429,
+      "step": 19816
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1582225092144637e-05,
+      "loss": 0.6352,
+      "step": 19817
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1581521235446274e-05,
+      "loss": 0.5946,
+      "step": 19818
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1580817370711734e-05,
+      "loss": 0.5944,
+      "step": 19819
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1580113497944603e-05,
+      "loss": 0.6054,
+      "step": 19820
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1579409617148449e-05,
+      "loss": 0.6199,
+      "step": 19821
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1578705728326855e-05,
+      "loss": 0.5666,
+      "step": 19822
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1578001831483394e-05,
+      "loss": 0.5847,
+      "step": 19823
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1577297926621644e-05,
+      "loss": 0.612,
+      "step": 19824
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1576594013745179e-05,
+      "loss": 0.5883,
+      "step": 19825
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.157589009285758e-05,
+      "loss": 0.6279,
+      "step": 19826
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1575186163962421e-05,
+      "loss": 0.6097,
+      "step": 19827
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.157448222706328e-05,
+      "loss": 0.6261,
+      "step": 19828
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1573778282163735e-05,
+      "loss": 0.5653,
+      "step": 19829
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.157307432926736e-05,
+      "loss": 0.5809,
+      "step": 19830
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1572370368377731e-05,
+      "loss": 0.6201,
+      "step": 19831
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1571666399498434e-05,
+      "loss": 0.5485,
+      "step": 19832
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1570962422633036e-05,
+      "loss": 0.55,
+      "step": 19833
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1570258437785118e-05,
+      "loss": 0.6269,
+      "step": 19834
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1569554444958258e-05,
+      "loss": 0.6188,
+      "step": 19835
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.156885044415603e-05,
+      "loss": 0.6363,
+      "step": 19836
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1568146435382017e-05,
+      "loss": 0.5962,
+      "step": 19837
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.156744241863979e-05,
+      "loss": 0.5958,
+      "step": 19838
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1566738393932933e-05,
+      "loss": 0.559,
+      "step": 19839
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1566034361265012e-05,
+      "loss": 0.5769,
+      "step": 19840
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1565330320639619e-05,
+      "loss": 0.5168,
+      "step": 19841
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.156462627206032e-05,
+      "loss": 0.5748,
+      "step": 19842
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1563922215530701e-05,
+      "loss": 0.5259,
+      "step": 19843
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.156321815105433e-05,
+      "loss": 0.6225,
+      "step": 19844
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1562514078634793e-05,
+      "loss": 0.5768,
+      "step": 19845
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1561809998275662e-05,
+      "loss": 0.5897,
+      "step": 19846
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1561105909980516e-05,
+      "loss": 0.6083,
+      "step": 19847
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.156040181375294e-05,
+      "loss": 0.6134,
+      "step": 19848
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1559697709596497e-05,
+      "loss": 0.6086,
+      "step": 19849
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1558993597514778e-05,
+      "loss": 0.5821,
+      "step": 19850
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1558289477511357e-05,
+      "loss": 0.5951,
+      "step": 19851
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1557585349589808e-05,
+      "loss": 0.5533,
+      "step": 19852
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1556881213753715e-05,
+      "loss": 0.5794,
+      "step": 19853
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.155617707000665e-05,
+      "loss": 0.5608,
+      "step": 19854
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1555472918352193e-05,
+      "loss": 0.6388,
+      "step": 19855
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1554768758793924e-05,
+      "loss": 0.6342,
+      "step": 19856
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1554064591335418e-05,
+      "loss": 0.6049,
+      "step": 19857
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1553360415980256e-05,
+      "loss": 0.5852,
+      "step": 19858
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1552656232732016e-05,
+      "loss": 0.6,
+      "step": 19859
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1551952041594274e-05,
+      "loss": 0.5606,
+      "step": 19860
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1551247842570609e-05,
+      "loss": 0.655,
+      "step": 19861
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.15505436356646e-05,
+      "loss": 0.6123,
+      "step": 19862
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1549839420879822e-05,
+      "loss": 0.6611,
+      "step": 19863
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.154913519821986e-05,
+      "loss": 0.5728,
+      "step": 19864
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1548430967688286e-05,
+      "loss": 0.652,
+      "step": 19865
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.154772672928868e-05,
+      "loss": 0.5873,
+      "step": 19866
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1547022483024624e-05,
+      "loss": 0.5745,
+      "step": 19867
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.154631822889969e-05,
+      "loss": 0.5924,
+      "step": 19868
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1545613966917461e-05,
+      "loss": 0.637,
+      "step": 19869
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1544909697081519e-05,
+      "loss": 0.5819,
+      "step": 19870
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1544205419395434e-05,
+      "loss": 0.5968,
+      "step": 19871
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1543501133862791e-05,
+      "loss": 0.601,
+      "step": 19872
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1542796840487168e-05,
+      "loss": 0.5864,
+      "step": 19873
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1542092539272139e-05,
+      "loss": 0.6255,
+      "step": 19874
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1541388230221288e-05,
+      "loss": 0.5895,
+      "step": 19875
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1540683913338191e-05,
+      "loss": 0.5848,
+      "step": 19876
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1539979588626426e-05,
+      "loss": 0.5661,
+      "step": 19877
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.153927525608958e-05,
+      "loss": 0.5592,
+      "step": 19878
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.153857091573122e-05,
+      "loss": 0.6231,
+      "step": 19879
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1537866567554933e-05,
+      "loss": 0.6608,
+      "step": 19880
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1537162211564296e-05,
+      "loss": 0.6236,
+      "step": 19881
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1536457847762888e-05,
+      "loss": 0.6021,
+      "step": 19882
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1535753476154286e-05,
+      "loss": 0.6287,
+      "step": 19883
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.153504909674207e-05,
+      "loss": 0.6747,
+      "step": 19884
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1534344709529821e-05,
+      "loss": 0.5578,
+      "step": 19885
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.153364031452112e-05,
+      "loss": 0.6405,
+      "step": 19886
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1532935911719539e-05,
+      "loss": 0.6454,
+      "step": 19887
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1532231501128662e-05,
+      "loss": 0.5604,
+      "step": 19888
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.153152708275207e-05,
+      "loss": 0.631,
+      "step": 19889
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1530822656593339e-05,
+      "loss": 0.6197,
+      "step": 19890
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1530118222656053e-05,
+      "loss": 0.5839,
+      "step": 19891
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1529413780943783e-05,
+      "loss": 0.6064,
+      "step": 19892
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1528709331460118e-05,
+      "loss": 0.6078,
+      "step": 19893
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1528004874208632e-05,
+      "loss": 0.6404,
+      "step": 19894
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1527300409192903e-05,
+      "loss": 0.6561,
+      "step": 19895
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1526595936416516e-05,
+      "loss": 0.6029,
+      "step": 19896
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1525891455883049e-05,
+      "loss": 0.578,
+      "step": 19897
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1525186967596077e-05,
+      "loss": 0.5606,
+      "step": 19898
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1524482471559188e-05,
+      "loss": 0.6171,
+      "step": 19899
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1523777967775956e-05,
+      "loss": 0.5769,
+      "step": 19900
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.152307345624996e-05,
+      "loss": 0.6211,
+      "step": 19901
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1522368936984784e-05,
+      "loss": 0.5774,
+      "step": 19902
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1521664409984004e-05,
+      "loss": 0.5891,
+      "step": 19903
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.15209598752512e-05,
+      "loss": 0.6107,
+      "step": 19904
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1520255332789955e-05,
+      "loss": 0.5556,
+      "step": 19905
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1519550782603846e-05,
+      "loss": 0.5572,
+      "step": 19906
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1518846224696456e-05,
+      "loss": 0.608,
+      "step": 19907
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1518141659071365e-05,
+      "loss": 0.5912,
+      "step": 19908
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1517437085732149e-05,
+      "loss": 0.6194,
+      "step": 19909
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1516732504682393e-05,
+      "loss": 0.623,
+      "step": 19910
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1516027915925672e-05,
+      "loss": 0.5667,
+      "step": 19911
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.151532331946557e-05,
+      "loss": 0.5368,
+      "step": 19912
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1514618715305667e-05,
+      "loss": 0.5823,
+      "step": 19913
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1513914103449543e-05,
+      "loss": 0.6309,
+      "step": 19914
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1513209483900778e-05,
+      "loss": 0.5456,
+      "step": 19915
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1512504856662954e-05,
+      "loss": 0.5511,
+      "step": 19916
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1511800221739646e-05,
+      "loss": 0.6594,
+      "step": 19917
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1511095579134441e-05,
+      "loss": 0.6494,
+      "step": 19918
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1510390928850916e-05,
+      "loss": 0.6592,
+      "step": 19919
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1509686270892655e-05,
+      "loss": 0.6473,
+      "step": 19920
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1508981605263234e-05,
+      "loss": 0.6071,
+      "step": 19921
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1508276931966236e-05,
+      "loss": 0.585,
+      "step": 19922
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1507572251005239e-05,
+      "loss": 0.6158,
+      "step": 19923
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1506867562383828e-05,
+      "loss": 0.6313,
+      "step": 19924
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.150616286610558e-05,
+      "loss": 0.6036,
+      "step": 19925
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1505458162174077e-05,
+      "loss": 0.5711,
+      "step": 19926
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1504753450592903e-05,
+      "loss": 0.597,
+      "step": 19927
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1504048731365633e-05,
+      "loss": 0.6034,
+      "step": 19928
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1503344004495855e-05,
+      "loss": 0.6226,
+      "step": 19929
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1502639269987141e-05,
+      "loss": 0.5952,
+      "step": 19930
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.150193452784308e-05,
+      "loss": 0.5627,
+      "step": 19931
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.150122977806725e-05,
+      "loss": 0.6476,
+      "step": 19932
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.150052502066323e-05,
+      "loss": 0.6145,
+      "step": 19933
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1499820255634602e-05,
+      "loss": 0.6257,
+      "step": 19934
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.149911548298495e-05,
+      "loss": 0.6051,
+      "step": 19935
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1498410702717853e-05,
+      "loss": 0.6145,
+      "step": 19936
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1497705914836895e-05,
+      "loss": 0.5856,
+      "step": 19937
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.149700111934565e-05,
+      "loss": 0.6073,
+      "step": 19938
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1496296316247709e-05,
+      "loss": 0.5895,
+      "step": 19939
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1495591505546643e-05,
+      "loss": 0.5754,
+      "step": 19940
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.149488668724604e-05,
+      "loss": 0.5816,
+      "step": 19941
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.149418186134948e-05,
+      "loss": 0.5929,
+      "step": 19942
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1493477027860546e-05,
+      "loss": 0.6104,
+      "step": 19943
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1492772186782816e-05,
+      "loss": 0.6273,
+      "step": 19944
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1492067338119877e-05,
+      "loss": 0.5887,
+      "step": 19945
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1491362481875301e-05,
+      "loss": 0.5983,
+      "step": 19946
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1490657618052678e-05,
+      "loss": 0.6198,
+      "step": 19947
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1489952746655588e-05,
+      "loss": 0.5493,
+      "step": 19948
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1489247867687612e-05,
+      "loss": 0.6204,
+      "step": 19949
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1488542981152329e-05,
+      "loss": 0.6524,
+      "step": 19950
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1487838087053328e-05,
+      "loss": 0.6323,
+      "step": 19951
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.148713318539418e-05,
+      "loss": 0.596,
+      "step": 19952
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1486428276178474e-05,
+      "loss": 0.6187,
+      "step": 19953
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1485723359409795e-05,
+      "loss": 0.6358,
+      "step": 19954
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1485018435091714e-05,
+      "loss": 0.5714,
+      "step": 19955
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1484313503227827e-05,
+      "loss": 0.6432,
+      "step": 19956
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1483608563821703e-05,
+      "loss": 0.6216,
+      "step": 19957
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1482903616876933e-05,
+      "loss": 0.6791,
+      "step": 19958
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1482198662397091e-05,
+      "loss": 0.6464,
+      "step": 19959
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1481493700385768e-05,
+      "loss": 0.5936,
+      "step": 19960
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1480788730846536e-05,
+      "loss": 0.5871,
+      "step": 19961
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1480083753782989e-05,
+      "loss": 0.5811,
+      "step": 19962
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1479378769198699e-05,
+      "loss": 0.5207,
+      "step": 19963
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1478673777097255e-05,
+      "loss": 0.5728,
+      "step": 19964
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1477968777482235e-05,
+      "loss": 0.6095,
+      "step": 19965
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.147726377035722e-05,
+      "loss": 0.625,
+      "step": 19966
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1476558755725799e-05,
+      "loss": 0.6149,
+      "step": 19967
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.147585373359155e-05,
+      "loss": 0.5383,
+      "step": 19968
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1475148703958055e-05,
+      "loss": 0.5638,
+      "step": 19969
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1474443666828898e-05,
+      "loss": 0.5453,
+      "step": 19970
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1473738622207659e-05,
+      "loss": 0.6599,
+      "step": 19971
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1473033570097924e-05,
+      "loss": 0.6481,
+      "step": 19972
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1472328510503276e-05,
+      "loss": 0.6319,
+      "step": 19973
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1471623443427293e-05,
+      "loss": 0.5886,
+      "step": 19974
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1470918368873563e-05,
+      "loss": 0.5949,
+      "step": 19975
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1470213286845662e-05,
+      "loss": 0.6255,
+      "step": 19976
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1469508197347182e-05,
+      "loss": 0.6042,
+      "step": 19977
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1468803100381696e-05,
+      "loss": 0.6341,
+      "step": 19978
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1468097995952791e-05,
+      "loss": 0.5743,
+      "step": 19979
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1467392884064053e-05,
+      "loss": 0.608,
+      "step": 19980
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1466687764719062e-05,
+      "loss": 0.5263,
+      "step": 19981
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1465982637921398e-05,
+      "loss": 0.5433,
+      "step": 19982
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.146527750367465e-05,
+      "loss": 0.6073,
+      "step": 19983
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1464572361982395e-05,
+      "loss": 0.6436,
+      "step": 19984
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1463867212848221e-05,
+      "loss": 0.6104,
+      "step": 19985
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.146316205627571e-05,
+      "loss": 0.5555,
+      "step": 19986
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.146245689226844e-05,
+      "loss": 0.6197,
+      "step": 19987
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1461751720829999e-05,
+      "loss": 0.6468,
+      "step": 19988
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1461046541963974e-05,
+      "loss": 0.5641,
+      "step": 19989
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1460341355673939e-05,
+      "loss": 0.6243,
+      "step": 19990
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1459636161963485e-05,
+      "loss": 0.5624,
+      "step": 19991
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.145893096083619e-05,
+      "loss": 0.5775,
+      "step": 19992
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1458225752295639e-05,
+      "loss": 0.6039,
+      "step": 19993
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1457520536345419e-05,
+      "loss": 0.5894,
+      "step": 19994
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1456815312989107e-05,
+      "loss": 0.5844,
+      "step": 19995
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1456110082230292e-05,
+      "loss": 0.5734,
+      "step": 19996
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1455404844072553e-05,
+      "loss": 0.5476,
+      "step": 19997
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1454699598519479e-05,
+      "loss": 0.6659,
+      "step": 19998
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1453994345574648e-05,
+      "loss": 0.5956,
+      "step": 19999
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1453289085241648e-05,
+      "loss": 0.5946,
+      "step": 20000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1452583817524059e-05,
+      "loss": 0.6027,
+      "step": 20001
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1451878542425465e-05,
+      "loss": 0.5681,
+      "step": 20002
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1451173259949452e-05,
+      "loss": 0.509,
+      "step": 20003
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1450467970099602e-05,
+      "loss": 0.5899,
+      "step": 20004
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1449762672879502e-05,
+      "loss": 0.6339,
+      "step": 20005
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.144905736829273e-05,
+      "loss": 0.6355,
+      "step": 20006
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1448352056342876e-05,
+      "loss": 0.601,
+      "step": 20007
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.144764673703352e-05,
+      "loss": 0.6273,
+      "step": 20008
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1446941410368245e-05,
+      "loss": 0.5316,
+      "step": 20009
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.144623607635064e-05,
+      "loss": 0.572,
+      "step": 20010
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1445530734984283e-05,
+      "loss": 0.5878,
+      "step": 20011
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1444825386272763e-05,
+      "loss": 0.5583,
+      "step": 20012
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1444120030219661e-05,
+      "loss": 0.5672,
+      "step": 20013
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1443414666828563e-05,
+      "loss": 0.5536,
+      "step": 20014
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.144270929610305e-05,
+      "loss": 0.585,
+      "step": 20015
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1442003918046714e-05,
+      "loss": 0.6231,
+      "step": 20016
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.144129853266313e-05,
+      "loss": 0.6052,
+      "step": 20017
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1440593139955887e-05,
+      "loss": 0.6555,
+      "step": 20018
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1439887739928568e-05,
+      "loss": 0.6118,
+      "step": 20019
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1439182332584757e-05,
+      "loss": 0.5856,
+      "step": 20020
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1438476917928041e-05,
+      "loss": 0.647,
+      "step": 20021
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1437771495962e-05,
+      "loss": 0.6471,
+      "step": 20022
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.143706606669022e-05,
+      "loss": 0.5764,
+      "step": 20023
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1436360630116292e-05,
+      "loss": 0.6563,
+      "step": 20024
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.143565518624379e-05,
+      "loss": 0.5859,
+      "step": 20025
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1434949735076303e-05,
+      "loss": 0.643,
+      "step": 20026
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1434244276617422e-05,
+      "loss": 0.6003,
+      "step": 20027
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.143353881087072e-05,
+      "loss": 0.5645,
+      "step": 20028
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1432833337839791e-05,
+      "loss": 0.5944,
+      "step": 20029
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1432127857528213e-05,
+      "loss": 0.6043,
+      "step": 20030
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1431422369939575e-05,
+      "loss": 0.6371,
+      "step": 20031
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1430716875077464e-05,
+      "loss": 0.575,
+      "step": 20032
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1430011372945455e-05,
+      "loss": 0.6253,
+      "step": 20033
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1429305863547145e-05,
+      "loss": 0.5325,
+      "step": 20034
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1428600346886112e-05,
+      "loss": 0.5988,
+      "step": 20035
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1427894822965944e-05,
+      "loss": 0.6331,
+      "step": 20036
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1427189291790219e-05,
+      "loss": 0.5882,
+      "step": 20037
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1426483753362531e-05,
+      "loss": 0.5755,
+      "step": 20038
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1425778207686463e-05,
+      "loss": 0.6149,
+      "step": 20039
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1425072654765596e-05,
+      "loss": 0.5906,
+      "step": 20040
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1424367094603517e-05,
+      "loss": 0.6055,
+      "step": 20041
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1423661527203812e-05,
+      "loss": 0.6284,
+      "step": 20042
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1422955952570065e-05,
+      "loss": 0.5769,
+      "step": 20043
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1422250370705862e-05,
+      "loss": 0.6028,
+      "step": 20044
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1421544781614791e-05,
+      "loss": 0.6163,
+      "step": 20045
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1420839185300432e-05,
+      "loss": 0.5859,
+      "step": 20046
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1420133581766373e-05,
+      "loss": 0.5826,
+      "step": 20047
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1419427971016202e-05,
+      "loss": 0.582,
+      "step": 20048
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.14187223530535e-05,
+      "loss": 0.6068,
+      "step": 20049
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1418016727881853e-05,
+      "loss": 0.5856,
+      "step": 20050
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.141731109550485e-05,
+      "loss": 0.6756,
+      "step": 20051
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1416605455926072e-05,
+      "loss": 0.5879,
+      "step": 20052
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.141589980914911e-05,
+      "loss": 0.5519,
+      "step": 20053
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1415194155177545e-05,
+      "loss": 0.5622,
+      "step": 20054
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1414488494014961e-05,
+      "loss": 0.5931,
+      "step": 20055
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1413782825664952e-05,
+      "loss": 0.587,
+      "step": 20056
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.1413077150131097e-05,
+      "loss": 0.5718,
+      "step": 20057
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.141237146741698e-05,
+      "loss": 0.6107,
+      "step": 20058
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1411665777526193e-05,
+      "loss": 0.6171,
+      "step": 20059
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.141096008046232e-05,
+      "loss": 0.5609,
+      "step": 20060
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1410254376228939e-05,
+      "loss": 0.618,
+      "step": 20061
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1409548664829648e-05,
+      "loss": 0.5828,
+      "step": 20062
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1408842946268025e-05,
+      "loss": 0.6163,
+      "step": 20063
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1408137220547658e-05,
+      "loss": 0.5978,
+      "step": 20064
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1407431487672138e-05,
+      "loss": 0.5879,
+      "step": 20065
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1406725747645043e-05,
+      "loss": 0.6093,
+      "step": 20066
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1406020000469965e-05,
+      "loss": 0.564,
+      "step": 20067
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1405314246150482e-05,
+      "loss": 0.6966,
+      "step": 20068
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1404608484690189e-05,
+      "loss": 0.681,
+      "step": 20069
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1403902716092672e-05,
+      "loss": 0.5283,
+      "step": 20070
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.140319694036151e-05,
+      "loss": 0.614,
+      "step": 20071
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1402491157500293e-05,
+      "loss": 0.6712,
+      "step": 20072
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.140178536751261e-05,
+      "loss": 0.6142,
+      "step": 20073
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1401079570402045e-05,
+      "loss": 0.6827,
+      "step": 20074
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1400373766172183e-05,
+      "loss": 0.5929,
+      "step": 20075
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1399667954826614e-05,
+      "loss": 0.5299,
+      "step": 20076
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1398962136368921e-05,
+      "loss": 0.659,
+      "step": 20077
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1398256310802694e-05,
+      "loss": 0.6003,
+      "step": 20078
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1397550478131514e-05,
+      "loss": 0.5907,
+      "step": 20079
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1396844638358969e-05,
+      "loss": 0.5887,
+      "step": 20080
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1396138791488651e-05,
+      "loss": 0.677,
+      "step": 20081
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.139543293752414e-05,
+      "loss": 0.6006,
+      "step": 20082
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.139472707646903e-05,
+      "loss": 0.6268,
+      "step": 20083
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1394021208326898e-05,
+      "loss": 0.5598,
+      "step": 20084
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1393315333101338e-05,
+      "loss": 0.6084,
+      "step": 20085
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1392609450795937e-05,
+      "loss": 0.599,
+      "step": 20086
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1391903561414278e-05,
+      "loss": 0.5979,
+      "step": 20087
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1391197664959948e-05,
+      "loss": 0.6152,
+      "step": 20088
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1390491761436535e-05,
+      "loss": 0.6013,
+      "step": 20089
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1389785850847627e-05,
+      "loss": 0.6308,
+      "step": 20090
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.138907993319681e-05,
+      "loss": 0.5427,
+      "step": 20091
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1388374008487673e-05,
+      "loss": 0.6363,
+      "step": 20092
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1387668076723799e-05,
+      "loss": 0.6006,
+      "step": 20093
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1386962137908779e-05,
+      "loss": 0.598,
+      "step": 20094
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1386256192046194e-05,
+      "loss": 0.5922,
+      "step": 20095
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.138555023913964e-05,
+      "loss": 0.6109,
+      "step": 20096
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.13848442791927e-05,
+      "loss": 0.6052,
+      "step": 20097
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1384138312208957e-05,
+      "loss": 0.5605,
+      "step": 20098
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1383432338192002e-05,
+      "loss": 0.5376,
+      "step": 20099
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1382726357145426e-05,
+      "loss": 0.5668,
+      "step": 20100
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1382020369072808e-05,
+      "loss": 0.5957,
+      "step": 20101
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1381314373977744e-05,
+      "loss": 0.6301,
+      "step": 20102
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1380608371863813e-05,
+      "loss": 0.6387,
+      "step": 20103
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1379902362734607e-05,
+      "loss": 0.6562,
+      "step": 20104
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1379196346593715e-05,
+      "loss": 0.6375,
+      "step": 20105
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1378490323444722e-05,
+      "loss": 0.5786,
+      "step": 20106
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1377784293291214e-05,
+      "loss": 0.564,
+      "step": 20107
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1377078256136783e-05,
+      "loss": 0.5896,
+      "step": 20108
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1376372211985011e-05,
+      "loss": 0.5869,
+      "step": 20109
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1375666160839492e-05,
+      "loss": 0.5897,
+      "step": 20110
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.137496010270381e-05,
+      "loss": 0.6129,
+      "step": 20111
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.137425403758155e-05,
+      "loss": 0.6569,
+      "step": 20112
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1373547965476308e-05,
+      "loss": 0.6248,
+      "step": 20113
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.137284188639166e-05,
+      "loss": 0.5875,
+      "step": 20114
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1372135800331208e-05,
+      "loss": 0.6217,
+      "step": 20115
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1371429707298527e-05,
+      "loss": 0.6108,
+      "step": 20116
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1370723607297211e-05,
+      "loss": 0.6292,
+      "step": 20117
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1370017500330845e-05,
+      "loss": 0.6142,
+      "step": 20118
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1369311386403024e-05,
+      "loss": 0.5791,
+      "step": 20119
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1368605265517324e-05,
+      "loss": 0.626,
+      "step": 20120
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1367899137677347e-05,
+      "loss": 0.6499,
+      "step": 20121
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1367193002886669e-05,
+      "loss": 0.6139,
+      "step": 20122
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1366486861148884e-05,
+      "loss": 0.6289,
+      "step": 20123
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.136578071246758e-05,
+      "loss": 0.6309,
+      "step": 20124
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1365074556846342e-05,
+      "loss": 0.625,
+      "step": 20125
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.136436839428876e-05,
+      "loss": 0.5398,
+      "step": 20126
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1363662224798428e-05,
+      "loss": 0.6003,
+      "step": 20127
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1362956048378923e-05,
+      "loss": 0.6014,
+      "step": 20128
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1362249865033843e-05,
+      "loss": 0.5791,
+      "step": 20129
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.136154367476677e-05,
+      "loss": 0.5768,
+      "step": 20130
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1360837477581295e-05,
+      "loss": 0.6225,
+      "step": 20131
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.136013127348101e-05,
+      "loss": 0.6206,
+      "step": 20132
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1359425062469496e-05,
+      "loss": 0.6685,
+      "step": 20133
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1358718844550347e-05,
+      "loss": 0.6161,
+      "step": 20134
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1358012619727148e-05,
+      "loss": 0.6379,
+      "step": 20135
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1357306388003491e-05,
+      "loss": 0.6402,
+      "step": 20136
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.135660014938296e-05,
+      "loss": 0.5792,
+      "step": 20137
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1355893903869148e-05,
+      "loss": 0.5942,
+      "step": 20138
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1355187651465642e-05,
+      "loss": 0.5787,
+      "step": 20139
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1354481392176032e-05,
+      "loss": 0.6112,
+      "step": 20140
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1353775126003903e-05,
+      "loss": 0.5797,
+      "step": 20141
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1353068852952844e-05,
+      "loss": 0.6329,
+      "step": 20142
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1352362573026452e-05,
+      "loss": 0.6206,
+      "step": 20143
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1351656286228306e-05,
+      "loss": 0.5578,
+      "step": 20144
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1350949992561997e-05,
+      "loss": 0.6432,
+      "step": 20145
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1350243692031118e-05,
+      "loss": 0.5939,
+      "step": 20146
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1349537384639255e-05,
+      "loss": 0.5886,
+      "step": 20147
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1348831070389998e-05,
+      "loss": 0.6101,
+      "step": 20148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1348124749286936e-05,
+      "loss": 0.5323,
+      "step": 20149
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1347418421333653e-05,
+      "loss": 0.604,
+      "step": 20150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1346712086533747e-05,
+      "loss": 0.5815,
+      "step": 20151
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.13460057448908e-05,
+      "loss": 0.6069,
+      "step": 20152
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1345299396408404e-05,
+      "loss": 0.6139,
+      "step": 20153
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1344593041090148e-05,
+      "loss": 0.542,
+      "step": 20154
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1343886678939621e-05,
+      "loss": 0.6519,
+      "step": 20155
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1343180309960413e-05,
+      "loss": 0.6148,
+      "step": 20156
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1342473934156113e-05,
+      "loss": 0.548,
+      "step": 20157
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1341767551530308e-05,
+      "loss": 0.5776,
+      "step": 20158
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.134106116208659e-05,
+      "loss": 0.5808,
+      "step": 20159
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1340354765828547e-05,
+      "loss": 0.62,
+      "step": 20160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1339648362759768e-05,
+      "loss": 0.6202,
+      "step": 20161
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1338941952883846e-05,
+      "loss": 0.6496,
+      "step": 20162
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1338235536204365e-05,
+      "loss": 0.571,
+      "step": 20163
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1337529112724917e-05,
+      "loss": 0.615,
+      "step": 20164
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1336822682449096e-05,
+      "loss": 0.5899,
+      "step": 20165
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1336116245380481e-05,
+      "loss": 0.607,
+      "step": 20166
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1335409801522675e-05,
+      "loss": 0.557,
+      "step": 20167
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1334703350879256e-05,
+      "loss": 0.6169,
+      "step": 20168
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.133399689345382e-05,
+      "loss": 0.5341,
+      "step": 20169
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1333290429249956e-05,
+      "loss": 0.5986,
+      "step": 20170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.133258395827125e-05,
+      "loss": 0.5988,
+      "step": 20171
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1331877480521299e-05,
+      "loss": 0.5494,
+      "step": 20172
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1331170996003685e-05,
+      "loss": 0.6247,
+      "step": 20173
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1330464504722003e-05,
+      "loss": 0.5974,
+      "step": 20174
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.132975800667984e-05,
+      "loss": 0.572,
+      "step": 20175
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.132905150188079e-05,
+      "loss": 0.6041,
+      "step": 20176
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1328344990328437e-05,
+      "loss": 0.6616,
+      "step": 20177
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1327638472026376e-05,
+      "loss": 0.6743,
+      "step": 20178
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1326931946978193e-05,
+      "loss": 0.6274,
+      "step": 20179
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.132622541518748e-05,
+      "loss": 0.6147,
+      "step": 20180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1325518876657831e-05,
+      "loss": 0.6036,
+      "step": 20181
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1324812331392826e-05,
+      "loss": 0.6165,
+      "step": 20182
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1324105779396066e-05,
+      "loss": 0.579,
+      "step": 20183
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1323399220671136e-05,
+      "loss": 0.5881,
+      "step": 20184
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1322692655221627e-05,
+      "loss": 0.5684,
+      "step": 20185
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1321986083051131e-05,
+      "loss": 0.5999,
+      "step": 20186
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1321279504163235e-05,
+      "loss": 0.6062,
+      "step": 20187
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1320572918561527e-05,
+      "loss": 0.5981,
+      "step": 20188
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1319866326249609e-05,
+      "loss": 0.5919,
+      "step": 20189
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1319159727231058e-05,
+      "loss": 0.5958,
+      "step": 20190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1318453121509471e-05,
+      "loss": 0.5933,
+      "step": 20191
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1317746509088439e-05,
+      "loss": 0.5723,
+      "step": 20192
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1317039889971553e-05,
+      "loss": 0.5616,
+      "step": 20193
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1316333264162397e-05,
+      "loss": 0.6473,
+      "step": 20194
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.131562663166457e-05,
+      "loss": 0.6472,
+      "step": 20195
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1314919992481656e-05,
+      "loss": 0.595,
+      "step": 20196
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.131421334661725e-05,
+      "loss": 0.598,
+      "step": 20197
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.131350669407494e-05,
+      "loss": 0.6019,
+      "step": 20198
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1312800034858317e-05,
+      "loss": 0.6554,
+      "step": 20199
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1312093368970976e-05,
+      "loss": 0.6072,
+      "step": 20200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1311386696416501e-05,
+      "loss": 0.5759,
+      "step": 20201
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1310680017198486e-05,
+      "loss": 0.5605,
+      "step": 20202
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1309973331320525e-05,
+      "loss": 0.6324,
+      "step": 20203
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.1309266638786203e-05,
+      "loss": 0.6284,
+      "step": 20204
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1308559939599117e-05,
+      "loss": 0.6277,
+      "step": 20205
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.130785323376285e-05,
+      "loss": 0.5605,
+      "step": 20206
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1307146521281001e-05,
+      "loss": 0.6848,
+      "step": 20207
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.130643980215716e-05,
+      "loss": 0.6045,
+      "step": 20208
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1305733076394909e-05,
+      "loss": 0.5774,
+      "step": 20209
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.130502634399785e-05,
+      "loss": 0.6222,
+      "step": 20210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1304319604969569e-05,
+      "loss": 0.6132,
+      "step": 20211
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.130361285931366e-05,
+      "loss": 0.5989,
+      "step": 20212
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.130290610703371e-05,
+      "loss": 0.5866,
+      "step": 20213
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1302199348133315e-05,
+      "loss": 0.5764,
+      "step": 20214
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1301492582616063e-05,
+      "loss": 0.6331,
+      "step": 20215
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1300785810485547e-05,
+      "loss": 0.6056,
+      "step": 20216
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1300079031745355e-05,
+      "loss": 0.5868,
+      "step": 20217
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1299372246399078e-05,
+      "loss": 0.623,
+      "step": 20218
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1298665454450315e-05,
+      "loss": 0.5816,
+      "step": 20219
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.129795865590265e-05,
+      "loss": 0.6163,
+      "step": 20220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.129725185075968e-05,
+      "loss": 0.5482,
+      "step": 20221
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.129654503902499e-05,
+      "loss": 0.6153,
+      "step": 20222
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1295838220702175e-05,
+      "loss": 0.6225,
+      "step": 20223
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.129513139579483e-05,
+      "loss": 0.6643,
+      "step": 20224
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1294424564306538e-05,
+      "loss": 0.5795,
+      "step": 20225
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1293717726240899e-05,
+      "loss": 0.6205,
+      "step": 20226
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1293010881601503e-05,
+      "loss": 0.6378,
+      "step": 20227
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1292304030391936e-05,
+      "loss": 0.5898,
+      "step": 20228
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1291597172615796e-05,
+      "loss": 0.5591,
+      "step": 20229
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1290890308276672e-05,
+      "loss": 0.6553,
+      "step": 20230
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1290183437378155e-05,
+      "loss": 0.6666,
+      "step": 20231
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1289476559923842e-05,
+      "loss": 0.5872,
+      "step": 20232
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.128876967591732e-05,
+      "loss": 0.6339,
+      "step": 20233
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.128806278536218e-05,
+      "loss": 0.5634,
+      "step": 20234
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1287355888262015e-05,
+      "loss": 0.6021,
+      "step": 20235
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1286648984620418e-05,
+      "loss": 0.6036,
+      "step": 20236
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.128594207444098e-05,
+      "loss": 0.5692,
+      "step": 20237
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1285235157727297e-05,
+      "loss": 0.6147,
+      "step": 20238
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1284528234482957e-05,
+      "loss": 0.6043,
+      "step": 20239
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.128382130471155e-05,
+      "loss": 0.5896,
+      "step": 20240
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1283114368416674e-05,
+      "loss": 0.5947,
+      "step": 20241
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1282407425601916e-05,
+      "loss": 0.6564,
+      "step": 20242
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1281700476270872e-05,
+      "loss": 0.5824,
+      "step": 20243
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1280993520427131e-05,
+      "loss": 0.6351,
+      "step": 20244
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1280286558074287e-05,
+      "loss": 0.559,
+      "step": 20245
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1279579589215932e-05,
+      "loss": 0.5805,
+      "step": 20246
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1278872613855661e-05,
+      "loss": 0.6028,
+      "step": 20247
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1278165631997063e-05,
+      "loss": 0.6225,
+      "step": 20248
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1277458643643726e-05,
+      "loss": 0.5863,
+      "step": 20249
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1276751648799251e-05,
+      "loss": 0.6659,
+      "step": 20250
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1276044647467228e-05,
+      "loss": 0.5417,
+      "step": 20251
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1275337639651246e-05,
+      "loss": 0.5587,
+      "step": 20252
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1274630625354903e-05,
+      "loss": 0.5946,
+      "step": 20253
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1273923604581789e-05,
+      "loss": 0.5877,
+      "step": 20254
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1273216577335493e-05,
+      "loss": 0.6038,
+      "step": 20255
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1272509543619612e-05,
+      "loss": 0.585,
+      "step": 20256
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1271802503437739e-05,
+      "loss": 0.5749,
+      "step": 20257
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1271095456793461e-05,
+      "loss": 0.5999,
+      "step": 20258
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1270388403690377e-05,
+      "loss": 0.589,
+      "step": 20259
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1269681344132078e-05,
+      "loss": 0.6208,
+      "step": 20260
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1268974278122156e-05,
+      "loss": 0.5646,
+      "step": 20261
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1268267205664202e-05,
+      "loss": 0.558,
+      "step": 20262
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1267560126761814e-05,
+      "loss": 0.6493,
+      "step": 20263
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.126685304141858e-05,
+      "loss": 0.6053,
+      "step": 20264
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1266145949638094e-05,
+      "loss": 0.619,
+      "step": 20265
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.126543885142395e-05,
+      "loss": 0.7285,
+      "step": 20266
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1264731746779743e-05,
+      "loss": 0.5716,
+      "step": 20267
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.126402463570906e-05,
+      "loss": 0.6216,
+      "step": 20268
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.12633175182155e-05,
+      "loss": 0.6422,
+      "step": 20269
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1262610394302653e-05,
+      "loss": 0.5232,
+      "step": 20270
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1261903263974113e-05,
+      "loss": 0.5242,
+      "step": 20271
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1261196127233474e-05,
+      "loss": 0.5785,
+      "step": 20272
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.126048898408433e-05,
+      "loss": 0.6017,
+      "step": 20273
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1259781834530268e-05,
+      "loss": 0.5957,
+      "step": 20274
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1259074678574887e-05,
+      "loss": 0.5898,
+      "step": 20275
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.125836751622178e-05,
+      "loss": 0.5725,
+      "step": 20276
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1257660347474534e-05,
+      "loss": 0.5425,
+      "step": 20277
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1256953172336753e-05,
+      "loss": 0.7086,
+      "step": 20278
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1256245990812022e-05,
+      "loss": 0.544,
+      "step": 20279
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1255538802903938e-05,
+      "loss": 0.5839,
+      "step": 20280
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1254831608616098e-05,
+      "loss": 0.6225,
+      "step": 20281
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1254124407952086e-05,
+      "loss": 0.5985,
+      "step": 20282
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.12534172009155e-05,
+      "loss": 0.5627,
+      "step": 20283
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1252709987509936e-05,
+      "loss": 0.69,
+      "step": 20284
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1252002767738985e-05,
+      "loss": 0.5902,
+      "step": 20285
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1251295541606244e-05,
+      "loss": 0.6207,
+      "step": 20286
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.12505883091153e-05,
+      "loss": 0.5653,
+      "step": 20287
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.124988107026975e-05,
+      "loss": 0.6055,
+      "step": 20288
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1249173825073191e-05,
+      "loss": 0.6217,
+      "step": 20289
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1248466573529215e-05,
+      "loss": 0.6124,
+      "step": 20290
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1247759315641414e-05,
+      "loss": 0.6575,
+      "step": 20291
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.124705205141338e-05,
+      "loss": 0.6337,
+      "step": 20292
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1246344780848713e-05,
+      "loss": 0.6242,
+      "step": 20293
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1245637503951001e-05,
+      "loss": 0.6148,
+      "step": 20294
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.124493022072384e-05,
+      "loss": 0.5919,
+      "step": 20295
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1244222931170822e-05,
+      "loss": 0.6607,
+      "step": 20296
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1243515635295549e-05,
+      "loss": 0.5872,
+      "step": 20297
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1242808333101602e-05,
+      "loss": 0.5556,
+      "step": 20298
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1242101024592585e-05,
+      "loss": 0.5526,
+      "step": 20299
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.124139370977209e-05,
+      "loss": 0.5765,
+      "step": 20300
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1240686388643706e-05,
+      "loss": 0.5761,
+      "step": 20301
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1239979061211034e-05,
+      "loss": 0.5534,
+      "step": 20302
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1239271727477667e-05,
+      "loss": 0.6189,
+      "step": 20303
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1238564387447194e-05,
+      "loss": 0.627,
+      "step": 20304
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1237857041123214e-05,
+      "loss": 0.5422,
+      "step": 20305
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1237149688509322e-05,
+      "loss": 0.6378,
+      "step": 20306
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1236442329609105e-05,
+      "loss": 0.56,
+      "step": 20307
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1235734964426169e-05,
+      "loss": 0.5881,
+      "step": 20308
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1235027592964096e-05,
+      "loss": 0.5894,
+      "step": 20309
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.123432021522649e-05,
+      "loss": 0.6259,
+      "step": 20310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1233612831216938e-05,
+      "loss": 0.61,
+      "step": 20311
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1232905440939041e-05,
+      "loss": 0.5931,
+      "step": 20312
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.123219804439639e-05,
+      "loss": 0.5979,
+      "step": 20313
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.123149064159258e-05,
+      "loss": 0.5561,
+      "step": 20314
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1230783232531201e-05,
+      "loss": 0.6409,
+      "step": 20315
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1230075817215857e-05,
+      "loss": 0.6658,
+      "step": 20316
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1229368395650135e-05,
+      "loss": 0.6029,
+      "step": 20317
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1228660967837632e-05,
+      "loss": 0.6078,
+      "step": 20318
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1227953533781945e-05,
+      "loss": 0.6297,
+      "step": 20319
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1227246093486666e-05,
+      "loss": 0.635,
+      "step": 20320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1226538646955387e-05,
+      "loss": 0.6168,
+      "step": 20321
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.122583119419171e-05,
+      "loss": 0.567,
+      "step": 20322
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1225123735199224e-05,
+      "loss": 0.5794,
+      "step": 20323
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1224416269981526e-05,
+      "loss": 0.5605,
+      "step": 20324
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.122370879854221e-05,
+      "loss": 0.6028,
+      "step": 20325
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.122300132088487e-05,
+      "loss": 0.5797,
+      "step": 20326
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1222293837013103e-05,
+      "loss": 0.6782,
+      "step": 20327
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1221586346930503e-05,
+      "loss": 0.6819,
+      "step": 20328
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1220878850640667e-05,
+      "loss": 0.6473,
+      "step": 20329
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1220171348147186e-05,
+      "loss": 0.6123,
+      "step": 20330
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1219463839453658e-05,
+      "loss": 0.6191,
+      "step": 20331
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1218756324563675e-05,
+      "loss": 0.5708,
+      "step": 20332
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1218048803480837e-05,
+      "loss": 0.6346,
+      "step": 20333
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1217341276208736e-05,
+      "loss": 0.5488,
+      "step": 20334
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1216633742750966e-05,
+      "loss": 0.6005,
+      "step": 20335
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1215926203111123e-05,
+      "loss": 0.5775,
+      "step": 20336
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1215218657292803e-05,
+      "loss": 0.6142,
+      "step": 20337
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1214511105299603e-05,
+      "loss": 0.5929,
+      "step": 20338
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1213803547135116e-05,
+      "loss": 0.5903,
+      "step": 20339
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1213095982802938e-05,
+      "loss": 0.7196,
+      "step": 20340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.121238841230666e-05,
+      "loss": 0.5726,
+      "step": 20341
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1211680835649885e-05,
+      "loss": 0.5547,
+      "step": 20342
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1210973252836205e-05,
+      "loss": 0.5796,
+      "step": 20343
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1210265663869216e-05,
+      "loss": 0.6852,
+      "step": 20344
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.120955806875251e-05,
+      "loss": 0.5715,
+      "step": 20345
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1208850467489686e-05,
+      "loss": 0.6246,
+      "step": 20346
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.120814286008434e-05,
+      "loss": 0.6107,
+      "step": 20347
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1207435246540064e-05,
+      "loss": 0.5863,
+      "step": 20348
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1206727626860459e-05,
+      "loss": 0.5896,
+      "step": 20349
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1206020001049117e-05,
+      "loss": 0.6013,
+      "step": 20350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.1205312369109634e-05,
+      "loss": 0.561,
+      "step": 20351
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1204604731045606e-05,
+      "loss": 0.6458,
+      "step": 20352
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1203897086860625e-05,
+      "loss": 0.6757,
+      "step": 20353
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1203189436558295e-05,
+      "loss": 0.6036,
+      "step": 20354
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1202481780142204e-05,
+      "loss": 0.6898,
+      "step": 20355
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1201774117615951e-05,
+      "loss": 0.6582,
+      "step": 20356
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1201066448983133e-05,
+      "loss": 0.5822,
+      "step": 20357
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1200358774247344e-05,
+      "loss": 0.5811,
+      "step": 20358
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1199651093412182e-05,
+      "loss": 0.5724,
+      "step": 20359
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.119894340648124e-05,
+      "loss": 0.6019,
+      "step": 20360
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1198235713458114e-05,
+      "loss": 0.6356,
+      "step": 20361
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1197528014346406e-05,
+      "loss": 0.5633,
+      "step": 20362
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1196820309149703e-05,
+      "loss": 0.6369,
+      "step": 20363
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1196112597871606e-05,
+      "loss": 0.6183,
+      "step": 20364
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1195404880515713e-05,
+      "loss": 0.6179,
+      "step": 20365
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1194697157085616e-05,
+      "loss": 0.5971,
+      "step": 20366
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1193989427584913e-05,
+      "loss": 0.6496,
+      "step": 20367
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1193281692017202e-05,
+      "loss": 0.587,
+      "step": 20368
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1192573950386076e-05,
+      "loss": 0.5625,
+      "step": 20369
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1191866202695132e-05,
+      "loss": 0.6456,
+      "step": 20370
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1191158448947968e-05,
+      "loss": 0.6094,
+      "step": 20371
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1190450689148175e-05,
+      "loss": 0.6049,
+      "step": 20372
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1189742923299359e-05,
+      "loss": 0.6236,
+      "step": 20373
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1189035151405108e-05,
+      "loss": 0.5945,
+      "step": 20374
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1188327373469018e-05,
+      "loss": 0.6277,
+      "step": 20375
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1187619589494694e-05,
+      "loss": 0.6519,
+      "step": 20376
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1186911799485725e-05,
+      "loss": 0.6144,
+      "step": 20377
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.118620400344571e-05,
+      "loss": 0.65,
+      "step": 20378
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1185496201378243e-05,
+      "loss": 0.6016,
+      "step": 20379
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1184788393286925e-05,
+      "loss": 0.5977,
+      "step": 20380
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.118408057917535e-05,
+      "loss": 0.5282,
+      "step": 20381
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1183372759047112e-05,
+      "loss": 0.6265,
+      "step": 20382
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1182664932905814e-05,
+      "loss": 0.5379,
+      "step": 20383
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1181957100755048e-05,
+      "loss": 0.638,
+      "step": 20384
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1181249262598411e-05,
+      "loss": 0.5862,
+      "step": 20385
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1180541418439502e-05,
+      "loss": 0.6025,
+      "step": 20386
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1179833568281915e-05,
+      "loss": 0.6074,
+      "step": 20387
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1179125712129247e-05,
+      "loss": 0.5887,
+      "step": 20388
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1178417849985099e-05,
+      "loss": 0.6327,
+      "step": 20389
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1177709981853065e-05,
+      "loss": 0.579,
+      "step": 20390
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1177002107736738e-05,
+      "loss": 0.629,
+      "step": 20391
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1176294227639722e-05,
+      "loss": 0.5292,
+      "step": 20392
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1175586341565608e-05,
+      "loss": 0.6087,
+      "step": 20393
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1174878449517997e-05,
+      "loss": 0.5432,
+      "step": 20394
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1174170551500486e-05,
+      "loss": 0.5812,
+      "step": 20395
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1173462647516666e-05,
+      "loss": 0.6275,
+      "step": 20396
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1172754737570143e-05,
+      "loss": 0.5786,
+      "step": 20397
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1172046821664507e-05,
+      "loss": 0.5766,
+      "step": 20398
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1171338899803357e-05,
+      "loss": 0.61,
+      "step": 20399
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1170630971990295e-05,
+      "loss": 0.5747,
+      "step": 20400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1169923038228912e-05,
+      "loss": 0.5848,
+      "step": 20401
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1169215098522806e-05,
+      "loss": 0.6405,
+      "step": 20402
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1168507152875578e-05,
+      "loss": 0.5332,
+      "step": 20403
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1167799201290821e-05,
+      "loss": 0.59,
+      "step": 20404
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1167091243772137e-05,
+      "loss": 0.6796,
+      "step": 20405
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1166383280323117e-05,
+      "loss": 0.5632,
+      "step": 20406
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1165675310947364e-05,
+      "loss": 0.6283,
+      "step": 20407
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1164967335648473e-05,
+      "loss": 0.559,
+      "step": 20408
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1164259354430044e-05,
+      "loss": 0.5564,
+      "step": 20409
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1163551367295668e-05,
+      "loss": 0.5797,
+      "step": 20410
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.116284337424895e-05,
+      "loss": 0.5549,
+      "step": 20411
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1162135375293484e-05,
+      "loss": 0.5768,
+      "step": 20412
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1161427370432866e-05,
+      "loss": 0.6237,
+      "step": 20413
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1160719359670698e-05,
+      "loss": 0.5681,
+      "step": 20414
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1160011343010573e-05,
+      "loss": 0.5785,
+      "step": 20415
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1159303320456091e-05,
+      "loss": 0.5376,
+      "step": 20416
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.115859529201085e-05,
+      "loss": 0.5502,
+      "step": 20417
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1157887257678446e-05,
+      "loss": 0.5817,
+      "step": 20418
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.115717921746248e-05,
+      "loss": 0.5698,
+      "step": 20419
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1156471171366544e-05,
+      "loss": 0.5684,
+      "step": 20420
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1155763119394238e-05,
+      "loss": 0.6279,
+      "step": 20421
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1155055061549168e-05,
+      "loss": 0.586,
+      "step": 20422
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1154346997834919e-05,
+      "loss": 0.6415,
+      "step": 20423
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1153638928255098e-05,
+      "loss": 0.6048,
+      "step": 20424
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1152930852813297e-05,
+      "loss": 0.6001,
+      "step": 20425
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1152222771513118e-05,
+      "loss": 0.5966,
+      "step": 20426
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.115151468435816e-05,
+      "loss": 0.5663,
+      "step": 20427
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1150806591352015e-05,
+      "loss": 0.6536,
+      "step": 20428
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1150098492498288e-05,
+      "loss": 0.6508,
+      "step": 20429
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1149390387800572e-05,
+      "loss": 0.5977,
+      "step": 20430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1148682277262466e-05,
+      "loss": 0.587,
+      "step": 20431
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1147974160887572e-05,
+      "loss": 0.5656,
+      "step": 20432
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1147266038679481e-05,
+      "loss": 0.5872,
+      "step": 20433
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1146557910641798e-05,
+      "loss": 0.5925,
+      "step": 20434
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1145849776778117e-05,
+      "loss": 0.6298,
+      "step": 20435
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1145141637092038e-05,
+      "loss": 0.5708,
+      "step": 20436
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1144433491587158e-05,
+      "loss": 0.5895,
+      "step": 20437
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.114372534026708e-05,
+      "loss": 0.5849,
+      "step": 20438
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1143017183135394e-05,
+      "loss": 0.62,
+      "step": 20439
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1142309020195705e-05,
+      "loss": 0.5666,
+      "step": 20440
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.114160085145161e-05,
+      "loss": 0.6381,
+      "step": 20441
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1140892676906705e-05,
+      "loss": 0.5451,
+      "step": 20442
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1140184496564592e-05,
+      "loss": 0.6137,
+      "step": 20443
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1139476310428868e-05,
+      "loss": 0.5611,
+      "step": 20444
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.113876811850313e-05,
+      "loss": 0.6391,
+      "step": 20445
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1138059920790978e-05,
+      "loss": 0.6213,
+      "step": 20446
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.113735171729601e-05,
+      "loss": 0.6757,
+      "step": 20447
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1136643508021826e-05,
+      "loss": 0.6029,
+      "step": 20448
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.113593529297202e-05,
+      "loss": 0.5808,
+      "step": 20449
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1135227072150199e-05,
+      "loss": 0.6005,
+      "step": 20450
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1134518845559954e-05,
+      "loss": 0.5586,
+      "step": 20451
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1133810613204887e-05,
+      "loss": 0.549,
+      "step": 20452
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1133102375088595e-05,
+      "loss": 0.6049,
+      "step": 20453
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1132394131214679e-05,
+      "loss": 0.5852,
+      "step": 20454
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1131685881586738e-05,
+      "loss": 0.6631,
+      "step": 20455
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1130977626208366e-05,
+      "loss": 0.6258,
+      "step": 20456
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1130269365083168e-05,
+      "loss": 0.559,
+      "step": 20457
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.112956109821474e-05,
+      "loss": 0.5988,
+      "step": 20458
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1128852825606681e-05,
+      "loss": 0.6076,
+      "step": 20459
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1128144547262591e-05,
+      "loss": 0.5982,
+      "step": 20460
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1127436263186069e-05,
+      "loss": 0.618,
+      "step": 20461
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.112672797338071e-05,
+      "loss": 0.5702,
+      "step": 20462
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1126019677850119e-05,
+      "loss": 0.5524,
+      "step": 20463
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.112531137659789e-05,
+      "loss": 0.5849,
+      "step": 20464
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1124603069627628e-05,
+      "loss": 0.5875,
+      "step": 20465
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1123894756942925e-05,
+      "loss": 0.6021,
+      "step": 20466
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1123186438547387e-05,
+      "loss": 0.6494,
+      "step": 20467
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1122478114444606e-05,
+      "loss": 0.5874,
+      "step": 20468
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1121769784638187e-05,
+      "loss": 0.5633,
+      "step": 20469
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1121061449131727e-05,
+      "loss": 0.5803,
+      "step": 20470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1120353107928826e-05,
+      "loss": 0.6024,
+      "step": 20471
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1119644761033079e-05,
+      "loss": 0.6128,
+      "step": 20472
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1118936408448093e-05,
+      "loss": 0.6059,
+      "step": 20473
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1118228050177462e-05,
+      "loss": 0.7012,
+      "step": 20474
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1117519686224788e-05,
+      "loss": 0.6465,
+      "step": 20475
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1116811316593667e-05,
+      "loss": 0.6109,
+      "step": 20476
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1116102941287703e-05,
+      "loss": 0.5827,
+      "step": 20477
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1115394560310493e-05,
+      "loss": 0.5547,
+      "step": 20478
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1114686173665634e-05,
+      "loss": 0.5619,
+      "step": 20479
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.111397778135673e-05,
+      "loss": 0.5733,
+      "step": 20480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1113269383387379e-05,
+      "loss": 0.5565,
+      "step": 20481
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1112560979761177e-05,
+      "loss": 0.5785,
+      "step": 20482
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.111185257048173e-05,
+      "loss": 0.5225,
+      "step": 20483
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1111144155552635e-05,
+      "loss": 0.6321,
+      "step": 20484
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.111043573497749e-05,
+      "loss": 0.5889,
+      "step": 20485
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1109727308759896e-05,
+      "loss": 0.5977,
+      "step": 20486
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1109018876903453e-05,
+      "loss": 0.6451,
+      "step": 20487
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.110831043941176e-05,
+      "loss": 0.6399,
+      "step": 20488
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1107601996288416e-05,
+      "loss": 0.5923,
+      "step": 20489
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1106893547537023e-05,
+      "loss": 0.6249,
+      "step": 20490
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1106185093161179e-05,
+      "loss": 0.611,
+      "step": 20491
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1105476633164487e-05,
+      "loss": 0.6338,
+      "step": 20492
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1104768167550542e-05,
+      "loss": 0.631,
+      "step": 20493
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1104059696322944e-05,
+      "loss": 0.645,
+      "step": 20494
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.11033512194853e-05,
+      "loss": 0.5876,
+      "step": 20495
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1102642737041202e-05,
+      "loss": 0.537,
+      "step": 20496
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1101934248994257e-05,
+      "loss": 0.5425,
+      "step": 20497
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.1101225755348058e-05,
+      "loss": 0.5342,
+      "step": 20498
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1100517256106208e-05,
+      "loss": 0.5794,
+      "step": 20499
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1099808751272312e-05,
+      "loss": 0.5695,
+      "step": 20500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.109910024084996e-05,
+      "loss": 0.5527,
+      "step": 20501
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.109839172484276e-05,
+      "loss": 0.5921,
+      "step": 20502
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1097683203254313e-05,
+      "loss": 0.6003,
+      "step": 20503
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1096974676088213e-05,
+      "loss": 0.5656,
+      "step": 20504
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1096266143348064e-05,
+      "loss": 0.5905,
+      "step": 20505
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1095557605037466e-05,
+      "loss": 0.5743,
+      "step": 20506
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.109484906116002e-05,
+      "loss": 0.648,
+      "step": 20507
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1094140511719323e-05,
+      "loss": 0.5261,
+      "step": 20508
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1093431956718978e-05,
+      "loss": 0.5786,
+      "step": 20509
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1092723396162586e-05,
+      "loss": 0.6081,
+      "step": 20510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1092014830053748e-05,
+      "loss": 0.5414,
+      "step": 20511
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.109130625839606e-05,
+      "loss": 0.59,
+      "step": 20512
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1090597681193124e-05,
+      "loss": 0.5721,
+      "step": 20513
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1089889098448546e-05,
+      "loss": 0.6406,
+      "step": 20514
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.108918051016592e-05,
+      "loss": 0.5525,
+      "step": 20515
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1088471916348851e-05,
+      "loss": 0.5754,
+      "step": 20516
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1087763317000935e-05,
+      "loss": 0.5651,
+      "step": 20517
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1087054712125773e-05,
+      "loss": 0.5804,
+      "step": 20518
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1086346101726973e-05,
+      "loss": 0.6147,
+      "step": 20519
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1085637485808124e-05,
+      "loss": 0.615,
+      "step": 20520
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1084928864372838e-05,
+      "loss": 0.5735,
+      "step": 20521
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.108422023742471e-05,
+      "loss": 0.585,
+      "step": 20522
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1083511604967338e-05,
+      "loss": 0.5813,
+      "step": 20523
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1082802967004331e-05,
+      "loss": 0.6102,
+      "step": 20524
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1082094323539283e-05,
+      "loss": 0.5703,
+      "step": 20525
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1081385674575797e-05,
+      "loss": 0.5924,
+      "step": 20526
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1080677020117472e-05,
+      "loss": 0.5944,
+      "step": 20527
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1079968360167914e-05,
+      "loss": 0.5664,
+      "step": 20528
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1079259694730718e-05,
+      "loss": 0.6598,
+      "step": 20529
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.107855102380949e-05,
+      "loss": 0.586,
+      "step": 20530
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1077842347407823e-05,
+      "loss": 0.6524,
+      "step": 20531
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1077133665529328e-05,
+      "loss": 0.6438,
+      "step": 20532
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.10764249781776e-05,
+      "loss": 0.6093,
+      "step": 20533
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.107571628535624e-05,
+      "loss": 0.6479,
+      "step": 20534
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1075007587068853e-05,
+      "loss": 0.5937,
+      "step": 20535
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1074298883319037e-05,
+      "loss": 0.6408,
+      "step": 20536
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1073590174110392e-05,
+      "loss": 0.5896,
+      "step": 20537
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1072881459446522e-05,
+      "loss": 0.6583,
+      "step": 20538
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1072172739331028e-05,
+      "loss": 0.6493,
+      "step": 20539
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1071464013767508e-05,
+      "loss": 0.6187,
+      "step": 20540
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1070755282759568e-05,
+      "loss": 0.5964,
+      "step": 20541
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1070046546310806e-05,
+      "loss": 0.5809,
+      "step": 20542
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1069337804424824e-05,
+      "loss": 0.5456,
+      "step": 20543
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1068629057105225e-05,
+      "loss": 0.5785,
+      "step": 20544
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1067920304355607e-05,
+      "loss": 0.5967,
+      "step": 20545
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1067211546179573e-05,
+      "loss": 0.5751,
+      "step": 20546
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1066502782580727e-05,
+      "loss": 0.6387,
+      "step": 20547
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1065794013562665e-05,
+      "loss": 0.5843,
+      "step": 20548
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1065085239128994e-05,
+      "loss": 0.6045,
+      "step": 20549
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.106437645928331e-05,
+      "loss": 0.5441,
+      "step": 20550
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1063667674029219e-05,
+      "loss": 0.59,
+      "step": 20551
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1062958883370323e-05,
+      "loss": 0.632,
+      "step": 20552
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1062250087310217e-05,
+      "loss": 0.6301,
+      "step": 20553
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1061541285852513e-05,
+      "loss": 0.5705,
+      "step": 20554
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1060832479000803e-05,
+      "loss": 0.6311,
+      "step": 20555
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.106012366675869e-05,
+      "loss": 0.5993,
+      "step": 20556
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1059414849129784e-05,
+      "loss": 0.6216,
+      "step": 20557
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1058706026117678e-05,
+      "loss": 0.6295,
+      "step": 20558
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1057997197725976e-05,
+      "loss": 0.5729,
+      "step": 20559
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1057288363958282e-05,
+      "loss": 0.6331,
+      "step": 20560
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1056579524818197e-05,
+      "loss": 0.6121,
+      "step": 20561
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.105587068030932e-05,
+      "loss": 0.5824,
+      "step": 20562
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1055161830435254e-05,
+      "loss": 0.6625,
+      "step": 20563
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1054452975199602e-05,
+      "loss": 0.6057,
+      "step": 20564
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.105374411460597e-05,
+      "loss": 0.5817,
+      "step": 20565
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1053035248657953e-05,
+      "loss": 0.6069,
+      "step": 20566
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1052326377359153e-05,
+      "loss": 0.5697,
+      "step": 20567
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1051617500713179e-05,
+      "loss": 0.6345,
+      "step": 20568
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1050908618723624e-05,
+      "loss": 0.5988,
+      "step": 20569
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1050199731394099e-05,
+      "loss": 0.6676,
+      "step": 20570
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1049490838728197e-05,
+      "loss": 0.5654,
+      "step": 20571
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1048781940729524e-05,
+      "loss": 0.6059,
+      "step": 20572
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1048073037401688e-05,
+      "loss": 0.6179,
+      "step": 20573
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1047364128748283e-05,
+      "loss": 0.5843,
+      "step": 20574
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1046655214772914e-05,
+      "loss": 0.601,
+      "step": 20575
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1045946295479185e-05,
+      "loss": 0.6077,
+      "step": 20576
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1045237370870694e-05,
+      "loss": 0.6368,
+      "step": 20577
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1044528440951045e-05,
+      "loss": 0.608,
+      "step": 20578
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1043819505723845e-05,
+      "loss": 0.6575,
+      "step": 20579
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.104311056519269e-05,
+      "loss": 0.6463,
+      "step": 20580
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1042401619361188e-05,
+      "loss": 0.607,
+      "step": 20581
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1041692668232935e-05,
+      "loss": 0.6146,
+      "step": 20582
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1040983711811535e-05,
+      "loss": 0.6332,
+      "step": 20583
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1040274750100595e-05,
+      "loss": 0.5589,
+      "step": 20584
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1039565783103711e-05,
+      "loss": 0.5878,
+      "step": 20585
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1038856810824494e-05,
+      "loss": 0.5835,
+      "step": 20586
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.103814783326654e-05,
+      "loss": 0.626,
+      "step": 20587
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.103743885043345e-05,
+      "loss": 0.5853,
+      "step": 20588
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1036729862328832e-05,
+      "loss": 0.6427,
+      "step": 20589
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1036020868956283e-05,
+      "loss": 0.5795,
+      "step": 20590
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.103531187031941e-05,
+      "loss": 0.5759,
+      "step": 20591
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1034602866421816e-05,
+      "loss": 0.5956,
+      "step": 20592
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1033893857267098e-05,
+      "loss": 0.619,
+      "step": 20593
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1033184842858865e-05,
+      "loss": 0.6193,
+      "step": 20594
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1032475823200716e-05,
+      "loss": 0.613,
+      "step": 20595
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1031766798296256e-05,
+      "loss": 0.6053,
+      "step": 20596
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1031057768149085e-05,
+      "loss": 0.5853,
+      "step": 20597
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.103034873276281e-05,
+      "loss": 0.5921,
+      "step": 20598
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.102963969214103e-05,
+      "loss": 0.5689,
+      "step": 20599
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.102893064628735e-05,
+      "loss": 0.5824,
+      "step": 20600
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1028221595205371e-05,
+      "loss": 0.5279,
+      "step": 20601
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1027512538898697e-05,
+      "loss": 0.5798,
+      "step": 20602
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1026803477370933e-05,
+      "loss": 0.5559,
+      "step": 20603
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1026094410625677e-05,
+      "loss": 0.6279,
+      "step": 20604
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1025385338666537e-05,
+      "loss": 0.5895,
+      "step": 20605
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1024676261497113e-05,
+      "loss": 0.5964,
+      "step": 20606
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1023967179121008e-05,
+      "loss": 0.6385,
+      "step": 20607
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1023258091541829e-05,
+      "loss": 0.5921,
+      "step": 20608
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1022548998763172e-05,
+      "loss": 0.5791,
+      "step": 20609
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1021839900788645e-05,
+      "loss": 0.6407,
+      "step": 20610
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1021130797621854e-05,
+      "loss": 0.583,
+      "step": 20611
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1020421689266392e-05,
+      "loss": 0.6669,
+      "step": 20612
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1019712575725872e-05,
+      "loss": 0.5848,
+      "step": 20613
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1019003457003895e-05,
+      "loss": 0.6056,
+      "step": 20614
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1018294333104061e-05,
+      "loss": 0.5923,
+      "step": 20615
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1017585204029974e-05,
+      "loss": 0.6439,
+      "step": 20616
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1016876069785241e-05,
+      "loss": 0.6024,
+      "step": 20617
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1016166930373463e-05,
+      "loss": 0.5562,
+      "step": 20618
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1015457785798242e-05,
+      "loss": 0.6138,
+      "step": 20619
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1014748636063182e-05,
+      "loss": 0.5628,
+      "step": 20620
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1014039481171887e-05,
+      "loss": 0.5744,
+      "step": 20621
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1013330321127962e-05,
+      "loss": 0.5922,
+      "step": 20622
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1012621155935007e-05,
+      "loss": 0.6107,
+      "step": 20623
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.101191198559663e-05,
+      "loss": 0.576,
+      "step": 20624
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.101120281011643e-05,
+      "loss": 0.6048,
+      "step": 20625
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1010493629498013e-05,
+      "loss": 0.5861,
+      "step": 20626
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1009784443744985e-05,
+      "loss": 0.587,
+      "step": 20627
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1009075252860942e-05,
+      "loss": 0.6016,
+      "step": 20628
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1008366056849494e-05,
+      "loss": 0.5671,
+      "step": 20629
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1007656855714241e-05,
+      "loss": 0.6844,
+      "step": 20630
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.100694764945879e-05,
+      "loss": 0.6004,
+      "step": 20631
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.100623843808674e-05,
+      "loss": 0.6278,
+      "step": 20632
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1005529221601702e-05,
+      "loss": 0.5509,
+      "step": 20633
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1004820000007274e-05,
+      "loss": 0.5665,
+      "step": 20634
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1004110773307062e-05,
+      "loss": 0.6895,
+      "step": 20635
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1003401541504668e-05,
+      "loss": 0.5856,
+      "step": 20636
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1002692304603697e-05,
+      "loss": 0.5809,
+      "step": 20637
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1001983062607755e-05,
+      "loss": 0.5815,
+      "step": 20638
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.1001273815520439e-05,
+      "loss": 0.5513,
+      "step": 20639
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.100056456334536e-05,
+      "loss": 0.6783,
+      "step": 20640
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0999855306086121e-05,
+      "loss": 0.5511,
+      "step": 20641
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0999146043746323e-05,
+      "loss": 0.6178,
+      "step": 20642
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0998436776329574e-05,
+      "loss": 0.6569,
+      "step": 20643
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.099772750383947e-05,
+      "loss": 0.6365,
+      "step": 20644
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0997018226279623e-05,
+      "loss": 0.6737,
+      "step": 20645
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0996308943653637e-05,
+      "loss": 0.6277,
+      "step": 20646
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.099559965596511e-05,
+      "loss": 0.6721,
+      "step": 20647
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.099489036321765e-05,
+      "loss": 0.5614,
+      "step": 20648
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0994181065414862e-05,
+      "loss": 0.6027,
+      "step": 20649
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0993471762560347e-05,
+      "loss": 0.5603,
+      "step": 20650
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0992762454657711e-05,
+      "loss": 0.5695,
+      "step": 20651
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0992053141710561e-05,
+      "loss": 0.6335,
+      "step": 20652
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0991343823722496e-05,
+      "loss": 0.6417,
+      "step": 20653
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0990634500697124e-05,
+      "loss": 0.6134,
+      "step": 20654
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0989925172638047e-05,
+      "loss": 0.6553,
+      "step": 20655
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.098921583954887e-05,
+      "loss": 0.5089,
+      "step": 20656
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0988506501433198e-05,
+      "loss": 0.5769,
+      "step": 20657
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0987797158294635e-05,
+      "loss": 0.5961,
+      "step": 20658
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0987087810136785e-05,
+      "loss": 0.6148,
+      "step": 20659
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0986378456963254e-05,
+      "loss": 0.6206,
+      "step": 20660
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0985669098777644e-05,
+      "loss": 0.6045,
+      "step": 20661
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0984959735583563e-05,
+      "loss": 0.5735,
+      "step": 20662
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.098425036738461e-05,
+      "loss": 0.6263,
+      "step": 20663
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0983540994184393e-05,
+      "loss": 0.6684,
+      "step": 20664
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0982831615986517e-05,
+      "loss": 0.694,
+      "step": 20665
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0982122232794588e-05,
+      "loss": 0.5762,
+      "step": 20666
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0981412844612204e-05,
+      "loss": 0.564,
+      "step": 20667
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.098070345144298e-05,
+      "loss": 0.6187,
+      "step": 20668
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0979994053290509e-05,
+      "loss": 0.6128,
+      "step": 20669
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0979284650158402e-05,
+      "loss": 0.564,
+      "step": 20670
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0978575242050265e-05,
+      "loss": 0.5647,
+      "step": 20671
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0977865828969698e-05,
+      "loss": 0.6608,
+      "step": 20672
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0977156410920311e-05,
+      "loss": 0.5722,
+      "step": 20673
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0976446987905705e-05,
+      "loss": 0.6692,
+      "step": 20674
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0975737559929485e-05,
+      "loss": 0.6208,
+      "step": 20675
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0975028126995257e-05,
+      "loss": 0.6614,
+      "step": 20676
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0974318689106624e-05,
+      "loss": 0.5596,
+      "step": 20677
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0973609246267193e-05,
+      "loss": 0.6186,
+      "step": 20678
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0972899798480572e-05,
+      "loss": 0.6156,
+      "step": 20679
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0972190345750359e-05,
+      "loss": 0.5919,
+      "step": 20680
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0971480888080164e-05,
+      "loss": 0.6056,
+      "step": 20681
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.097077142547359e-05,
+      "loss": 0.5881,
+      "step": 20682
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0970061957934241e-05,
+      "loss": 0.6159,
+      "step": 20683
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0969352485465723e-05,
+      "loss": 0.5523,
+      "step": 20684
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0968643008071643e-05,
+      "loss": 0.5724,
+      "step": 20685
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.09679335257556e-05,
+      "loss": 0.5996,
+      "step": 20686
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0967224038521206e-05,
+      "loss": 0.6588,
+      "step": 20687
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0966514546372064e-05,
+      "loss": 0.5609,
+      "step": 20688
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0965805049311778e-05,
+      "loss": 0.642,
+      "step": 20689
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0965095547343953e-05,
+      "loss": 0.6171,
+      "step": 20690
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0964386040472194e-05,
+      "loss": 0.5708,
+      "step": 20691
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0963676528700111e-05,
+      "loss": 0.6202,
+      "step": 20692
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.09629670120313e-05,
+      "loss": 0.603,
+      "step": 20693
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0962257490469374e-05,
+      "loss": 0.567,
+      "step": 20694
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0961547964017938e-05,
+      "loss": 0.6547,
+      "step": 20695
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0960838432680592e-05,
+      "loss": 0.5702,
+      "step": 20696
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0960128896460947e-05,
+      "loss": 0.5921,
+      "step": 20697
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0959419355362604e-05,
+      "loss": 0.5791,
+      "step": 20698
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0958709809389172e-05,
+      "loss": 0.5949,
+      "step": 20699
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0958000258544255e-05,
+      "loss": 0.6585,
+      "step": 20700
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0957290702831456e-05,
+      "loss": 0.5865,
+      "step": 20701
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0956581142254383e-05,
+      "loss": 0.5174,
+      "step": 20702
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0955871576816642e-05,
+      "loss": 0.56,
+      "step": 20703
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0955162006521838e-05,
+      "loss": 0.5753,
+      "step": 20704
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0954452431373575e-05,
+      "loss": 0.5936,
+      "step": 20705
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.095374285137546e-05,
+      "loss": 0.6057,
+      "step": 20706
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0953033266531097e-05,
+      "loss": 0.5772,
+      "step": 20707
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0952323676844097e-05,
+      "loss": 0.5814,
+      "step": 20708
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0951614082318055e-05,
+      "loss": 0.5769,
+      "step": 20709
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0950904482956589e-05,
+      "loss": 0.5886,
+      "step": 20710
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0950194878763295e-05,
+      "loss": 0.6932,
+      "step": 20711
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0949485269741784e-05,
+      "loss": 0.5405,
+      "step": 20712
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.094877565589566e-05,
+      "loss": 0.641,
+      "step": 20713
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0948066037228531e-05,
+      "loss": 0.6026,
+      "step": 20714
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0947356413743996e-05,
+      "loss": 0.5721,
+      "step": 20715
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.094664678544567e-05,
+      "loss": 0.6407,
+      "step": 20716
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0945937152337154e-05,
+      "loss": 0.5647,
+      "step": 20717
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0945227514422052e-05,
+      "loss": 0.5936,
+      "step": 20718
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0944517871703978e-05,
+      "loss": 0.5864,
+      "step": 20719
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0943808224186525e-05,
+      "loss": 0.606,
+      "step": 20720
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0943098571873307e-05,
+      "loss": 0.6586,
+      "step": 20721
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0942388914767932e-05,
+      "loss": 0.6517,
+      "step": 20722
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0941679252874005e-05,
+      "loss": 0.5058,
+      "step": 20723
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0940969586195125e-05,
+      "loss": 0.5853,
+      "step": 20724
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0940259914734905e-05,
+      "loss": 0.6305,
+      "step": 20725
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0939550238496947e-05,
+      "loss": 0.5526,
+      "step": 20726
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0938840557484862e-05,
+      "loss": 0.664,
+      "step": 20727
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0938130871702253e-05,
+      "loss": 0.619,
+      "step": 20728
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0937421181152723e-05,
+      "loss": 0.5623,
+      "step": 20729
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0936711485839884e-05,
+      "loss": 0.5503,
+      "step": 20730
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0936001785767338e-05,
+      "loss": 0.6082,
+      "step": 20731
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0935292080938696e-05,
+      "loss": 0.6245,
+      "step": 20732
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0934582371357558e-05,
+      "loss": 0.6127,
+      "step": 20733
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0933872657027534e-05,
+      "loss": 0.6181,
+      "step": 20734
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0933162937952229e-05,
+      "loss": 0.5998,
+      "step": 20735
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0932453214135252e-05,
+      "loss": 0.5675,
+      "step": 20736
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0931743485580204e-05,
+      "loss": 0.6061,
+      "step": 20737
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0931033752290699e-05,
+      "loss": 0.5905,
+      "step": 20738
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0930324014270335e-05,
+      "loss": 0.5915,
+      "step": 20739
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0929614271522721e-05,
+      "loss": 0.5937,
+      "step": 20740
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0928904524051469e-05,
+      "loss": 0.6259,
+      "step": 20741
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0928194771860181e-05,
+      "loss": 0.6242,
+      "step": 20742
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.092748501495246e-05,
+      "loss": 0.5929,
+      "step": 20743
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0926775253331918e-05,
+      "loss": 0.5801,
+      "step": 20744
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.092606548700216e-05,
+      "loss": 0.6985,
+      "step": 20745
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0925355715966792e-05,
+      "loss": 0.5641,
+      "step": 20746
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.092464594022942e-05,
+      "loss": 0.6401,
+      "step": 20747
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0923936159793649e-05,
+      "loss": 0.6327,
+      "step": 20748
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.092322637466309e-05,
+      "loss": 0.6488,
+      "step": 20749
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0922516584841347e-05,
+      "loss": 0.5928,
+      "step": 20750
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0921806790332026e-05,
+      "loss": 0.6413,
+      "step": 20751
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0921096991138737e-05,
+      "loss": 0.6208,
+      "step": 20752
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0920387187265082e-05,
+      "loss": 0.6427,
+      "step": 20753
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0919677378714672e-05,
+      "loss": 0.6945,
+      "step": 20754
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0918967565491111e-05,
+      "loss": 0.558,
+      "step": 20755
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0918257747598007e-05,
+      "loss": 0.6014,
+      "step": 20756
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0917547925038968e-05,
+      "loss": 0.6033,
+      "step": 20757
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0916838097817597e-05,
+      "loss": 0.5618,
+      "step": 20758
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0916128265937502e-05,
+      "loss": 0.5914,
+      "step": 20759
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0915418429402295e-05,
+      "loss": 0.6425,
+      "step": 20760
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0914708588215575e-05,
+      "loss": 0.6013,
+      "step": 20761
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0913998742380956e-05,
+      "loss": 0.6433,
+      "step": 20762
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.091328889190204e-05,
+      "loss": 0.5743,
+      "step": 20763
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0912579036782437e-05,
+      "loss": 0.6107,
+      "step": 20764
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0911869177025754e-05,
+      "loss": 0.5765,
+      "step": 20765
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0911159312635593e-05,
+      "loss": 0.6099,
+      "step": 20766
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0910449443615565e-05,
+      "loss": 0.6587,
+      "step": 20767
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0909739569969278e-05,
+      "loss": 0.539,
+      "step": 20768
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0909029691700336e-05,
+      "loss": 0.6307,
+      "step": 20769
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.090831980881235e-05,
+      "loss": 0.5816,
+      "step": 20770
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0907609921308926e-05,
+      "loss": 0.6139,
+      "step": 20771
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0906900029193665e-05,
+      "loss": 0.5767,
+      "step": 20772
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0906190132470186e-05,
+      "loss": 0.6188,
+      "step": 20773
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0905480231142086e-05,
+      "loss": 0.6093,
+      "step": 20774
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0904770325212975e-05,
+      "loss": 0.5859,
+      "step": 20775
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0904060414686464e-05,
+      "loss": 0.6097,
+      "step": 20776
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0903350499566154e-05,
+      "loss": 0.5672,
+      "step": 20777
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0902640579855656e-05,
+      "loss": 0.5588,
+      "step": 20778
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.090193065555858e-05,
+      "loss": 0.6251,
+      "step": 20779
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0901220726678527e-05,
+      "loss": 0.6013,
+      "step": 20780
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.090051079321911e-05,
+      "loss": 0.5862,
+      "step": 20781
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.089980085518393e-05,
+      "loss": 0.6031,
+      "step": 20782
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0899090912576604e-05,
+      "loss": 0.6103,
+      "step": 20783
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0898380965400732e-05,
+      "loss": 0.5646,
+      "step": 20784
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.089767101365992e-05,
+      "loss": 0.5962,
+      "step": 20785
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.089696105735778e-05,
+      "loss": 0.5957,
+      "step": 20786
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.089625109649792e-05,
+      "loss": 0.6177,
+      "step": 20787
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0895541131083944e-05,
+      "loss": 0.5933,
+      "step": 20788
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.089483116111946e-05,
+      "loss": 0.5801,
+      "step": 20789
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0894121186608082e-05,
+      "loss": 0.6162,
+      "step": 20790
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0893411207553408e-05,
+      "loss": 0.6318,
+      "step": 20791
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0892701223959053e-05,
+      "loss": 0.6282,
+      "step": 20792
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0891991235828616e-05,
+      "loss": 0.6171,
+      "step": 20793
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0891281243165715e-05,
+      "loss": 0.5785,
+      "step": 20794
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0890571245973953e-05,
+      "loss": 0.5854,
+      "step": 20795
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0889861244256933e-05,
+      "loss": 0.5918,
+      "step": 20796
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0889151238018272e-05,
+      "loss": 0.58,
+      "step": 20797
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0888441227261572e-05,
+      "loss": 0.6175,
+      "step": 20798
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0887731211990441e-05,
+      "loss": 0.5986,
+      "step": 20799
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0887021192208491e-05,
+      "loss": 0.6697,
+      "step": 20800
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0886311167919323e-05,
+      "loss": 0.5654,
+      "step": 20801
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0885601139126551e-05,
+      "loss": 0.5838,
+      "step": 20802
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0884891105833782e-05,
+      "loss": 0.6063,
+      "step": 20803
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0884181068044618e-05,
+      "loss": 0.5745,
+      "step": 20804
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.088347102576267e-05,
+      "loss": 0.5899,
+      "step": 20805
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0882760978991552e-05,
+      "loss": 0.5921,
+      "step": 20806
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0882050927734862e-05,
+      "loss": 0.5816,
+      "step": 20807
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0881340871996216e-05,
+      "loss": 0.6201,
+      "step": 20808
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0880630811779218e-05,
+      "loss": 0.6407,
+      "step": 20809
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0879920747087476e-05,
+      "loss": 0.575,
+      "step": 20810
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0879210677924601e-05,
+      "loss": 0.6309,
+      "step": 20811
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0878500604294198e-05,
+      "loss": 0.5514,
+      "step": 20812
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0877790526199874e-05,
+      "loss": 0.5875,
+      "step": 20813
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0877080443645242e-05,
+      "loss": 0.5767,
+      "step": 20814
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0876370356633908e-05,
+      "loss": 0.6496,
+      "step": 20815
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0875660265169477e-05,
+      "loss": 0.5914,
+      "step": 20816
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0874950169255562e-05,
+      "loss": 0.6363,
+      "step": 20817
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0874240068895767e-05,
+      "loss": 0.6273,
+      "step": 20818
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0873529964093703e-05,
+      "loss": 0.6446,
+      "step": 20819
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0872819854852978e-05,
+      "loss": 0.6074,
+      "step": 20820
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0872109741177202e-05,
+      "loss": 0.5751,
+      "step": 20821
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0871399623069977e-05,
+      "loss": 0.5944,
+      "step": 20822
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0870689500534917e-05,
+      "loss": 0.6311,
+      "step": 20823
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0869979373575628e-05,
+      "loss": 0.5694,
+      "step": 20824
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0869269242195718e-05,
+      "loss": 0.5985,
+      "step": 20825
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.08685591063988e-05,
+      "loss": 0.5905,
+      "step": 20826
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0867848966188474e-05,
+      "loss": 0.5857,
+      "step": 20827
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0867138821568357e-05,
+      "loss": 0.6162,
+      "step": 20828
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.086642867254205e-05,
+      "loss": 0.5728,
+      "step": 20829
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.086571851911317e-05,
+      "loss": 0.6654,
+      "step": 20830
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0865008361285319e-05,
+      "loss": 0.6132,
+      "step": 20831
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0864298199062104e-05,
+      "loss": 0.575,
+      "step": 20832
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0863588032447141e-05,
+      "loss": 0.5609,
+      "step": 20833
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.086287786144403e-05,
+      "loss": 0.5501,
+      "step": 20834
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0862167686056386e-05,
+      "loss": 0.5589,
+      "step": 20835
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0861457506287818e-05,
+      "loss": 0.6127,
+      "step": 20836
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0860747322141928e-05,
+      "loss": 0.656,
+      "step": 20837
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0860037133622331e-05,
+      "loss": 0.6482,
+      "step": 20838
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0859326940732635e-05,
+      "loss": 0.6041,
+      "step": 20839
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0858616743476448e-05,
+      "loss": 0.6216,
+      "step": 20840
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0857906541857372e-05,
+      "loss": 0.6115,
+      "step": 20841
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0857196335879025e-05,
+      "loss": 0.6275,
+      "step": 20842
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0856486125545011e-05,
+      "loss": 0.6582,
+      "step": 20843
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0855775910858943e-05,
+      "loss": 0.5519,
+      "step": 20844
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0855065691824425e-05,
+      "loss": 0.6074,
+      "step": 20845
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.085435546844507e-05,
+      "loss": 0.6141,
+      "step": 20846
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0853645240724482e-05,
+      "loss": 0.5653,
+      "step": 20847
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0852935008666272e-05,
+      "loss": 0.6418,
+      "step": 20848
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0852224772274051e-05,
+      "loss": 0.5858,
+      "step": 20849
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0851514531551426e-05,
+      "loss": 0.558,
+      "step": 20850
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0850804286502005e-05,
+      "loss": 0.5826,
+      "step": 20851
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0850094037129401e-05,
+      "loss": 0.6109,
+      "step": 20852
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0849383783437219e-05,
+      "loss": 0.6258,
+      "step": 20853
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0848673525429068e-05,
+      "loss": 0.6271,
+      "step": 20854
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0847963263108561e-05,
+      "loss": 0.6268,
+      "step": 20855
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0847252996479301e-05,
+      "loss": 0.5754,
+      "step": 20856
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0846542725544904e-05,
+      "loss": 0.6182,
+      "step": 20857
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0845832450308972e-05,
+      "loss": 0.638,
+      "step": 20858
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.084512217077512e-05,
+      "loss": 0.604,
+      "step": 20859
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0844411886946953e-05,
+      "loss": 0.6524,
+      "step": 20860
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0843701598828084e-05,
+      "loss": 0.5976,
+      "step": 20861
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0842991306422119e-05,
+      "loss": 0.57,
+      "step": 20862
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0842281009732669e-05,
+      "loss": 0.6713,
+      "step": 20863
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.084157070876334e-05,
+      "loss": 0.6581,
+      "step": 20864
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0840860403517748e-05,
+      "loss": 0.5553,
+      "step": 20865
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0840150093999495e-05,
+      "loss": 0.6451,
+      "step": 20866
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0839439780212193e-05,
+      "loss": 0.5469,
+      "step": 20867
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0838729462159454e-05,
+      "loss": 0.5889,
+      "step": 20868
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0838019139844883e-05,
+      "loss": 0.6057,
+      "step": 20869
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.083730881327209e-05,
+      "loss": 0.5679,
+      "step": 20870
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.083659848244469e-05,
+      "loss": 0.6635,
+      "step": 20871
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0835888147366282e-05,
+      "loss": 0.6177,
+      "step": 20872
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0835177808040485e-05,
+      "loss": 0.5556,
+      "step": 20873
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0834467464470905e-05,
+      "loss": 0.5953,
+      "step": 20874
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0833757116661152e-05,
+      "loss": 0.5755,
+      "step": 20875
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0833046764614836e-05,
+      "loss": 0.5262,
+      "step": 20876
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0832336408335563e-05,
+      "loss": 0.5389,
+      "step": 20877
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0831626047826944e-05,
+      "loss": 0.6827,
+      "step": 20878
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0830915683092593e-05,
+      "loss": 0.5774,
+      "step": 20879
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0830205314136114e-05,
+      "loss": 0.6684,
+      "step": 20880
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.082949494096112e-05,
+      "loss": 0.5964,
+      "step": 20881
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.082878456357122e-05,
+      "loss": 0.6038,
+      "step": 20882
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.082807418197002e-05,
+      "loss": 0.6353,
+      "step": 20883
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0827363796161134e-05,
+      "loss": 0.6536,
+      "step": 20884
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.082665340614817e-05,
+      "loss": 0.6236,
+      "step": 20885
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0825943011934737e-05,
+      "loss": 0.6639,
+      "step": 20886
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0825232613524448e-05,
+      "loss": 0.5984,
+      "step": 20887
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0824522210920909e-05,
+      "loss": 0.6097,
+      "step": 20888
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0823811804127731e-05,
+      "loss": 0.5642,
+      "step": 20889
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0823101393148522e-05,
+      "loss": 0.5811,
+      "step": 20890
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0822390977986896e-05,
+      "loss": 0.6205,
+      "step": 20891
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0821680558646464e-05,
+      "loss": 0.5901,
+      "step": 20892
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0820970135130827e-05,
+      "loss": 0.5862,
+      "step": 20893
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0820259707443601e-05,
+      "loss": 0.6019,
+      "step": 20894
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.08195492755884e-05,
+      "loss": 0.6071,
+      "step": 20895
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0818838839568825e-05,
+      "loss": 0.6322,
+      "step": 20896
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0818128399388488e-05,
+      "loss": 0.6104,
+      "step": 20897
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0817417955051004e-05,
+      "loss": 0.5972,
+      "step": 20898
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0816707506559982e-05,
+      "loss": 0.6218,
+      "step": 20899
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0815997053919025e-05,
+      "loss": 0.5765,
+      "step": 20900
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0815286597131753e-05,
+      "loss": 0.6264,
+      "step": 20901
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0814576136201769e-05,
+      "loss": 0.5536,
+      "step": 20902
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0813865671132684e-05,
+      "loss": 0.5754,
+      "step": 20903
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.081315520192811e-05,
+      "loss": 0.6462,
+      "step": 20904
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0812444728591656e-05,
+      "loss": 0.6188,
+      "step": 20905
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0811734251126932e-05,
+      "loss": 0.5672,
+      "step": 20906
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.081102376953755e-05,
+      "loss": 0.563,
+      "step": 20907
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0810313283827116e-05,
+      "loss": 0.6335,
+      "step": 20908
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0809602793999245e-05,
+      "loss": 0.5805,
+      "step": 20909
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0808892300057544e-05,
+      "loss": 0.6527,
+      "step": 20910
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0808181802005627e-05,
+      "loss": 0.6212,
+      "step": 20911
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0807471299847099e-05,
+      "loss": 0.5998,
+      "step": 20912
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0806760793585573e-05,
+      "loss": 0.573,
+      "step": 20913
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0806050283224661e-05,
+      "loss": 0.5779,
+      "step": 20914
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0805339768767968e-05,
+      "loss": 0.5785,
+      "step": 20915
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.080462925021911e-05,
+      "loss": 0.5579,
+      "step": 20916
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0803918727581697e-05,
+      "loss": 0.6316,
+      "step": 20917
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0803208200859334e-05,
+      "loss": 0.5972,
+      "step": 20918
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0802497670055639e-05,
+      "loss": 0.6128,
+      "step": 20919
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0801787135174215e-05,
+      "loss": 0.5747,
+      "step": 20920
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0801076596218675e-05,
+      "loss": 0.6461,
+      "step": 20921
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0800366053192635e-05,
+      "loss": 0.6076,
+      "step": 20922
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0799655506099695e-05,
+      "loss": 0.6367,
+      "step": 20923
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0798944954943474e-05,
+      "loss": 0.567,
+      "step": 20924
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0798234399727581e-05,
+      "loss": 0.5993,
+      "step": 20925
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0797523840455622e-05,
+      "loss": 0.5699,
+      "step": 20926
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.079681327713121e-05,
+      "loss": 0.6207,
+      "step": 20927
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.079610270975796e-05,
+      "loss": 0.6703,
+      "step": 20928
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0795392138339477e-05,
+      "loss": 0.6226,
+      "step": 20929
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0794681562879377e-05,
+      "loss": 0.5587,
+      "step": 20930
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0793970983381264e-05,
+      "loss": 0.6036,
+      "step": 20931
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0793260399848751e-05,
+      "loss": 0.6383,
+      "step": 20932
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0792549812285453e-05,
+      "loss": 0.6212,
+      "step": 20933
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0791839220694974e-05,
+      "loss": 0.5707,
+      "step": 20934
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0791128625080929e-05,
+      "loss": 0.6407,
+      "step": 20935
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.079041802544693e-05,
+      "loss": 0.6013,
+      "step": 20936
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0789707421796582e-05,
+      "loss": 0.6038,
+      "step": 20937
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0788996814133502e-05,
+      "loss": 0.6083,
+      "step": 20938
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0788286202461298e-05,
+      "loss": 0.6091,
+      "step": 20939
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0787575586783578e-05,
+      "loss": 0.6037,
+      "step": 20940
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.078686496710396e-05,
+      "loss": 0.6033,
+      "step": 20941
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0786154343426047e-05,
+      "loss": 0.6131,
+      "step": 20942
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0785443715753455e-05,
+      "loss": 0.6248,
+      "step": 20943
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0784733084089796e-05,
+      "loss": 0.548,
+      "step": 20944
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0784022448438674e-05,
+      "loss": 0.6496,
+      "step": 20945
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0783311808803705e-05,
+      "loss": 0.6315,
+      "step": 20946
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0782601165188503e-05,
+      "loss": 0.6106,
+      "step": 20947
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0781890517596672e-05,
+      "loss": 0.5879,
+      "step": 20948
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0781179866031827e-05,
+      "loss": 0.5478,
+      "step": 20949
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0780469210497577e-05,
+      "loss": 0.6113,
+      "step": 20950
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0779758550997537e-05,
+      "loss": 0.5166,
+      "step": 20951
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0779047887535315e-05,
+      "loss": 0.6199,
+      "step": 20952
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0778337220114521e-05,
+      "loss": 0.6129,
+      "step": 20953
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0777626548738766e-05,
+      "loss": 0.5633,
+      "step": 20954
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0776915873411667e-05,
+      "loss": 0.6144,
+      "step": 20955
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0776205194136828e-05,
+      "loss": 0.6236,
+      "step": 20956
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0775494510917865e-05,
+      "loss": 0.5698,
+      "step": 20957
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0774783823758385e-05,
+      "loss": 0.5632,
+      "step": 20958
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0774073132662004e-05,
+      "loss": 0.6125,
+      "step": 20959
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.077336243763233e-05,
+      "loss": 0.6158,
+      "step": 20960
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0772651738672976e-05,
+      "loss": 0.5581,
+      "step": 20961
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.077194103578755e-05,
+      "loss": 0.6105,
+      "step": 20962
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0771230328979667e-05,
+      "loss": 0.5548,
+      "step": 20963
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0770519618252934e-05,
+      "loss": 0.5835,
+      "step": 20964
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0769808903610967e-05,
+      "loss": 0.6555,
+      "step": 20965
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0769098185057378e-05,
+      "loss": 0.6068,
+      "step": 20966
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0768387462595773e-05,
+      "loss": 0.5766,
+      "step": 20967
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0767676736229768e-05,
+      "loss": 0.6407,
+      "step": 20968
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.076696600596297e-05,
+      "loss": 0.5866,
+      "step": 20969
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0766255271798996e-05,
+      "loss": 0.5946,
+      "step": 20970
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0765544533741455e-05,
+      "loss": 0.5913,
+      "step": 20971
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0764833791793954e-05,
+      "loss": 0.6462,
+      "step": 20972
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0764123045960111e-05,
+      "loss": 0.6165,
+      "step": 20973
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0763412296243536e-05,
+      "loss": 0.6052,
+      "step": 20974
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0762701542647838e-05,
+      "loss": 0.655,
+      "step": 20975
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.076199078517663e-05,
+      "loss": 0.6074,
+      "step": 20976
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0761280023833525e-05,
+      "loss": 0.5365,
+      "step": 20977
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0760569258622134e-05,
+      "loss": 0.5881,
+      "step": 20978
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0759858489546067e-05,
+      "loss": 0.6336,
+      "step": 20979
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0759147716608934e-05,
+      "loss": 0.6147,
+      "step": 20980
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.075843693981435e-05,
+      "loss": 0.6539,
+      "step": 20981
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0757726159165929e-05,
+      "loss": 0.6069,
+      "step": 20982
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0757015374667275e-05,
+      "loss": 0.6058,
+      "step": 20983
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0756304586322008e-05,
+      "loss": 0.6376,
+      "step": 20984
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0755593794133734e-05,
+      "loss": 0.6324,
+      "step": 20985
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0754882998106064e-05,
+      "loss": 0.5856,
+      "step": 20986
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0754172198242618e-05,
+      "loss": 0.6155,
+      "step": 20987
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0753461394546998e-05,
+      "loss": 0.6489,
+      "step": 20988
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.075275058702282e-05,
+      "loss": 0.606,
+      "step": 20989
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.07520397756737e-05,
+      "loss": 0.6108,
+      "step": 20990
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0751328960503242e-05,
+      "loss": 0.6117,
+      "step": 20991
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0750618141515061e-05,
+      "loss": 0.5607,
+      "step": 20992
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0749907318712773e-05,
+      "loss": 0.6282,
+      "step": 20993
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0749196492099982e-05,
+      "loss": 0.6257,
+      "step": 20994
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0748485661680308e-05,
+      "loss": 0.6199,
+      "step": 20995
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0747774827457358e-05,
+      "loss": 0.6598,
+      "step": 20996
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0747063989434745e-05,
+      "loss": 0.6133,
+      "step": 20997
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.074635314761608e-05,
+      "loss": 0.6341,
+      "step": 20998
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0745642302004977e-05,
+      "loss": 0.5939,
+      "step": 20999
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0744931452605048e-05,
+      "loss": 0.5953,
+      "step": 21000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0744220599419903e-05,
+      "loss": 0.5986,
+      "step": 21001
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0743509742453155e-05,
+      "loss": 0.6042,
+      "step": 21002
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0742798881708416e-05,
+      "loss": 0.6371,
+      "step": 21003
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0742088017189298e-05,
+      "loss": 0.5388,
+      "step": 21004
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0741377148899414e-05,
+      "loss": 0.6449,
+      "step": 21005
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0740666276842377e-05,
+      "loss": 0.5887,
+      "step": 21006
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0739955401021795e-05,
+      "loss": 0.6003,
+      "step": 21007
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0739244521441286e-05,
+      "loss": 0.6172,
+      "step": 21008
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0738533638104459e-05,
+      "loss": 0.5833,
+      "step": 21009
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0737822751014921e-05,
+      "loss": 0.5916,
+      "step": 21010
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0737111860176294e-05,
+      "loss": 0.5899,
+      "step": 21011
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0736400965592186e-05,
+      "loss": 0.5702,
+      "step": 21012
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0735690067266207e-05,
+      "loss": 0.5948,
+      "step": 21013
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0734979165201973e-05,
+      "loss": 0.6142,
+      "step": 21014
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0734268259403094e-05,
+      "loss": 0.6252,
+      "step": 21015
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0733557349873185e-05,
+      "loss": 0.6184,
+      "step": 21016
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0732846436615854e-05,
+      "loss": 0.6474,
+      "step": 21017
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0732135519634717e-05,
+      "loss": 0.6182,
+      "step": 21018
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0731424598933383e-05,
+      "loss": 0.6031,
+      "step": 21019
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0730713674515468e-05,
+      "loss": 0.6179,
+      "step": 21020
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.073000274638458e-05,
+      "loss": 0.5776,
+      "step": 21021
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0729291814544337e-05,
+      "loss": 0.6388,
+      "step": 21022
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0728580878998349e-05,
+      "loss": 0.5272,
+      "step": 21023
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0727869939750227e-05,
+      "loss": 0.5708,
+      "step": 21024
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0727158996803586e-05,
+      "loss": 0.5766,
+      "step": 21025
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0726448050162036e-05,
+      "loss": 0.6214,
+      "step": 21026
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.072573709982919e-05,
+      "loss": 0.6327,
+      "step": 21027
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.072502614580866e-05,
+      "loss": 0.6057,
+      "step": 21028
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0724315188104064e-05,
+      "loss": 0.5904,
+      "step": 21029
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0723604226719008e-05,
+      "loss": 0.6168,
+      "step": 21030
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0722893261657108e-05,
+      "loss": 0.5656,
+      "step": 21031
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0722182292921973e-05,
+      "loss": 0.5934,
+      "step": 21032
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0721471320517223e-05,
+      "loss": 0.612,
+      "step": 21033
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0720760344446461e-05,
+      "loss": 0.5763,
+      "step": 21034
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0720049364713308e-05,
+      "loss": 0.642,
+      "step": 21035
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0719338381321371e-05,
+      "loss": 0.529,
+      "step": 21036
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0718627394274269e-05,
+      "loss": 0.5897,
+      "step": 21037
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0717916403575605e-05,
+      "loss": 0.5943,
+      "step": 21038
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0717205409229001e-05,
+      "loss": 0.6279,
+      "step": 21039
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0716494411238065e-05,
+      "loss": 0.6187,
+      "step": 21040
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0715783409606412e-05,
+      "loss": 0.5775,
+      "step": 21041
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0715072404337654e-05,
+      "loss": 0.602,
+      "step": 21042
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0714361395435402e-05,
+      "loss": 0.6161,
+      "step": 21043
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0713650382903274e-05,
+      "loss": 0.6083,
+      "step": 21044
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0712939366744877e-05,
+      "loss": 0.527,
+      "step": 21045
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0712228346963825e-05,
+      "loss": 0.6468,
+      "step": 21046
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0711517323563734e-05,
+      "loss": 0.629,
+      "step": 21047
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0710806296548216e-05,
+      "loss": 0.648,
+      "step": 21048
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0710095265920882e-05,
+      "loss": 0.6353,
+      "step": 21049
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0709384231685345e-05,
+      "loss": 0.6418,
+      "step": 21050
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0708673193845219e-05,
+      "loss": 0.5967,
+      "step": 21051
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0707962152404118e-05,
+      "loss": 0.6199,
+      "step": 21052
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0707251107365652e-05,
+      "loss": 0.6419,
+      "step": 21053
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0706540058733437e-05,
+      "loss": 0.6338,
+      "step": 21054
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0705829006511087e-05,
+      "loss": 0.5437,
+      "step": 21055
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0705117950702213e-05,
+      "loss": 0.5713,
+      "step": 21056
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0704406891310426e-05,
+      "loss": 0.559,
+      "step": 21057
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0703695828339344e-05,
+      "loss": 0.5763,
+      "step": 21058
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0702984761792576e-05,
+      "loss": 0.5563,
+      "step": 21059
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0702273691673737e-05,
+      "loss": 0.6604,
+      "step": 21060
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0701562617986438e-05,
+      "loss": 0.6581,
+      "step": 21061
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0700851540734294e-05,
+      "loss": 0.6709,
+      "step": 21062
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.070014045992092e-05,
+      "loss": 0.5651,
+      "step": 21063
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0699429375549925e-05,
+      "loss": 0.6565,
+      "step": 21064
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0698718287624925e-05,
+      "loss": 0.6158,
+      "step": 21065
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0698007196149533e-05,
+      "loss": 0.6183,
+      "step": 21066
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0697296101127364e-05,
+      "loss": 0.6155,
+      "step": 21067
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0696585002562028e-05,
+      "loss": 0.5764,
+      "step": 21068
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.069587390045714e-05,
+      "loss": 0.6623,
+      "step": 21069
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.069516279481631e-05,
+      "loss": 0.5631,
+      "step": 21070
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0694451685643157e-05,
+      "loss": 0.6064,
+      "step": 21071
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0693740572941292e-05,
+      "loss": 0.5471,
+      "step": 21072
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0693029456714326e-05,
+      "loss": 0.5865,
+      "step": 21073
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0692318336965876e-05,
+      "loss": 0.6333,
+      "step": 21074
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0691607213699555e-05,
+      "loss": 0.5576,
+      "step": 21075
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0690896086918972e-05,
+      "loss": 0.6099,
+      "step": 21076
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0690184956627748e-05,
+      "loss": 0.5611,
+      "step": 21077
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0689473822829487e-05,
+      "loss": 0.5876,
+      "step": 21078
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0688762685527811e-05,
+      "loss": 0.5585,
+      "step": 21079
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0688051544726327e-05,
+      "loss": 0.6182,
+      "step": 21080
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0687340400428652e-05,
+      "loss": 0.6223,
+      "step": 21081
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0686629252638403e-05,
+      "loss": 0.5707,
+      "step": 21082
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0685918101359184e-05,
+      "loss": 0.6297,
+      "step": 21083
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.068520694659462e-05,
+      "loss": 0.5557,
+      "step": 21084
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0684495788348316e-05,
+      "loss": 0.5883,
+      "step": 21085
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0683784626623887e-05,
+      "loss": 0.6151,
+      "step": 21086
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0683073461424953e-05,
+      "loss": 0.6175,
+      "step": 21087
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0682362292755118e-05,
+      "loss": 0.608,
+      "step": 21088
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0681651120618002e-05,
+      "loss": 0.5185,
+      "step": 21089
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0680939945017217e-05,
+      "loss": 0.6677,
+      "step": 21090
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0680228765956375e-05,
+      "loss": 0.6382,
+      "step": 21091
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0679517583439092e-05,
+      "loss": 0.6336,
+      "step": 21092
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0678806397468984e-05,
+      "loss": 0.5662,
+      "step": 21093
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0678095208049659e-05,
+      "loss": 0.5949,
+      "step": 21094
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0677384015184738e-05,
+      "loss": 0.5577,
+      "step": 21095
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0676672818877827e-05,
+      "loss": 0.6256,
+      "step": 21096
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0675961619132542e-05,
+      "loss": 0.6734,
+      "step": 21097
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.06752504159525e-05,
+      "loss": 0.5499,
+      "step": 21098
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0674539209341312e-05,
+      "loss": 0.5894,
+      "step": 21099
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0673827999302593e-05,
+      "loss": 0.6187,
+      "step": 21100
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0673116785839958e-05,
+      "loss": 0.5789,
+      "step": 21101
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0672405568957016e-05,
+      "loss": 0.5213,
+      "step": 21102
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0671694348657386e-05,
+      "loss": 0.6242,
+      "step": 21103
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0670983124944683e-05,
+      "loss": 0.5865,
+      "step": 21104
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0670271897822516e-05,
+      "loss": 0.5671,
+      "step": 21105
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0669560667294502e-05,
+      "loss": 0.5775,
+      "step": 21106
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0668849433364252e-05,
+      "loss": 0.622,
+      "step": 21107
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0668138196035382e-05,
+      "loss": 0.5591,
+      "step": 21108
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.066742695531151e-05,
+      "loss": 0.5844,
+      "step": 21109
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0666715711196242e-05,
+      "loss": 0.5967,
+      "step": 21110
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0666004463693198e-05,
+      "loss": 0.62,
+      "step": 21111
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0665293212805991e-05,
+      "loss": 0.6265,
+      "step": 21112
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0664581958538234e-05,
+      "loss": 0.6207,
+      "step": 21113
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0663870700893542e-05,
+      "loss": 0.6378,
+      "step": 21114
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0663159439875525e-05,
+      "loss": 0.5836,
+      "step": 21115
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0662448175487804e-05,
+      "loss": 0.6042,
+      "step": 21116
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0661736907733993e-05,
+      "loss": 0.5871,
+      "step": 21117
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0661025636617697e-05,
+      "loss": 0.6098,
+      "step": 21118
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0660314362142538e-05,
+      "loss": 0.6038,
+      "step": 21119
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0659603084312132e-05,
+      "loss": 0.5707,
+      "step": 21120
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0658891803130082e-05,
+      "loss": 0.5707,
+      "step": 21121
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0658180518600016e-05,
+      "loss": 0.6278,
+      "step": 21122
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0657469230725538e-05,
+      "loss": 0.6154,
+      "step": 21123
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.065675793951027e-05,
+      "loss": 0.6114,
+      "step": 21124
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.065604664495782e-05,
+      "loss": 0.6026,
+      "step": 21125
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0655335347071806e-05,
+      "loss": 0.5437,
+      "step": 21126
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.065462404585584e-05,
+      "loss": 0.6333,
+      "step": 21127
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.065391274131354e-05,
+      "loss": 0.6458,
+      "step": 21128
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0653201433448515e-05,
+      "loss": 0.5976,
+      "step": 21129
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0652490122264382e-05,
+      "loss": 0.6176,
+      "step": 21130
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0651778807764759e-05,
+      "loss": 0.5771,
+      "step": 21131
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0651067489953256e-05,
+      "loss": 0.6051,
+      "step": 21132
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0650356168833487e-05,
+      "loss": 0.5545,
+      "step": 21133
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0649644844409069e-05,
+      "loss": 0.5857,
+      "step": 21134
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0648933516683616e-05,
+      "loss": 0.5933,
+      "step": 21135
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.064822218566074e-05,
+      "loss": 0.5788,
+      "step": 21136
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0647510851344057e-05,
+      "loss": 0.605,
+      "step": 21137
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.064679951373718e-05,
+      "loss": 0.6531,
+      "step": 21138
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0646088172843729e-05,
+      "loss": 0.6871,
+      "step": 21139
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0645376828667314e-05,
+      "loss": 0.5962,
+      "step": 21140
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0644665481211551e-05,
+      "loss": 0.6602,
+      "step": 21141
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0643954130480051e-05,
+      "loss": 0.6004,
+      "step": 21142
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0643242776476432e-05,
+      "loss": 0.5846,
+      "step": 21143
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0642531419204312e-05,
+      "loss": 0.5711,
+      "step": 21144
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0641820058667296e-05,
+      "loss": 0.5786,
+      "step": 21145
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0641108694869007e-05,
+      "loss": 0.6711,
+      "step": 21146
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.064039732781306e-05,
+      "loss": 0.6616,
+      "step": 21147
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.063968595750306e-05,
+      "loss": 0.5801,
+      "step": 21148
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0638974583942632e-05,
+      "loss": 0.6137,
+      "step": 21149
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0638263207135388e-05,
+      "loss": 0.602,
+      "step": 21150
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0637551827084939e-05,
+      "loss": 0.5802,
+      "step": 21151
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0636840443794907e-05,
+      "loss": 0.5519,
+      "step": 21152
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0636129057268898e-05,
+      "loss": 0.5968,
+      "step": 21153
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0635417667510534e-05,
+      "loss": 0.6305,
+      "step": 21154
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0634706274523425e-05,
+      "loss": 0.5917,
+      "step": 21155
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.063399487831119e-05,
+      "loss": 0.6242,
+      "step": 21156
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0633283478877436e-05,
+      "loss": 0.6319,
+      "step": 21157
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.063257207622579e-05,
+      "loss": 0.5897,
+      "step": 21158
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0631860670359853e-05,
+      "loss": 0.6329,
+      "step": 21159
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0631149261283252e-05,
+      "loss": 0.6293,
+      "step": 21160
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0630437848999594e-05,
+      "loss": 0.6221,
+      "step": 21161
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0629726433512494e-05,
+      "loss": 0.5813,
+      "step": 21162
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0629015014825577e-05,
+      "loss": 0.6054,
+      "step": 21163
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0628303592942446e-05,
+      "loss": 0.6507,
+      "step": 21164
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0627592167866719e-05,
+      "loss": 0.6438,
+      "step": 21165
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0626880739602017e-05,
+      "loss": 0.6486,
+      "step": 21166
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0626169308151945e-05,
+      "loss": 0.5939,
+      "step": 21167
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.062545787352013e-05,
+      "loss": 0.616,
+      "step": 21168
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0624746435710175e-05,
+      "loss": 0.5738,
+      "step": 21169
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.06240349947257e-05,
+      "loss": 0.5819,
+      "step": 21170
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0623323550570323e-05,
+      "loss": 0.5706,
+      "step": 21171
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0622612103247656e-05,
+      "loss": 0.6352,
+      "step": 21172
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0621900652761315e-05,
+      "loss": 0.609,
+      "step": 21173
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0621189199114914e-05,
+      "loss": 0.5719,
+      "step": 21174
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.062047774231207e-05,
+      "loss": 0.5613,
+      "step": 21175
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0619766282356395e-05,
+      "loss": 0.6111,
+      "step": 21176
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0619054819251507e-05,
+      "loss": 0.6086,
+      "step": 21177
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0618343353001019e-05,
+      "loss": 0.6292,
+      "step": 21178
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.061763188360855e-05,
+      "loss": 0.5962,
+      "step": 21179
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.061692041107771e-05,
+      "loss": 0.589,
+      "step": 21180
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0616208935412116e-05,
+      "loss": 0.6069,
+      "step": 21181
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0615497456615388e-05,
+      "loss": 0.5845,
+      "step": 21182
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.061478597469113e-05,
+      "loss": 0.6234,
+      "step": 21183
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.061407448964297e-05,
+      "loss": 0.6029,
+      "step": 21184
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0613363001474518e-05,
+      "loss": 0.6448,
+      "step": 21185
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0612651510189386e-05,
+      "loss": 0.5455,
+      "step": 21186
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0611940015791195e-05,
+      "loss": 0.5837,
+      "step": 21187
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0611228518283556e-05,
+      "loss": 0.5892,
+      "step": 21188
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0610517017670086e-05,
+      "loss": 0.5528,
+      "step": 21189
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0609805513954402e-05,
+      "loss": 0.6189,
+      "step": 21190
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0609094007140115e-05,
+      "loss": 0.6351,
+      "step": 21191
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0608382497230845e-05,
+      "loss": 0.5833,
+      "step": 21192
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0607670984230202e-05,
+      "loss": 0.6156,
+      "step": 21193
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0606959468141808e-05,
+      "loss": 0.6738,
+      "step": 21194
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0606247948969275e-05,
+      "loss": 0.6047,
+      "step": 21195
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0605536426716219e-05,
+      "loss": 0.6511,
+      "step": 21196
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0604824901386253e-05,
+      "loss": 0.6131,
+      "step": 21197
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0604113372982997e-05,
+      "loss": 0.6159,
+      "step": 21198
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0603401841510062e-05,
+      "loss": 0.5972,
+      "step": 21199
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0602690306971066e-05,
+      "loss": 0.6589,
+      "step": 21200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0601978769369626e-05,
+      "loss": 0.5898,
+      "step": 21201
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0601267228709351e-05,
+      "loss": 0.6612,
+      "step": 21202
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0600555684993862e-05,
+      "loss": 0.621,
+      "step": 21203
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0599844138226778e-05,
+      "loss": 0.6138,
+      "step": 21204
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0599132588411707e-05,
+      "loss": 0.6465,
+      "step": 21205
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.059842103555227e-05,
+      "loss": 0.6298,
+      "step": 21206
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0597709479652078e-05,
+      "loss": 0.639,
+      "step": 21207
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0596997920714751e-05,
+      "loss": 0.6377,
+      "step": 21208
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0596286358743904e-05,
+      "loss": 0.5936,
+      "step": 21209
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0595574793743146e-05,
+      "loss": 0.6598,
+      "step": 21210
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0594863225716101e-05,
+      "loss": 0.6216,
+      "step": 21211
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0594151654666385e-05,
+      "loss": 0.5514,
+      "step": 21212
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0593440080597608e-05,
+      "loss": 0.5795,
+      "step": 21213
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0592728503513388e-05,
+      "loss": 0.613,
+      "step": 21214
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0592016923417342e-05,
+      "loss": 0.5803,
+      "step": 21215
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0591305340313081e-05,
+      "loss": 0.6099,
+      "step": 21216
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0590593754204229e-05,
+      "loss": 0.5468,
+      "step": 21217
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0589882165094395e-05,
+      "loss": 0.6308,
+      "step": 21218
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0589170572987194e-05,
+      "loss": 0.624,
+      "step": 21219
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.058845897788625e-05,
+      "loss": 0.5947,
+      "step": 21220
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0587747379795168e-05,
+      "loss": 0.5357,
+      "step": 21221
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0587035778717575e-05,
+      "loss": 0.5628,
+      "step": 21222
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0586324174657079e-05,
+      "loss": 0.6015,
+      "step": 21223
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0585612567617297e-05,
+      "loss": 0.5789,
+      "step": 21224
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0584900957601847e-05,
+      "loss": 0.6463,
+      "step": 21225
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0584189344614344e-05,
+      "loss": 0.5902,
+      "step": 21226
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0583477728658402e-05,
+      "loss": 0.6173,
+      "step": 21227
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0582766109737643e-05,
+      "loss": 0.6741,
+      "step": 21228
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0582054487855676e-05,
+      "loss": 0.6384,
+      "step": 21229
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0581342863016118e-05,
+      "loss": 0.6042,
+      "step": 21230
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0580631235222589e-05,
+      "loss": 0.5693,
+      "step": 21231
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0579919604478705e-05,
+      "loss": 0.5923,
+      "step": 21232
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0579207970788077e-05,
+      "loss": 0.603,
+      "step": 21233
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0578496334154324e-05,
+      "loss": 0.5205,
+      "step": 21234
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.057778469458106e-05,
+      "loss": 0.6354,
+      "step": 21235
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0577073052071906e-05,
+      "loss": 0.6038,
+      "step": 21236
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0576361406630471e-05,
+      "loss": 0.6163,
+      "step": 21237
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0575649758260377e-05,
+      "loss": 0.6024,
+      "step": 21238
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0574938106965238e-05,
+      "loss": 0.5659,
+      "step": 21239
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.057422645274867e-05,
+      "loss": 0.6032,
+      "step": 21240
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0573514795614288e-05,
+      "loss": 0.6379,
+      "step": 21241
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0572803135565712e-05,
+      "loss": 0.5717,
+      "step": 21242
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0572091472606553e-05,
+      "loss": 0.6338,
+      "step": 21243
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0571379806740431e-05,
+      "loss": 0.6292,
+      "step": 21244
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.057066813797096e-05,
+      "loss": 0.6503,
+      "step": 21245
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0569956466301758e-05,
+      "loss": 0.6302,
+      "step": 21246
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0569244791736442e-05,
+      "loss": 0.6076,
+      "step": 21247
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0568533114278623e-05,
+      "loss": 0.6046,
+      "step": 21248
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0567821433931923e-05,
+      "loss": 0.6264,
+      "step": 21249
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0567109750699955e-05,
+      "loss": 0.6155,
+      "step": 21250
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0566398064586334e-05,
+      "loss": 0.6184,
+      "step": 21251
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0565686375594683e-05,
+      "loss": 0.5659,
+      "step": 21252
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0564974683728612e-05,
+      "loss": 0.5955,
+      "step": 21253
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.056426298899174e-05,
+      "loss": 0.5575,
+      "step": 21254
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0563551291387681e-05,
+      "loss": 0.6147,
+      "step": 21255
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0562839590920054e-05,
+      "loss": 0.5632,
+      "step": 21256
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0562127887592473e-05,
+      "loss": 0.5912,
+      "step": 21257
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0561416181408557e-05,
+      "loss": 0.5864,
+      "step": 21258
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0560704472371919e-05,
+      "loss": 0.5472,
+      "step": 21259
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0559992760486179e-05,
+      "loss": 0.6304,
+      "step": 21260
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0559281045754952e-05,
+      "loss": 0.6645,
+      "step": 21261
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0558569328181851e-05,
+      "loss": 0.5863,
+      "step": 21262
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.05578576077705e-05,
+      "loss": 0.5767,
+      "step": 21263
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0557145884524508e-05,
+      "loss": 0.6796,
+      "step": 21264
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0556434158447495e-05,
+      "loss": 0.5819,
+      "step": 21265
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.055572242954308e-05,
+      "loss": 0.65,
+      "step": 21266
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0555010697814874e-05,
+      "loss": 0.6047,
+      "step": 21267
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0554298963266496e-05,
+      "loss": 0.6201,
+      "step": 21268
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0553587225901566e-05,
+      "loss": 0.5891,
+      "step": 21269
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0552875485723693e-05,
+      "loss": 0.5461,
+      "step": 21270
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0552163742736499e-05,
+      "loss": 0.6036,
+      "step": 21271
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.05514519969436e-05,
+      "loss": 0.6282,
+      "step": 21272
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0550740248348612e-05,
+      "loss": 0.6063,
+      "step": 21273
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.055002849695515e-05,
+      "loss": 0.624,
+      "step": 21274
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0549316742766833e-05,
+      "loss": 0.5545,
+      "step": 21275
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0548604985787276e-05,
+      "loss": 0.5657,
+      "step": 21276
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0547893226020098e-05,
+      "loss": 0.6083,
+      "step": 21277
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0547181463468915e-05,
+      "loss": 0.5788,
+      "step": 21278
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.054646969813734e-05,
+      "loss": 0.663,
+      "step": 21279
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0545757930028992e-05,
+      "loss": 0.5828,
+      "step": 21280
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.054504615914749e-05,
+      "loss": 0.542,
+      "step": 21281
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0544334385496449e-05,
+      "loss": 0.6639,
+      "step": 21282
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0543622609079485e-05,
+      "loss": 0.6221,
+      "step": 21283
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0542910829900214e-05,
+      "loss": 0.6929,
+      "step": 21284
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0542199047962256e-05,
+      "loss": 0.5885,
+      "step": 21285
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0541487263269226e-05,
+      "loss": 0.6226,
+      "step": 21286
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0540775475824738e-05,
+      "loss": 0.6527,
+      "step": 21287
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0540063685632413e-05,
+      "loss": 0.6056,
+      "step": 21288
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0539351892695865e-05,
+      "loss": 0.5412,
+      "step": 21289
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0538640097018717e-05,
+      "loss": 0.5805,
+      "step": 21290
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0537928298604577e-05,
+      "loss": 0.5767,
+      "step": 21291
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0537216497457068e-05,
+      "loss": 0.6048,
+      "step": 21292
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0536504693579804e-05,
+      "loss": 0.6049,
+      "step": 21293
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0535792886976398e-05,
+      "loss": 0.6115,
+      "step": 21294
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0535081077650475e-05,
+      "loss": 0.5804,
+      "step": 21295
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.053436926560565e-05,
+      "loss": 0.5983,
+      "step": 21296
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0533657450845535e-05,
+      "loss": 0.6248,
+      "step": 21297
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0532945633373755e-05,
+      "loss": 0.5949,
+      "step": 21298
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0532233813193917e-05,
+      "loss": 0.6454,
+      "step": 21299
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0531521990309642e-05,
+      "loss": 0.6223,
+      "step": 21300
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0530810164724554e-05,
+      "loss": 0.5759,
+      "step": 21301
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.053009833644226e-05,
+      "loss": 0.531,
+      "step": 21302
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.052938650546638e-05,
+      "loss": 0.5953,
+      "step": 21303
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0528674671800538e-05,
+      "loss": 0.594,
+      "step": 21304
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.052796283544834e-05,
+      "loss": 0.6236,
+      "step": 21305
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0527250996413411e-05,
+      "loss": 0.561,
+      "step": 21306
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0526539154699362e-05,
+      "loss": 0.642,
+      "step": 21307
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0525827310309814e-05,
+      "loss": 0.6123,
+      "step": 21308
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0525115463248388e-05,
+      "loss": 0.5641,
+      "step": 21309
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0524403613518691e-05,
+      "loss": 0.6076,
+      "step": 21310
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0523691761124351e-05,
+      "loss": 0.5889,
+      "step": 21311
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0522979906068977e-05,
+      "loss": 0.6589,
+      "step": 21312
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0522268048356188e-05,
+      "loss": 0.5733,
+      "step": 21313
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0521556187989604e-05,
+      "loss": 0.5904,
+      "step": 21314
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0520844324972838e-05,
+      "loss": 0.6003,
+      "step": 21315
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0520132459309511e-05,
+      "loss": 0.6266,
+      "step": 21316
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.051942059100324e-05,
+      "loss": 0.6325,
+      "step": 21317
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0518708720057638e-05,
+      "loss": 0.6477,
+      "step": 21318
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0517996846476324e-05,
+      "loss": 0.6406,
+      "step": 21319
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.051728497026292e-05,
+      "loss": 0.6069,
+      "step": 21320
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0516573091421038e-05,
+      "loss": 0.5915,
+      "step": 21321
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0515861209954296e-05,
+      "loss": 0.6028,
+      "step": 21322
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0515149325866314e-05,
+      "loss": 0.6015,
+      "step": 21323
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0514437439160705e-05,
+      "loss": 0.565,
+      "step": 21324
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0513725549841091e-05,
+      "loss": 0.634,
+      "step": 21325
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0513013657911086e-05,
+      "loss": 0.6222,
+      "step": 21326
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0512301763374309e-05,
+      "loss": 0.6358,
+      "step": 21327
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0511589866234377e-05,
+      "loss": 0.6667,
+      "step": 21328
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0510877966494906e-05,
+      "loss": 0.6129,
+      "step": 21329
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0510166064159515e-05,
+      "loss": 0.6516,
+      "step": 21330
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.050945415923182e-05,
+      "loss": 0.5744,
+      "step": 21331
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.050874225171544e-05,
+      "loss": 0.6264,
+      "step": 21332
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.050803034161399e-05,
+      "loss": 0.6471,
+      "step": 21333
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0507318428931092e-05,
+      "loss": 0.6413,
+      "step": 21334
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.050660651367036e-05,
+      "loss": 0.6114,
+      "step": 21335
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.050589459583541e-05,
+      "loss": 0.5898,
+      "step": 21336
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0505182675429862e-05,
+      "loss": 0.6253,
+      "step": 21337
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0504470752457331e-05,
+      "loss": 0.6202,
+      "step": 21338
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0503758826921442e-05,
+      "loss": 0.6606,
+      "step": 21339
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0503046898825802e-05,
+      "loss": 0.5829,
+      "step": 21340
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0502334968174033e-05,
+      "loss": 0.5933,
+      "step": 21341
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0501623034969756e-05,
+      "loss": 0.6241,
+      "step": 21342
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0500911099216584e-05,
+      "loss": 0.6588,
+      "step": 21343
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0500199160918136e-05,
+      "loss": 0.643,
+      "step": 21344
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0499487220078031e-05,
+      "loss": 0.6103,
+      "step": 21345
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0498775276699882e-05,
+      "loss": 0.6017,
+      "step": 21346
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0498063330787314e-05,
+      "loss": 0.604,
+      "step": 21347
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0497351382343937e-05,
+      "loss": 0.5868,
+      "step": 21348
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0496639431373372e-05,
+      "loss": 0.5824,
+      "step": 21349
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0495927477879237e-05,
+      "loss": 0.6225,
+      "step": 21350
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0495215521865152e-05,
+      "loss": 0.5739,
+      "step": 21351
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0494503563334727e-05,
+      "loss": 0.5977,
+      "step": 21352
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.049379160229159e-05,
+      "loss": 0.5772,
+      "step": 21353
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.049307963873935e-05,
+      "loss": 0.65,
+      "step": 21354
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.049236767268163e-05,
+      "loss": 0.6224,
+      "step": 21355
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0491655704122043e-05,
+      "loss": 0.5943,
+      "step": 21356
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.049094373306421e-05,
+      "loss": 0.5735,
+      "step": 21357
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0490231759511748e-05,
+      "loss": 0.5897,
+      "step": 21358
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0489519783468277e-05,
+      "loss": 0.6329,
+      "step": 21359
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0488807804937408e-05,
+      "loss": 0.577,
+      "step": 21360
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0488095823922768e-05,
+      "loss": 0.5926,
+      "step": 21361
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.048738384042797e-05,
+      "loss": 0.6429,
+      "step": 21362
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.048667185445663e-05,
+      "loss": 0.5941,
+      "step": 21363
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0485959866012367e-05,
+      "loss": 0.617,
+      "step": 21364
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.04852478750988e-05,
+      "loss": 0.6172,
+      "step": 21365
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0484535881719548e-05,
+      "loss": 0.6445,
+      "step": 21366
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0483823885878228e-05,
+      "loss": 0.6576,
+      "step": 21367
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0483111887578456e-05,
+      "loss": 0.5619,
+      "step": 21368
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.048239988682385e-05,
+      "loss": 0.5683,
+      "step": 21369
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0481687883618031e-05,
+      "loss": 0.5906,
+      "step": 21370
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0480975877964612e-05,
+      "loss": 0.6029,
+      "step": 21371
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0480263869867218e-05,
+      "loss": 0.5462,
+      "step": 21372
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0479551859329458e-05,
+      "loss": 0.6017,
+      "step": 21373
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0478839846354957e-05,
+      "loss": 0.6384,
+      "step": 21374
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0478127830947332e-05,
+      "loss": 0.6068,
+      "step": 21375
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0477415813110194e-05,
+      "loss": 0.584,
+      "step": 21376
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0476703792847171e-05,
+      "loss": 0.5572,
+      "step": 21377
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0475991770161875e-05,
+      "loss": 0.5657,
+      "step": 21378
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0475279745057924e-05,
+      "loss": 0.5785,
+      "step": 21379
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0474567717538941e-05,
+      "loss": 0.5646,
+      "step": 21380
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0473855687608538e-05,
+      "loss": 0.5792,
+      "step": 21381
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0473143655270337e-05,
+      "loss": 0.5718,
+      "step": 21382
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0472431620527951e-05,
+      "loss": 0.6279,
+      "step": 21383
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0471719583385005e-05,
+      "loss": 0.6297,
+      "step": 21384
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0471007543845115e-05,
+      "loss": 0.6653,
+      "step": 21385
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0470295501911894e-05,
+      "loss": 0.6697,
+      "step": 21386
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0469583457588962e-05,
+      "loss": 0.5291,
+      "step": 21387
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0468871410879946e-05,
+      "loss": 0.6045,
+      "step": 21388
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0468159361788455e-05,
+      "loss": 0.6454,
+      "step": 21389
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0467447310318106e-05,
+      "loss": 0.6407,
+      "step": 21390
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0466735256472522e-05,
+      "loss": 0.6063,
+      "step": 21391
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.046602320025532e-05,
+      "loss": 0.6045,
+      "step": 21392
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0465311141670117e-05,
+      "loss": 0.567,
+      "step": 21393
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0464599080720533e-05,
+      "loss": 0.5601,
+      "step": 21394
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0463887017410182e-05,
+      "loss": 0.6176,
+      "step": 21395
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0463174951742686e-05,
+      "loss": 0.6426,
+      "step": 21396
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0462462883721664e-05,
+      "loss": 0.674,
+      "step": 21397
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0461750813350733e-05,
+      "loss": 0.5728,
+      "step": 21398
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.046103874063351e-05,
+      "loss": 0.5932,
+      "step": 21399
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0460326665573613e-05,
+      "loss": 0.5999,
+      "step": 21400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0459614588174663e-05,
+      "loss": 0.6409,
+      "step": 21401
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0458902508440276e-05,
+      "loss": 0.6156,
+      "step": 21402
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0458190426374068e-05,
+      "loss": 0.6519,
+      "step": 21403
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0457478341979665e-05,
+      "loss": 0.579,
+      "step": 21404
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.045676625526068e-05,
+      "loss": 0.6323,
+      "step": 21405
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0456054166220728e-05,
+      "loss": 0.5972,
+      "step": 21406
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0455342074863438e-05,
+      "loss": 0.6385,
+      "step": 21407
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0454629981192416e-05,
+      "loss": 0.5983,
+      "step": 21408
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0453917885211287e-05,
+      "loss": 0.5455,
+      "step": 21409
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.045320578692367e-05,
+      "loss": 0.5763,
+      "step": 21410
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.045249368633318e-05,
+      "loss": 0.6189,
+      "step": 21411
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0451781583443438e-05,
+      "loss": 0.627,
+      "step": 21412
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.045106947825806e-05,
+      "loss": 0.6015,
+      "step": 21413
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0450357370780666e-05,
+      "loss": 0.6281,
+      "step": 21414
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0449645261014879e-05,
+      "loss": 0.616,
+      "step": 21415
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0448933148964306e-05,
+      "loss": 0.5645,
+      "step": 21416
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0448221034632577e-05,
+      "loss": 0.5909,
+      "step": 21417
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0447508918023303e-05,
+      "loss": 0.6738,
+      "step": 21418
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0446796799140104e-05,
+      "loss": 0.6302,
+      "step": 21419
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0446084677986604e-05,
+      "loss": 0.6438,
+      "step": 21420
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0445372554566413e-05,
+      "loss": 0.6485,
+      "step": 21421
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0444660428883153e-05,
+      "loss": 0.5965,
+      "step": 21422
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0443948300940447e-05,
+      "loss": 0.6427,
+      "step": 21423
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0443236170741908e-05,
+      "loss": 0.6249,
+      "step": 21424
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0442524038291155e-05,
+      "loss": 0.6009,
+      "step": 21425
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.044181190359181e-05,
+      "loss": 0.6166,
+      "step": 21426
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0441099766647489e-05,
+      "loss": 0.6544,
+      "step": 21427
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.044038762746181e-05,
+      "loss": 0.5762,
+      "step": 21428
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0439675486038394e-05,
+      "loss": 0.6777,
+      "step": 21429
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0438963342380856e-05,
+      "loss": 0.6328,
+      "step": 21430
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0438251196492818e-05,
+      "loss": 0.6687,
+      "step": 21431
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0437539048377897e-05,
+      "loss": 0.6283,
+      "step": 21432
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0436826898039712e-05,
+      "loss": 0.5645,
+      "step": 21433
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0436114745481882e-05,
+      "loss": 0.6589,
+      "step": 21434
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0435402590708023e-05,
+      "loss": 0.6133,
+      "step": 21435
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.043469043372176e-05,
+      "loss": 0.6001,
+      "step": 21436
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0433978274526704e-05,
+      "loss": 0.5873,
+      "step": 21437
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0433266113126477e-05,
+      "loss": 0.6101,
+      "step": 21438
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.04325539495247e-05,
+      "loss": 0.6398,
+      "step": 21439
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.043184178372499e-05,
+      "loss": 0.6287,
+      "step": 21440
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0431129615730962e-05,
+      "loss": 0.6121,
+      "step": 21441
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0430417445546241e-05,
+      "loss": 0.6235,
+      "step": 21442
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0429705273174444e-05,
+      "loss": 0.5565,
+      "step": 21443
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0428993098619186e-05,
+      "loss": 0.6512,
+      "step": 21444
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.042828092188409e-05,
+      "loss": 0.5659,
+      "step": 21445
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0427568742972774e-05,
+      "loss": 0.6389,
+      "step": 21446
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0426856561888855e-05,
+      "loss": 0.5549,
+      "step": 21447
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0426144378635952e-05,
+      "loss": 0.5896,
+      "step": 21448
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0425432193217686e-05,
+      "loss": 0.5891,
+      "step": 21449
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0424720005637677e-05,
+      "loss": 0.6294,
+      "step": 21450
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0424007815899534e-05,
+      "loss": 0.6347,
+      "step": 21451
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0423295624006889e-05,
+      "loss": 0.5835,
+      "step": 21452
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0422583429963352e-05,
+      "loss": 0.5875,
+      "step": 21453
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0421871233772547e-05,
+      "loss": 0.5212,
+      "step": 21454
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.042115903543809e-05,
+      "loss": 0.561,
+      "step": 21455
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.04204468349636e-05,
+      "loss": 0.578,
+      "step": 21456
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0419734632352697e-05,
+      "loss": 0.6162,
+      "step": 21457
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0419022427608998e-05,
+      "loss": 0.5919,
+      "step": 21458
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0418310220736126e-05,
+      "loss": 0.569,
+      "step": 21459
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0417598011737693e-05,
+      "loss": 0.5784,
+      "step": 21460
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0416885800617326e-05,
+      "loss": 0.6435,
+      "step": 21461
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.041617358737864e-05,
+      "loss": 0.6157,
+      "step": 21462
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0415461372025253e-05,
+      "loss": 0.5954,
+      "step": 21463
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0414749154560783e-05,
+      "loss": 0.5513,
+      "step": 21464
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0414036934988854e-05,
+      "loss": 0.6549,
+      "step": 21465
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0413324713313082e-05,
+      "loss": 0.591,
+      "step": 21466
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0412612489537087e-05,
+      "loss": 0.5906,
+      "step": 21467
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0411900263664485e-05,
+      "loss": 0.5882,
+      "step": 21468
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.04111880356989e-05,
+      "loss": 0.5416,
+      "step": 21469
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0410475805643944e-05,
+      "loss": 0.5835,
+      "step": 21470
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0409763573503242e-05,
+      "loss": 0.5918,
+      "step": 21471
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0409051339280412e-05,
+      "loss": 0.6096,
+      "step": 21472
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.040833910297907e-05,
+      "loss": 0.6306,
+      "step": 21473
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0407626864602843e-05,
+      "loss": 0.5837,
+      "step": 21474
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.040691462415534e-05,
+      "loss": 0.5721,
+      "step": 21475
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0406202381640182e-05,
+      "loss": 0.6058,
+      "step": 21476
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0405490137060996e-05,
+      "loss": 0.607,
+      "step": 21477
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0404777890421393e-05,
+      "loss": 0.6131,
+      "step": 21478
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0404065641724997e-05,
+      "loss": 0.5456,
+      "step": 21479
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0403353390975424e-05,
+      "loss": 0.6133,
+      "step": 21480
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0402641138176295e-05,
+      "loss": 0.6043,
+      "step": 21481
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0401928883331227e-05,
+      "loss": 0.5852,
+      "step": 21482
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0401216626443841e-05,
+      "loss": 0.5997,
+      "step": 21483
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0400504367517756e-05,
+      "loss": 0.5936,
+      "step": 21484
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0399792106556592e-05,
+      "loss": 0.4916,
+      "step": 21485
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0399079843563965e-05,
+      "loss": 0.6495,
+      "step": 21486
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.03983675785435e-05,
+      "loss": 0.5498,
+      "step": 21487
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0397655311498812e-05,
+      "loss": 0.6594,
+      "step": 21488
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0396943042433518e-05,
+      "loss": 0.6485,
+      "step": 21489
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0396230771351244e-05,
+      "loss": 0.5656,
+      "step": 21490
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0395518498255603e-05,
+      "loss": 0.5668,
+      "step": 21491
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0394806223150216e-05,
+      "loss": 0.5797,
+      "step": 21492
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0394093946038703e-05,
+      "loss": 0.564,
+      "step": 21493
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0393381666924685e-05,
+      "loss": 0.6078,
+      "step": 21494
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0392669385811778e-05,
+      "loss": 0.5696,
+      "step": 21495
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0391957102703604e-05,
+      "loss": 0.6179,
+      "step": 21496
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.039124481760378e-05,
+      "loss": 0.6035,
+      "step": 21497
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0390532530515926e-05,
+      "loss": 0.5933,
+      "step": 21498
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0389820241443665e-05,
+      "loss": 0.573,
+      "step": 21499
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0389107950390609e-05,
+      "loss": 0.6204,
+      "step": 21500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0388395657360385e-05,
+      "loss": 0.5392,
+      "step": 21501
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0387683362356606e-05,
+      "loss": 0.5891,
+      "step": 21502
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0386971065382896e-05,
+      "loss": 0.5909,
+      "step": 21503
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0386258766442874e-05,
+      "loss": 0.599,
+      "step": 21504
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0385546465540155e-05,
+      "loss": 0.672,
+      "step": 21505
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0384834162678365e-05,
+      "loss": 0.5934,
+      "step": 21506
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0384121857861115e-05,
+      "loss": 0.5975,
+      "step": 21507
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0383409551092037e-05,
+      "loss": 0.5752,
+      "step": 21508
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0382697242374737e-05,
+      "loss": 0.5978,
+      "step": 21509
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0381984931712841e-05,
+      "loss": 0.5746,
+      "step": 21510
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0381272619109968e-05,
+      "loss": 0.6248,
+      "step": 21511
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0380560304569738e-05,
+      "loss": 0.609,
+      "step": 21512
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0379847988095768e-05,
+      "loss": 0.573,
+      "step": 21513
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.037913566969168e-05,
+      "loss": 0.532,
+      "step": 21514
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0378423349361094e-05,
+      "loss": 0.5755,
+      "step": 21515
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0377711027107625e-05,
+      "loss": 0.6625,
+      "step": 21516
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0376998702934898e-05,
+      "loss": 0.6095,
+      "step": 21517
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.037628637684653e-05,
+      "loss": 0.6425,
+      "step": 21518
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.037557404884614e-05,
+      "loss": 0.6096,
+      "step": 21519
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0374861718937348e-05,
+      "loss": 0.5763,
+      "step": 21520
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0374149387123776e-05,
+      "loss": 0.5819,
+      "step": 21521
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0373437053409037e-05,
+      "loss": 0.6218,
+      "step": 21522
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.037272471779676e-05,
+      "loss": 0.6406,
+      "step": 21523
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0372012380290556e-05,
+      "loss": 0.6019,
+      "step": 21524
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.037130004089405e-05,
+      "loss": 0.5612,
+      "step": 21525
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0370587699610859e-05,
+      "loss": 0.6393,
+      "step": 21526
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0369875356444605e-05,
+      "loss": 0.6112,
+      "step": 21527
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0369163011398905e-05,
+      "loss": 0.5937,
+      "step": 21528
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0368450664477378e-05,
+      "loss": 0.6215,
+      "step": 21529
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0367738315683646e-05,
+      "loss": 0.6157,
+      "step": 21530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0367025965021331e-05,
+      "loss": 0.5973,
+      "step": 21531
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0366313612494046e-05,
+      "loss": 0.5797,
+      "step": 21532
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0365601258105414e-05,
+      "loss": 0.5692,
+      "step": 21533
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0364888901859057e-05,
+      "loss": 0.5707,
+      "step": 21534
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.036417654375859e-05,
+      "loss": 0.5765,
+      "step": 21535
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0363464183807638e-05,
+      "loss": 0.6683,
+      "step": 21536
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0362751822009818e-05,
+      "loss": 0.6288,
+      "step": 21537
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0362039458368749e-05,
+      "loss": 0.5937,
+      "step": 21538
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0361327092888051e-05,
+      "loss": 0.6201,
+      "step": 21539
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0360614725571345e-05,
+      "loss": 0.6419,
+      "step": 21540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035990235642225e-05,
+      "loss": 0.585,
+      "step": 21541
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0359189985444384e-05,
+      "loss": 0.6428,
+      "step": 21542
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035847761264137e-05,
+      "loss": 0.6599,
+      "step": 21543
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0357765238016824e-05,
+      "loss": 0.5805,
+      "step": 21544
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0357052861574373e-05,
+      "loss": 0.5802,
+      "step": 21545
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035634048331763e-05,
+      "loss": 0.6161,
+      "step": 21546
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0355628103250216e-05,
+      "loss": 0.6175,
+      "step": 21547
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0354915721375751e-05,
+      "loss": 0.5816,
+      "step": 21548
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0354203337697855e-05,
+      "loss": 0.5682,
+      "step": 21549
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035349095222015e-05,
+      "loss": 0.6075,
+      "step": 21550
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0352778564946251e-05,
+      "loss": 0.5888,
+      "step": 21551
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035206617587978e-05,
+      "loss": 0.6023,
+      "step": 21552
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0351353785024362e-05,
+      "loss": 0.6044,
+      "step": 21553
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.035064139238361e-05,
+      "loss": 0.6526,
+      "step": 21554
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0349928997961148e-05,
+      "loss": 0.6613,
+      "step": 21555
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0349216601760593e-05,
+      "loss": 0.657,
+      "step": 21556
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0348504203785563e-05,
+      "loss": 0.5173,
+      "step": 21557
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0347791804039686e-05,
+      "loss": 0.5612,
+      "step": 21558
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0347079402526574e-05,
+      "loss": 0.5948,
+      "step": 21559
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.034636699924985e-05,
+      "loss": 0.6152,
+      "step": 21560
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0345654594213137e-05,
+      "loss": 0.6111,
+      "step": 21561
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.034494218742005e-05,
+      "loss": 0.5811,
+      "step": 21562
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.034422977887421e-05,
+      "loss": 0.592,
+      "step": 21563
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.034351736857924e-05,
+      "loss": 0.624,
+      "step": 21564
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0342804956538757e-05,
+      "loss": 0.6807,
+      "step": 21565
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0342092542756378e-05,
+      "loss": 0.5909,
+      "step": 21566
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0341380127235732e-05,
+      "loss": 0.6063,
+      "step": 21567
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0340667709980428e-05,
+      "loss": 0.5845,
+      "step": 21568
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0339955290994098e-05,
+      "loss": 0.5928,
+      "step": 21569
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0339242870280348e-05,
+      "loss": 0.6261,
+      "step": 21570
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0338530447842809e-05,
+      "loss": 0.6058,
+      "step": 21571
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.03378180236851e-05,
+      "loss": 0.6187,
+      "step": 21572
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0337105597810836e-05,
+      "loss": 0.6377,
+      "step": 21573
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0336393170223642e-05,
+      "loss": 0.6343,
+      "step": 21574
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0335680740927134e-05,
+      "loss": 0.6513,
+      "step": 21575
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0334968309924933e-05,
+      "loss": 0.6284,
+      "step": 21576
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0334255877220662e-05,
+      "loss": 0.6207,
+      "step": 21577
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0333543442817937e-05,
+      "loss": 0.6384,
+      "step": 21578
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.033283100672038e-05,
+      "loss": 0.5506,
+      "step": 21579
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0332118568931615e-05,
+      "loss": 0.6177,
+      "step": 21580
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0331406129455254e-05,
+      "loss": 0.5845,
+      "step": 21581
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0330693688294924e-05,
+      "loss": 0.5949,
+      "step": 21582
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0329981245454242e-05,
+      "loss": 0.5973,
+      "step": 21583
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0329268800936827e-05,
+      "loss": 0.6022,
+      "step": 21584
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0328556354746302e-05,
+      "loss": 0.5825,
+      "step": 21585
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0327843906886288e-05,
+      "loss": 0.5668,
+      "step": 21586
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.03271314573604e-05,
+      "loss": 0.6173,
+      "step": 21587
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0326419006172263e-05,
+      "loss": 0.5485,
+      "step": 21588
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0325706553325494e-05,
+      "loss": 0.6347,
+      "step": 21589
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0324994098823716e-05,
+      "loss": 0.6007,
+      "step": 21590
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0324281642670548e-05,
+      "loss": 0.569,
+      "step": 21591
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0323569184869609e-05,
+      "loss": 0.5613,
+      "step": 21592
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0322856725424522e-05,
+      "loss": 0.595,
+      "step": 21593
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0322144264338902e-05,
+      "loss": 0.5963,
+      "step": 21594
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0321431801616373e-05,
+      "loss": 0.5999,
+      "step": 21595
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0320719337260559e-05,
+      "loss": 0.6232,
+      "step": 21596
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0320006871275072e-05,
+      "loss": 0.5612,
+      "step": 21597
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0319294403663536e-05,
+      "loss": 0.6476,
+      "step": 21598
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0318581934429574e-05,
+      "loss": 0.5569,
+      "step": 21599
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0317869463576804e-05,
+      "loss": 0.641,
+      "step": 21600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0317156991108848e-05,
+      "loss": 0.6006,
+      "step": 21601
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.031644451702932e-05,
+      "loss": 0.5637,
+      "step": 21602
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0315732041341846e-05,
+      "loss": 0.6064,
+      "step": 21603
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0315019564050047e-05,
+      "loss": 0.569,
+      "step": 21604
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.031430708515754e-05,
+      "loss": 0.5587,
+      "step": 21605
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0313594604667947e-05,
+      "loss": 0.5851,
+      "step": 21606
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0312882122584888e-05,
+      "loss": 0.5521,
+      "step": 21607
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0312169638911981e-05,
+      "loss": 0.5798,
+      "step": 21608
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.031145715365285e-05,
+      "loss": 0.5799,
+      "step": 21609
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0310744666811117e-05,
+      "loss": 0.6137,
+      "step": 21610
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0310032178390397e-05,
+      "loss": 0.6281,
+      "step": 21611
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0309319688394313e-05,
+      "loss": 0.6076,
+      "step": 21612
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0308607196826483e-05,
+      "loss": 0.5536,
+      "step": 21613
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0307894703690531e-05,
+      "loss": 0.5085,
+      "step": 21614
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0307182208990078e-05,
+      "loss": 0.5551,
+      "step": 21615
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0306469712728739e-05,
+      "loss": 0.6464,
+      "step": 21616
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0305757214910139e-05,
+      "loss": 0.5775,
+      "step": 21617
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0305044715537898e-05,
+      "loss": 0.6113,
+      "step": 21618
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0304332214615633e-05,
+      "loss": 0.5829,
+      "step": 21619
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.030361971214697e-05,
+      "loss": 0.5989,
+      "step": 21620
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0302907208135524e-05,
+      "loss": 0.5477,
+      "step": 21621
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0302194702584918e-05,
+      "loss": 0.5849,
+      "step": 21622
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0301482195498774e-05,
+      "loss": 0.5801,
+      "step": 21623
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0300769686880708e-05,
+      "loss": 0.5738,
+      "step": 21624
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0300057176734347e-05,
+      "loss": 0.5926,
+      "step": 21625
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0299344665063304e-05,
+      "loss": 0.6249,
+      "step": 21626
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0298632151871204e-05,
+      "loss": 0.5773,
+      "step": 21627
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0297919637161668e-05,
+      "loss": 0.6599,
+      "step": 21628
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0297207120938313e-05,
+      "loss": 0.615,
+      "step": 21629
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0296494603204761e-05,
+      "loss": 0.6107,
+      "step": 21630
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0295782083964636e-05,
+      "loss": 0.579,
+      "step": 21631
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0295069563221554e-05,
+      "loss": 0.6267,
+      "step": 21632
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0294357040979135e-05,
+      "loss": 0.6127,
+      "step": 21633
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0293644517241006e-05,
+      "loss": 0.6829,
+      "step": 21634
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.029293199201078e-05,
+      "loss": 0.639,
+      "step": 21635
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.029221946529208e-05,
+      "loss": 0.5968,
+      "step": 21636
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.029150693708853e-05,
+      "loss": 0.6076,
+      "step": 21637
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0290794407403749e-05,
+      "loss": 0.6424,
+      "step": 21638
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0290081876241353e-05,
+      "loss": 0.6097,
+      "step": 21639
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0289369343604967e-05,
+      "loss": 0.5573,
+      "step": 21640
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.028865680949821e-05,
+      "loss": 0.6247,
+      "step": 21641
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0287944273924706e-05,
+      "loss": 0.6643,
+      "step": 21642
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0287231736888071e-05,
+      "loss": 0.5855,
+      "step": 21643
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0286519198391929e-05,
+      "loss": 0.6222,
+      "step": 21644
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0285806658439898e-05,
+      "loss": 0.5852,
+      "step": 21645
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0285094117035599e-05,
+      "loss": 0.5936,
+      "step": 21646
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0284381574182653e-05,
+      "loss": 0.6073,
+      "step": 21647
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0283669029884682e-05,
+      "loss": 0.6115,
+      "step": 21648
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0282956484145304e-05,
+      "loss": 0.6222,
+      "step": 21649
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0282243936968143e-05,
+      "loss": 0.6427,
+      "step": 21650
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0281531388356818e-05,
+      "loss": 0.5865,
+      "step": 21651
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0280818838314948e-05,
+      "loss": 0.5868,
+      "step": 21652
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0280106286846157e-05,
+      "loss": 0.5486,
+      "step": 21653
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0279393733954062e-05,
+      "loss": 0.5555,
+      "step": 21654
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0278681179642285e-05,
+      "loss": 0.6349,
+      "step": 21655
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0277968623914452e-05,
+      "loss": 0.5899,
+      "step": 21656
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0277256066774172e-05,
+      "loss": 0.5662,
+      "step": 21657
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.027654350822508e-05,
+      "loss": 0.5813,
+      "step": 21658
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0275830948270784e-05,
+      "loss": 0.5646,
+      "step": 21659
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0275118386914913e-05,
+      "loss": 0.5878,
+      "step": 21660
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0274405824161084e-05,
+      "loss": 0.5481,
+      "step": 21661
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0273693260012918e-05,
+      "loss": 0.5817,
+      "step": 21662
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.027298069447404e-05,
+      "loss": 0.5831,
+      "step": 21663
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0272268127548062e-05,
+      "loss": 0.5321,
+      "step": 21664
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0271555559238613e-05,
+      "loss": 0.6138,
+      "step": 21665
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0270842989549312e-05,
+      "loss": 0.5934,
+      "step": 21666
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0270130418483775e-05,
+      "loss": 0.6697,
+      "step": 21667
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0269417846045625e-05,
+      "loss": 0.6248,
+      "step": 21668
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.026870527223849e-05,
+      "loss": 0.5653,
+      "step": 21669
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0267992697065981e-05,
+      "loss": 0.6263,
+      "step": 21670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0267280120531722e-05,
+      "loss": 0.6151,
+      "step": 21671
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0266567542639336e-05,
+      "loss": 0.6227,
+      "step": 21672
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.026585496339244e-05,
+      "loss": 0.5481,
+      "step": 21673
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0265142382794661e-05,
+      "loss": 0.6128,
+      "step": 21674
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0264429800849613e-05,
+      "loss": 0.6182,
+      "step": 21675
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.026371721756092e-05,
+      "loss": 0.5759,
+      "step": 21676
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0263004632932206e-05,
+      "loss": 0.6334,
+      "step": 21677
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0262292046967084e-05,
+      "loss": 0.6361,
+      "step": 21678
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.026157945966918e-05,
+      "loss": 0.6096,
+      "step": 21679
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0260866871042114e-05,
+      "loss": 0.5911,
+      "step": 21680
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0260154281089508e-05,
+      "loss": 0.5765,
+      "step": 21681
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0259441689814983e-05,
+      "loss": 0.6095,
+      "step": 21682
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0258729097222157e-05,
+      "loss": 0.6393,
+      "step": 21683
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0258016503314654e-05,
+      "loss": 0.6112,
+      "step": 21684
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0257303908096092e-05,
+      "loss": 0.6292,
+      "step": 21685
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0256591311570097e-05,
+      "loss": 0.5622,
+      "step": 21686
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0255878713740281e-05,
+      "loss": 0.6767,
+      "step": 21687
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0255166114610275e-05,
+      "loss": 0.6242,
+      "step": 21688
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0254453514183693e-05,
+      "loss": 0.6434,
+      "step": 21689
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0253740912464158e-05,
+      "loss": 0.5791,
+      "step": 21690
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0253028309455293e-05,
+      "loss": 0.5616,
+      "step": 21691
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0252315705160713e-05,
+      "loss": 0.604,
+      "step": 21692
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.025160309958405e-05,
+      "loss": 0.5966,
+      "step": 21693
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0250890492728911e-05,
+      "loss": 0.6362,
+      "step": 21694
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0250177884598927e-05,
+      "loss": 0.6048,
+      "step": 21695
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0249465275197716e-05,
+      "loss": 0.6435,
+      "step": 21696
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0248752664528899e-05,
+      "loss": 0.6276,
+      "step": 21697
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0248040052596093e-05,
+      "loss": 0.5737,
+      "step": 21698
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0247327439402927e-05,
+      "loss": 0.5961,
+      "step": 21699
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0246614824953017e-05,
+      "loss": 0.6363,
+      "step": 21700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0245902209249988e-05,
+      "loss": 0.6098,
+      "step": 21701
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0245189592297451e-05,
+      "loss": 0.5812,
+      "step": 21702
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.024447697409904e-05,
+      "loss": 0.5847,
+      "step": 21703
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0243764354658365e-05,
+      "loss": 0.5804,
+      "step": 21704
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0243051733979056e-05,
+      "loss": 0.595,
+      "step": 21705
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0242339112064727e-05,
+      "loss": 0.6184,
+      "step": 21706
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0241626488919006e-05,
+      "loss": 0.5395,
+      "step": 21707
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0240913864545507e-05,
+      "loss": 0.6031,
+      "step": 21708
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0240201238947854e-05,
+      "loss": 0.5833,
+      "step": 21709
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0239488612129669e-05,
+      "loss": 0.6131,
+      "step": 21710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0238775984094571e-05,
+      "loss": 0.5932,
+      "step": 21711
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0238063354846184e-05,
+      "loss": 0.563,
+      "step": 21712
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0237350724388126e-05,
+      "loss": 0.5986,
+      "step": 21713
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0236638092724018e-05,
+      "loss": 0.594,
+      "step": 21714
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0235925459857487e-05,
+      "loss": 0.6392,
+      "step": 21715
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0235212825792147e-05,
+      "loss": 0.5935,
+      "step": 21716
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.023450019053162e-05,
+      "loss": 0.5556,
+      "step": 21717
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0233787554079533e-05,
+      "loss": 0.5581,
+      "step": 21718
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0233074916439499e-05,
+      "loss": 0.6229,
+      "step": 21719
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0232362277615147e-05,
+      "loss": 0.5849,
+      "step": 21720
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.023164963761009e-05,
+      "loss": 0.5932,
+      "step": 21721
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0230936996427959e-05,
+      "loss": 0.6977,
+      "step": 21722
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0230224354072362e-05,
+      "loss": 0.5707,
+      "step": 21723
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0229511710546933e-05,
+      "loss": 0.6401,
+      "step": 21724
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0228799065855285e-05,
+      "loss": 0.5797,
+      "step": 21725
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0228086420001046e-05,
+      "loss": 0.6395,
+      "step": 21726
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.022737377298783e-05,
+      "loss": 0.6106,
+      "step": 21727
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.022666112481926e-05,
+      "loss": 0.6161,
+      "step": 21728
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0225948475498961e-05,
+      "loss": 0.6116,
+      "step": 21729
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.022523582503055e-05,
+      "loss": 0.6127,
+      "step": 21730
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0224523173417651e-05,
+      "loss": 0.6051,
+      "step": 21731
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0223810520663884e-05,
+      "loss": 0.5866,
+      "step": 21732
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.022309786677287e-05,
+      "loss": 0.6337,
+      "step": 21733
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.022238521174823e-05,
+      "loss": 0.6036,
+      "step": 21734
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0221672555593585e-05,
+      "loss": 0.6659,
+      "step": 21735
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0220959898312556e-05,
+      "loss": 0.5865,
+      "step": 21736
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0220247239908766e-05,
+      "loss": 0.6346,
+      "step": 21737
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0219534580385838e-05,
+      "loss": 0.6767,
+      "step": 21738
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0218821919747389e-05,
+      "loss": 0.5815,
+      "step": 21739
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.021810925799704e-05,
+      "loss": 0.5751,
+      "step": 21740
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0217396595138416e-05,
+      "loss": 0.5648,
+      "step": 21741
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0216683931175135e-05,
+      "loss": 0.5739,
+      "step": 21742
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.021597126611082e-05,
+      "loss": 0.6594,
+      "step": 21743
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0215258599949092e-05,
+      "loss": 0.5867,
+      "step": 21744
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0214545932693572e-05,
+      "loss": 0.644,
+      "step": 21745
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0213833264347879e-05,
+      "loss": 0.6839,
+      "step": 21746
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0213120594915637e-05,
+      "loss": 0.6273,
+      "step": 21747
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0212407924400472e-05,
+      "loss": 0.5774,
+      "step": 21748
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0211695252805996e-05,
+      "loss": 0.644,
+      "step": 21749
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0210982580135835e-05,
+      "loss": 0.5293,
+      "step": 21750
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0210269906393609e-05,
+      "loss": 0.6204,
+      "step": 21751
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0209557231582938e-05,
+      "loss": 0.6057,
+      "step": 21752
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0208844555707449e-05,
+      "loss": 0.6752,
+      "step": 21753
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.020813187877076e-05,
+      "loss": 0.6438,
+      "step": 21754
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0207419200776487e-05,
+      "loss": 0.6041,
+      "step": 21755
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0206706521728263e-05,
+      "loss": 0.5957,
+      "step": 21756
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0205993841629698e-05,
+      "loss": 0.6329,
+      "step": 21757
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0205281160484419e-05,
+      "loss": 0.5614,
+      "step": 21758
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0204568478296046e-05,
+      "loss": 0.6053,
+      "step": 21759
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.02038557950682e-05,
+      "loss": 0.6297,
+      "step": 21760
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0203143110804508e-05,
+      "loss": 0.5272,
+      "step": 21761
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0202430425508583e-05,
+      "loss": 0.6356,
+      "step": 21762
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.020171773918405e-05,
+      "loss": 0.5547,
+      "step": 21763
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.020100505183453e-05,
+      "loss": 0.5929,
+      "step": 21764
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0200292363463641e-05,
+      "loss": 0.6283,
+      "step": 21765
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0199579674075014e-05,
+      "loss": 0.6366,
+      "step": 21766
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.019886698367226e-05,
+      "loss": 0.5956,
+      "step": 21767
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0198154292259004e-05,
+      "loss": 0.5714,
+      "step": 21768
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.019744159983887e-05,
+      "loss": 0.5896,
+      "step": 21769
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0196728906415478e-05,
+      "loss": 0.5434,
+      "step": 21770
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0196016211992447e-05,
+      "loss": 0.6235,
+      "step": 21771
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0195303516573402e-05,
+      "loss": 0.5969,
+      "step": 21772
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.019459082016196e-05,
+      "loss": 0.5945,
+      "step": 21773
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0193878122761746e-05,
+      "loss": 0.6108,
+      "step": 21774
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0193165424376383e-05,
+      "loss": 0.5766,
+      "step": 21775
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0192452725009487e-05,
+      "loss": 0.5528,
+      "step": 21776
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0191740024664685e-05,
+      "loss": 0.5278,
+      "step": 21777
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0191027323345595e-05,
+      "loss": 0.6051,
+      "step": 21778
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0190314621055837e-05,
+      "loss": 0.6612,
+      "step": 21779
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0189601917799037e-05,
+      "loss": 0.5485,
+      "step": 21780
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0188889213578812e-05,
+      "loss": 0.5987,
+      "step": 21781
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0188176508398788e-05,
+      "loss": 0.6225,
+      "step": 21782
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0187463802262585e-05,
+      "loss": 0.5857,
+      "step": 21783
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0186751095173821e-05,
+      "loss": 0.5805,
+      "step": 21784
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0186038387136122e-05,
+      "loss": 0.5959,
+      "step": 21785
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0185325678153104e-05,
+      "loss": 0.5786,
+      "step": 21786
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0184612968228393e-05,
+      "loss": 0.6373,
+      "step": 21787
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.018390025736561e-05,
+      "loss": 0.5768,
+      "step": 21788
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0183187545568379e-05,
+      "loss": 0.6329,
+      "step": 21789
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0182474832840312e-05,
+      "loss": 0.6081,
+      "step": 21790
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0181762119185043e-05,
+      "loss": 0.5691,
+      "step": 21791
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0181049404606186e-05,
+      "loss": 0.5892,
+      "step": 21792
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0180336689107359e-05,
+      "loss": 0.6002,
+      "step": 21793
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0179623972692194e-05,
+      "loss": 0.561,
+      "step": 21794
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0178911255364306e-05,
+      "loss": 0.5933,
+      "step": 21795
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0178198537127317e-05,
+      "loss": 0.6358,
+      "step": 21796
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.017748581798485e-05,
+      "loss": 0.5996,
+      "step": 21797
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0176773097940522e-05,
+      "loss": 0.5483,
+      "step": 21798
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0176060376997962e-05,
+      "loss": 0.6262,
+      "step": 21799
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0175347655160785e-05,
+      "loss": 0.6824,
+      "step": 21800
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0174634932432619e-05,
+      "loss": 0.6036,
+      "step": 21801
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0173922208817082e-05,
+      "loss": 0.5691,
+      "step": 21802
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0173209484317792e-05,
+      "loss": 0.6367,
+      "step": 21803
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0172496758938375e-05,
+      "loss": 0.6237,
+      "step": 21804
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.017178403268245e-05,
+      "loss": 0.5821,
+      "step": 21805
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.017107130555364e-05,
+      "loss": 0.5721,
+      "step": 21806
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.017035857755557e-05,
+      "loss": 0.6317,
+      "step": 21807
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0169645848691854e-05,
+      "loss": 0.5992,
+      "step": 21808
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0168933118966118e-05,
+      "loss": 0.6044,
+      "step": 21809
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0168220388381986e-05,
+      "loss": 0.585,
+      "step": 21810
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0167507656943077e-05,
+      "loss": 0.6082,
+      "step": 21811
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0166794924653011e-05,
+      "loss": 0.5627,
+      "step": 21812
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0166082191515411e-05,
+      "loss": 0.5969,
+      "step": 21813
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0165369457533897e-05,
+      "loss": 0.5576,
+      "step": 21814
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0164656722712098e-05,
+      "loss": 0.5549,
+      "step": 21815
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0163943987053626e-05,
+      "loss": 0.6439,
+      "step": 21816
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0163231250562106e-05,
+      "loss": 0.5949,
+      "step": 21817
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0162518513241162e-05,
+      "loss": 0.5211,
+      "step": 21818
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0161805775094414e-05,
+      "loss": 0.6625,
+      "step": 21819
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0161093036125482e-05,
+      "loss": 0.5634,
+      "step": 21820
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.016038029633799e-05,
+      "loss": 0.5962,
+      "step": 21821
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0159667555735559e-05,
+      "loss": 0.6293,
+      "step": 21822
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0158954814321813e-05,
+      "loss": 0.6579,
+      "step": 21823
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0158242072100366e-05,
+      "loss": 0.5753,
+      "step": 21824
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0157529329074846e-05,
+      "loss": 0.5893,
+      "step": 21825
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0156816585248876e-05,
+      "loss": 0.6974,
+      "step": 21826
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0156103840626072e-05,
+      "loss": 0.5897,
+      "step": 21827
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0155391095210059e-05,
+      "loss": 0.6009,
+      "step": 21828
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0154678349004458e-05,
+      "loss": 0.5627,
+      "step": 21829
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0153965602012891e-05,
+      "loss": 0.6062,
+      "step": 21830
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0153252854238984e-05,
+      "loss": 0.5818,
+      "step": 21831
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.015254010568635e-05,
+      "loss": 0.5823,
+      "step": 21832
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0151827356358615e-05,
+      "loss": 0.5673,
+      "step": 21833
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0151114606259404e-05,
+      "loss": 0.5843,
+      "step": 21834
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.015040185539233e-05,
+      "loss": 0.6191,
+      "step": 21835
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0149689103761025e-05,
+      "loss": 0.5759,
+      "step": 21836
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0148976351369104e-05,
+      "loss": 0.5639,
+      "step": 21837
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0148263598220191e-05,
+      "loss": 0.6503,
+      "step": 21838
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0147550844317908e-05,
+      "loss": 0.5406,
+      "step": 21839
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0146838089665874e-05,
+      "loss": 0.617,
+      "step": 21840
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0146125334267717e-05,
+      "loss": 0.7155,
+      "step": 21841
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0145412578127052e-05,
+      "loss": 0.6595,
+      "step": 21842
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0144699821247501e-05,
+      "loss": 0.594,
+      "step": 21843
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0143987063632687e-05,
+      "loss": 0.5791,
+      "step": 21844
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0143274305286236e-05,
+      "loss": 0.627,
+      "step": 21845
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0142561546211764e-05,
+      "loss": 0.6236,
+      "step": 21846
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0141848786412896e-05,
+      "loss": 0.6312,
+      "step": 21847
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0141136025893252e-05,
+      "loss": 0.6128,
+      "step": 21848
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0140423264656456e-05,
+      "loss": 0.646,
+      "step": 21849
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0139710502706129e-05,
+      "loss": 0.5395,
+      "step": 21850
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0138997740045889e-05,
+      "loss": 0.538,
+      "step": 21851
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.013828497667936e-05,
+      "loss": 0.6353,
+      "step": 21852
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.013757221261017e-05,
+      "loss": 0.6372,
+      "step": 21853
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0136859447841929e-05,
+      "loss": 0.5834,
+      "step": 21854
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0136146682378267e-05,
+      "loss": 0.5995,
+      "step": 21855
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0135433916222808e-05,
+      "loss": 0.5752,
+      "step": 21856
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0134721149379165e-05,
+      "loss": 0.5382,
+      "step": 21857
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0134008381850969e-05,
+      "loss": 0.6374,
+      "step": 21858
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0133295613641831e-05,
+      "loss": 0.6426,
+      "step": 21859
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0132582844755383e-05,
+      "loss": 0.6443,
+      "step": 21860
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0131870075195243e-05,
+      "loss": 0.6175,
+      "step": 21861
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0131157304965032e-05,
+      "loss": 0.6194,
+      "step": 21862
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.013044453406837e-05,
+      "loss": 0.5481,
+      "step": 21863
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0129731762508885e-05,
+      "loss": 0.6723,
+      "step": 21864
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0129018990290191e-05,
+      "loss": 0.634,
+      "step": 21865
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0128306217415914e-05,
+      "loss": 0.6524,
+      "step": 21866
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0127593443889678e-05,
+      "loss": 0.5636,
+      "step": 21867
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0126880669715102e-05,
+      "loss": 0.6327,
+      "step": 21868
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0126167894895809e-05,
+      "loss": 0.6508,
+      "step": 21869
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0125455119435417e-05,
+      "loss": 0.5659,
+      "step": 21870
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0124742343337551e-05,
+      "loss": 0.6899,
+      "step": 21871
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0124029566605835e-05,
+      "loss": 0.5656,
+      "step": 21872
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0123316789243886e-05,
+      "loss": 0.5862,
+      "step": 21873
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0122604011255331e-05,
+      "loss": 0.6137,
+      "step": 21874
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0121891232643788e-05,
+      "loss": 0.6255,
+      "step": 21875
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0121178453412878e-05,
+      "loss": 0.585,
+      "step": 21876
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0120465673566229e-05,
+      "loss": 0.6588,
+      "step": 21877
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0119752893107453e-05,
+      "loss": 0.5932,
+      "step": 21878
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0119040112040184e-05,
+      "loss": 0.5779,
+      "step": 21879
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0118327330368032e-05,
+      "loss": 0.6082,
+      "step": 21880
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0117614548094628e-05,
+      "loss": 0.6194,
+      "step": 21881
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0116901765223585e-05,
+      "loss": 0.6079,
+      "step": 21882
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0116188981758535e-05,
+      "loss": 0.6171,
+      "step": 21883
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0115476197703091e-05,
+      "loss": 0.6084,
+      "step": 21884
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0114763413060882e-05,
+      "loss": 0.6235,
+      "step": 21885
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0114050627835525e-05,
+      "loss": 0.5948,
+      "step": 21886
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0113337842030641e-05,
+      "loss": 0.6042,
+      "step": 21887
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0112625055649858e-05,
+      "loss": 0.6218,
+      "step": 21888
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0111912268696793e-05,
+      "loss": 0.6027,
+      "step": 21889
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0111199481175066e-05,
+      "loss": 0.6114,
+      "step": 21890
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0110486693088306e-05,
+      "loss": 0.5833,
+      "step": 21891
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.010977390444013e-05,
+      "loss": 0.6301,
+      "step": 21892
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0109061115234158e-05,
+      "loss": 0.6673,
+      "step": 21893
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0108348325474018e-05,
+      "loss": 0.6185,
+      "step": 21894
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0107635535163324e-05,
+      "loss": 0.6403,
+      "step": 21895
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0106922744305708e-05,
+      "loss": 0.5824,
+      "step": 21896
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0106209952904782e-05,
+      "loss": 0.6239,
+      "step": 21897
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0105497160964175e-05,
+      "loss": 0.6226,
+      "step": 21898
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0104784368487504e-05,
+      "loss": 0.6127,
+      "step": 21899
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0104071575478393e-05,
+      "loss": 0.624,
+      "step": 21900
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0103358781940465e-05,
+      "loss": 0.6555,
+      "step": 21901
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0102645987877342e-05,
+      "loss": 0.5616,
+      "step": 21902
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0101933193292642e-05,
+      "loss": 0.6131,
+      "step": 21903
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.010122039818999e-05,
+      "loss": 0.6401,
+      "step": 21904
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0100507602573006e-05,
+      "loss": 0.6507,
+      "step": 21905
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0099794806445315e-05,
+      "loss": 0.5755,
+      "step": 21906
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.009908200981054e-05,
+      "loss": 0.6173,
+      "step": 21907
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0098369212672297e-05,
+      "loss": 0.6653,
+      "step": 21908
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0097656415034212e-05,
+      "loss": 0.6237,
+      "step": 21909
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0096943616899905e-05,
+      "loss": 0.5828,
+      "step": 21910
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0096230818273001e-05,
+      "loss": 0.5855,
+      "step": 21911
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0095518019157118e-05,
+      "loss": 0.5905,
+      "step": 21912
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0094805219555884e-05,
+      "loss": 0.6453,
+      "step": 21913
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0094092419472911e-05,
+      "loss": 0.602,
+      "step": 21914
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0093379618911831e-05,
+      "loss": 0.5648,
+      "step": 21915
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0092666817876262e-05,
+      "loss": 0.5763,
+      "step": 21916
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0091954016369823e-05,
+      "loss": 0.6186,
+      "step": 21917
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0091241214396141e-05,
+      "loss": 0.6101,
+      "step": 21918
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0090528411958837e-05,
+      "loss": 0.6093,
+      "step": 21919
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0089815609061528e-05,
+      "loss": 0.614,
+      "step": 21920
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0089102805707845e-05,
+      "loss": 0.6062,
+      "step": 21921
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0088390001901398e-05,
+      "loss": 0.5823,
+      "step": 21922
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008767719764582e-05,
+      "loss": 0.6007,
+      "step": 21923
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0086964392944727e-05,
+      "loss": 0.6257,
+      "step": 21924
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008625158780174e-05,
+      "loss": 0.603,
+      "step": 21925
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0085538782220487e-05,
+      "loss": 0.6002,
+      "step": 21926
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0084825976204584e-05,
+      "loss": 0.6284,
+      "step": 21927
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0084113169757657e-05,
+      "loss": 0.6482,
+      "step": 21928
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0083400362883327e-05,
+      "loss": 0.5643,
+      "step": 21929
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0082687555585214e-05,
+      "loss": 0.5546,
+      "step": 21930
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0081974747866938e-05,
+      "loss": 0.6628,
+      "step": 21931
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008126193973213e-05,
+      "loss": 0.5831,
+      "step": 21932
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0080549131184402e-05,
+      "loss": 0.659,
+      "step": 21933
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0079836322227386e-05,
+      "loss": 0.6145,
+      "step": 21934
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0079123512864693e-05,
+      "loss": 0.5493,
+      "step": 21935
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007841070309995e-05,
+      "loss": 0.5719,
+      "step": 21936
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0077697892936783e-05,
+      "loss": 0.5941,
+      "step": 21937
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007698508237881e-05,
+      "loss": 0.624,
+      "step": 21938
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007627227142965e-05,
+      "loss": 0.5831,
+      "step": 21939
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0075559460092932e-05,
+      "loss": 0.5746,
+      "step": 21940
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0074846648372272e-05,
+      "loss": 0.6029,
+      "step": 21941
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0074133836271293e-05,
+      "loss": 0.5904,
+      "step": 21942
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007342102379362e-05,
+      "loss": 0.6443,
+      "step": 21943
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0072708210942872e-05,
+      "loss": 0.6371,
+      "step": 21944
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0071995397722673e-05,
+      "loss": 0.5862,
+      "step": 21945
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0071282584136645e-05,
+      "loss": 0.5782,
+      "step": 21946
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0070569770188407e-05,
+      "loss": 0.5767,
+      "step": 21947
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0069856955881587e-05,
+      "loss": 0.6063,
+      "step": 21948
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0069144141219801e-05,
+      "loss": 0.5312,
+      "step": 21949
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0068431326206676e-05,
+      "loss": 0.6186,
+      "step": 21950
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0067718510845827e-05,
+      "loss": 0.655,
+      "step": 21951
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0067005695140884e-05,
+      "loss": 0.6033,
+      "step": 21952
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0066292879095462e-05,
+      "loss": 0.5493,
+      "step": 21953
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0065580062713189e-05,
+      "loss": 0.5695,
+      "step": 21954
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0064867245997682e-05,
+      "loss": 0.6146,
+      "step": 21955
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.006415442895257e-05,
+      "loss": 0.6003,
+      "step": 21956
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0063441611581469e-05,
+      "loss": 0.6063,
+      "step": 21957
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0062728793888002e-05,
+      "loss": 0.6028,
+      "step": 21958
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0062015975875792e-05,
+      "loss": 0.5639,
+      "step": 21959
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0061303157548458e-05,
+      "loss": 0.6164,
+      "step": 21960
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.006059033890963e-05,
+      "loss": 0.6146,
+      "step": 21961
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005987751996292e-05,
+      "loss": 0.5921,
+      "step": 21962
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0059164700711956e-05,
+      "loss": 0.6173,
+      "step": 21963
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005845188116036e-05,
+      "loss": 0.5461,
+      "step": 21964
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0057739061311752e-05,
+      "loss": 0.5905,
+      "step": 21965
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0057026241169753e-05,
+      "loss": 0.613,
+      "step": 21966
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0056313420737992e-05,
+      "loss": 0.5188,
+      "step": 21967
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0055600600020078e-05,
+      "loss": 0.5529,
+      "step": 21968
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.005488777901965e-05,
+      "loss": 0.5028,
+      "step": 21969
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0054174957740314e-05,
+      "loss": 0.5779,
+      "step": 21970
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0053462136185704e-05,
+      "loss": 0.5622,
+      "step": 21971
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0052749314359436e-05,
+      "loss": 0.5766,
+      "step": 21972
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0052036492265132e-05,
+      "loss": 0.6117,
+      "step": 21973
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0051323669906414e-05,
+      "loss": 0.5762,
+      "step": 21974
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.005061084728691e-05,
+      "loss": 0.6219,
+      "step": 21975
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0049898024410234e-05,
+      "loss": 0.5761,
+      "step": 21976
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0049185201280013e-05,
+      "loss": 0.5516,
+      "step": 21977
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0048472377899869e-05,
+      "loss": 0.6166,
+      "step": 21978
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0047759554273419e-05,
+      "loss": 0.5932,
+      "step": 21979
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0047046730404293e-05,
+      "loss": 0.5813,
+      "step": 21980
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0046333906296105e-05,
+      "loss": 0.5819,
+      "step": 21981
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.004562108195248e-05,
+      "loss": 0.587,
+      "step": 21982
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0044908257377044e-05,
+      "loss": 0.5961,
+      "step": 21983
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0044195432573417e-05,
+      "loss": 0.6523,
+      "step": 21984
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0043482607545214e-05,
+      "loss": 0.6053,
+      "step": 21985
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0042769782296069e-05,
+      "loss": 0.5919,
+      "step": 21986
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0042056956829596e-05,
+      "loss": 0.582,
+      "step": 21987
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0041344131149422e-05,
+      "loss": 0.5968,
+      "step": 21988
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0040631305259163e-05,
+      "loss": 0.5161,
+      "step": 21989
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0039918479162443e-05,
+      "loss": 0.6473,
+      "step": 21990
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.003920565286289e-05,
+      "loss": 0.5895,
+      "step": 21991
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0038492826364119e-05,
+      "loss": 0.6257,
+      "step": 21992
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0037779999669755e-05,
+      "loss": 0.62,
+      "step": 21993
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.003706717278342e-05,
+      "loss": 0.6614,
+      "step": 21994
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0036354345708737e-05,
+      "loss": 0.6089,
+      "step": 21995
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0035641518449326e-05,
+      "loss": 0.5535,
+      "step": 21996
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0034928691008807e-05,
+      "loss": 0.6789,
+      "step": 21997
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0034215863390811e-05,
+      "loss": 0.5477,
+      "step": 21998
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0033503035598953e-05,
+      "loss": 0.6075,
+      "step": 21999
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0032790207636853e-05,
+      "loss": 0.5802,
+      "step": 22000
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0032077379508135e-05,
+      "loss": 0.5627,
+      "step": 22001
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0031364551216426e-05,
+      "loss": 0.6325,
+      "step": 22002
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0030651722765343e-05,
+      "loss": 0.539,
+      "step": 22003
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002993889415851e-05,
+      "loss": 0.6051,
+      "step": 22004
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0029226065399551e-05,
+      "loss": 0.6028,
+      "step": 22005
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0028513236492083e-05,
+      "loss": 0.6338,
+      "step": 22006
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0027800407439731e-05,
+      "loss": 0.5794,
+      "step": 22007
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0027087578246117e-05,
+      "loss": 0.625,
+      "step": 22008
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0026374748914865e-05,
+      "loss": 0.6213,
+      "step": 22009
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0025661919449593e-05,
+      "loss": 0.599,
+      "step": 22010
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0024949089853926e-05,
+      "loss": 0.5907,
+      "step": 22011
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0024236260131485e-05,
+      "loss": 0.5345,
+      "step": 22012
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0023523430285896e-05,
+      "loss": 0.5347,
+      "step": 22013
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0022810600320772e-05,
+      "loss": 0.6794,
+      "step": 22014
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0022097770239745e-05,
+      "loss": 0.5836,
+      "step": 22015
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002138494004643e-05,
+      "loss": 0.5658,
+      "step": 22016
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0020672109744454e-05,
+      "loss": 0.5754,
+      "step": 22017
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0019959279337434e-05,
+      "loss": 0.5725,
+      "step": 22018
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0019246448829e-05,
+      "loss": 0.577,
+      "step": 22019
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0018533618222766e-05,
+      "loss": 0.6254,
+      "step": 22020
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0017820787522356e-05,
+      "loss": 0.6791,
+      "step": 22021
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0017107956731395e-05,
+      "loss": 0.5922,
+      "step": 22022
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0016395125853501e-05,
+      "loss": 0.6183,
+      "step": 22023
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0015682294892303e-05,
+      "loss": 0.6524,
+      "step": 22024
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0014969463851417e-05,
+      "loss": 0.6059,
+      "step": 22025
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0014256632734467e-05,
+      "loss": 0.6078,
+      "step": 22026
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0013543801545072e-05,
+      "loss": 0.579,
+      "step": 22027
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0012830970286858e-05,
+      "loss": 0.6444,
+      "step": 22028
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0012118138963449e-05,
+      "loss": 0.6199,
+      "step": 22029
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0011405307578462e-05,
+      "loss": 0.6507,
+      "step": 22030
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001069247613552e-05,
+      "loss": 0.6167,
+      "step": 22031
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0009979644638248e-05,
+      "loss": 0.6027,
+      "step": 22032
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0009266813090266e-05,
+      "loss": 0.5763,
+      "step": 22033
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.00085539814952e-05,
+      "loss": 0.5622,
+      "step": 22034
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0007841149856662e-05,
+      "loss": 0.5693,
+      "step": 22035
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0007128318178287e-05,
+      "loss": 0.6144,
+      "step": 22036
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0006415486463687e-05,
+      "loss": 0.5742,
+      "step": 22037
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0005702654716492e-05,
+      "loss": 0.6072,
+      "step": 22038
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0004989822940316e-05,
+      "loss": 0.6341,
+      "step": 22039
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0004276991138789e-05,
+      "loss": 0.6014,
+      "step": 22040
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0003564159315527e-05,
+      "loss": 0.5597,
+      "step": 22041
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0002851327474155e-05,
+      "loss": 0.5697,
+      "step": 22042
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0002138495618294e-05,
+      "loss": 0.6126,
+      "step": 22043
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0001425663751566e-05,
+      "loss": 0.6111,
+      "step": 22044
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0000712831877595e-05,
+      "loss": 0.6121,
+      "step": 22045
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1e-05,
+      "loss": 0.5738,
+      "step": 22046
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.999287168122407e-06,
+      "loss": 0.6141,
+      "step": 22047
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.998574336248436e-06,
+      "loss": 0.6482,
+      "step": 22048
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.997861504381713e-06,
+      "loss": 0.6069,
+      "step": 22049
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.99714867252585e-06,
+      "loss": 0.573,
+      "step": 22050
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.996435840684476e-06,
+      "loss": 0.6359,
+      "step": 22051
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.995723008861214e-06,
+      "loss": 0.5888,
+      "step": 22052
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.995010177059684e-06,
+      "loss": 0.6141,
+      "step": 22053
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.994297345283511e-06,
+      "loss": 0.5566,
+      "step": 22054
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.993584513536315e-06,
+      "loss": 0.6412,
+      "step": 22055
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.992871681821715e-06,
+      "loss": 0.5561,
+      "step": 22056
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.99215885014334e-06,
+      "loss": 0.6271,
+      "step": 22057
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.991446018504805e-06,
+      "loss": 0.6248,
+      "step": 22058
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.990733186909735e-06,
+      "loss": 0.6067,
+      "step": 22059
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.990020355361757e-06,
+      "loss": 0.6413,
+      "step": 22060
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.989307523864484e-06,
+      "loss": 0.6313,
+      "step": 22061
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.988594692421541e-06,
+      "loss": 0.6028,
+      "step": 22062
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.987881861036554e-06,
+      "loss": 0.5915,
+      "step": 22063
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.987169029713143e-06,
+      "loss": 0.6041,
+      "step": 22064
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.986456198454934e-06,
+      "loss": 0.6319,
+      "step": 22065
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.985743367265538e-06,
+      "loss": 0.5881,
+      "step": 22066
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.985030536148588e-06,
+      "loss": 0.581,
+      "step": 22067
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.9843177051077e-06,
+      "loss": 0.5814,
+      "step": 22068
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.983604874146499e-06,
+      "loss": 0.6439,
+      "step": 22069
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.982892043268607e-06,
+      "loss": 0.5909,
+      "step": 22070
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.982179212477647e-06,
+      "loss": 0.605,
+      "step": 22071
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.981466381777239e-06,
+      "loss": 0.6353,
+      "step": 22072
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.980753551171002e-06,
+      "loss": 0.5918,
+      "step": 22073
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.980040720662567e-06,
+      "loss": 0.5786,
+      "step": 22074
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.979327890255548e-06,
+      "loss": 0.6331,
+      "step": 22075
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.978615059953572e-06,
+      "loss": 0.5567,
+      "step": 22076
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.977902229760257e-06,
+      "loss": 0.6032,
+      "step": 22077
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97718939967923e-06,
+      "loss": 0.574,
+      "step": 22078
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.976476569714107e-06,
+      "loss": 0.5771,
+      "step": 22079
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.975763739868515e-06,
+      "loss": 0.573,
+      "step": 22080
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.975050910146074e-06,
+      "loss": 0.5102,
+      "step": 22081
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97433808055041e-06,
+      "loss": 0.5825,
+      "step": 22082
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.973625251085139e-06,
+      "loss": 0.5375,
+      "step": 22083
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.972912421753885e-06,
+      "loss": 0.6834,
+      "step": 22084
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97219959256027e-06,
+      "loss": 0.5869,
+      "step": 22085
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.971486763507919e-06,
+      "loss": 0.5632,
+      "step": 22086
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970773934600454e-06,
+      "loss": 0.5583,
+      "step": 22087
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970061105841493e-06,
+      "loss": 0.6157,
+      "step": 22088
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.969348277234659e-06,
+      "loss": 0.5655,
+      "step": 22089
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.968635448783575e-06,
+      "loss": 0.6567,
+      "step": 22090
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.967922620491865e-06,
+      "loss": 0.6612,
+      "step": 22091
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.967209792363154e-06,
+      "loss": 0.6626,
+      "step": 22092
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.966496964401054e-06,
+      "loss": 0.5858,
+      "step": 22093
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.965784136609192e-06,
+      "loss": 0.6082,
+      "step": 22094
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.965071308991195e-06,
+      "loss": 0.6607,
+      "step": 22095
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.964358481550678e-06,
+      "loss": 0.5824,
+      "step": 22096
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.963645654291265e-06,
+      "loss": 0.6251,
+      "step": 22097
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.962932827216583e-06,
+      "loss": 0.5885,
+      "step": 22098
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.962220000330249e-06,
+      "loss": 0.6706,
+      "step": 22099
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.961507173635884e-06,
+      "loss": 0.6172,
+      "step": 22100
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.960794347137112e-06,
+      "loss": 0.5779,
+      "step": 22101
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.960081520837557e-06,
+      "loss": 0.6236,
+      "step": 22102
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.959368694740842e-06,
+      "loss": 0.6561,
+      "step": 22103
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.958655868850583e-06,
+      "loss": 0.6032,
+      "step": 22104
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.957943043170407e-06,
+      "loss": 0.5792,
+      "step": 22105
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.957230217703934e-06,
+      "loss": 0.6628,
+      "step": 22106
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.956517392454785e-06,
+      "loss": 0.5813,
+      "step": 22107
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.955804567426585e-06,
+      "loss": 0.5937,
+      "step": 22108
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.95509174262296e-06,
+      "loss": 0.5869,
+      "step": 22109
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.954378918047523e-06,
+      "loss": 0.6288,
+      "step": 22110
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.953666093703899e-06,
+      "loss": 0.6019,
+      "step": 22111
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.95295326959571e-06,
+      "loss": 0.6077,
+      "step": 22112
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.952240445726583e-06,
+      "loss": 0.5838,
+      "step": 22113
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.951527622100136e-06,
+      "loss": 0.5996,
+      "step": 22114
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95081479871999e-06,
+      "loss": 0.6079,
+      "step": 22115
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95010197558977e-06,
+      "loss": 0.5993,
+      "step": 22116
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.949389152713093e-06,
+      "loss": 0.5727,
+      "step": 22117
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.948676330093586e-06,
+      "loss": 0.6056,
+      "step": 22118
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94796350773487e-06,
+      "loss": 0.5837,
+      "step": 22119
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.947250685640569e-06,
+      "loss": 0.5732,
+      "step": 22120
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.946537863814301e-06,
+      "loss": 0.6071,
+      "step": 22121
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.945825042259687e-06,
+      "loss": 0.6086,
+      "step": 22122
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.945112220980354e-06,
+      "loss": 0.5655,
+      "step": 22123
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94439939997992e-06,
+      "loss": 0.5745,
+      "step": 22124
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.943686579262015e-06,
+      "loss": 0.5575,
+      "step": 22125
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94297375883025e-06,
+      "loss": 0.6148,
+      "step": 22126
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.942260938688252e-06,
+      "loss": 0.6679,
+      "step": 22127
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.941548118839643e-06,
+      "loss": 0.6212,
+      "step": 22128
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.940835299288045e-06,
+      "loss": 0.6466,
+      "step": 22129
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.940122480037085e-06,
+      "loss": 0.6986,
+      "step": 22130
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.939409661090375e-06,
+      "loss": 0.664,
+      "step": 22131
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.938696842451544e-06,
+      "loss": 0.6132,
+      "step": 22132
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.937984024124211e-06,
+      "loss": 0.5659,
+      "step": 22133
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.937271206112001e-06,
+      "loss": 0.5823,
+      "step": 22134
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.936558388418533e-06,
+      "loss": 0.5463,
+      "step": 22135
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.935845571047431e-06,
+      "loss": 0.5403,
+      "step": 22136
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93513275400232e-06,
+      "loss": 0.5961,
+      "step": 22137
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.934419937286813e-06,
+      "loss": 0.6126,
+      "step": 22138
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93370712090454e-06,
+      "loss": 0.6166,
+      "step": 22139
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.932994304859118e-06,
+      "loss": 0.6337,
+      "step": 22140
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.932281489154178e-06,
+      "loss": 0.5779,
+      "step": 22141
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.931568673793331e-06,
+      "loss": 0.6489,
+      "step": 22142
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.930855858780202e-06,
+      "loss": 0.5891,
+      "step": 22143
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.930143044118415e-06,
+      "loss": 0.5569,
+      "step": 22144
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.929430229811593e-06,
+      "loss": 0.6311,
+      "step": 22145
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.928717415863357e-06,
+      "loss": 0.5859,
+      "step": 22146
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92800460227733e-06,
+      "loss": 0.6019,
+      "step": 22147
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.927291789057131e-06,
+      "loss": 0.6405,
+      "step": 22148
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.926578976206382e-06,
+      "loss": 0.5751,
+      "step": 22149
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.925866163728709e-06,
+      "loss": 0.6254,
+      "step": 22150
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.925153351627733e-06,
+      "loss": 0.6601,
+      "step": 22151
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.924440539907073e-06,
+      "loss": 0.6053,
+      "step": 22152
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.923727728570353e-06,
+      "loss": 0.5762,
+      "step": 22153
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.923014917621193e-06,
+      "loss": 0.5709,
+      "step": 22154
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92230210706322e-06,
+      "loss": 0.6524,
+      "step": 22155
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92158929690005e-06,
+      "loss": 0.6623,
+      "step": 22156
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92087648713531e-06,
+      "loss": 0.6491,
+      "step": 22157
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92016367777262e-06,
+      "loss": 0.5664,
+      "step": 22158
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.9194508688156e-06,
+      "loss": 0.599,
+      "step": 22159
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.918738060267872e-06,
+      "loss": 0.6129,
+      "step": 22160
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91802525213306e-06,
+      "loss": 0.5784,
+      "step": 22161
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.917312444414788e-06,
+      "loss": 0.6074,
+      "step": 22162
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.916599637116678e-06,
+      "loss": 0.6074,
+      "step": 22163
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.915886830242346e-06,
+      "loss": 0.6188,
+      "step": 22164
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.915174023795419e-06,
+      "loss": 0.5646,
+      "step": 22165
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.914461217779516e-06,
+      "loss": 0.5768,
+      "step": 22166
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91374841219826e-06,
+      "loss": 0.6074,
+      "step": 22167
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91303560705528e-06,
+      "loss": 0.586,
+      "step": 22168
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.912322802354185e-06,
+      "loss": 0.6009,
+      "step": 22169
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.911609998098605e-06,
+      "loss": 0.6087,
+      "step": 22170
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.910897194292159e-06,
+      "loss": 0.5593,
+      "step": 22171
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.910184390938473e-06,
+      "loss": 0.563,
+      "step": 22172
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.909471588041165e-06,
+      "loss": 0.6001,
+      "step": 22173
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90875878560386e-06,
+      "loss": 0.6128,
+      "step": 22174
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90804598363018e-06,
+      "loss": 0.5982,
+      "step": 22175
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.907333182123742e-06,
+      "loss": 0.5608,
+      "step": 22176
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90662038108817e-06,
+      "loss": 0.6386,
+      "step": 22177
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.905907580527089e-06,
+      "loss": 0.6628,
+      "step": 22178
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.905194780444123e-06,
+      "loss": 0.6414,
+      "step": 22179
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.904481980842885e-06,
+      "loss": 0.5864,
+      "step": 22180
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.903769181727002e-06,
+      "loss": 0.6348,
+      "step": 22181
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.903056383100096e-06,
+      "loss": 0.5999,
+      "step": 22182
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90234358496579e-06,
+      "loss": 0.6307,
+      "step": 22183
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.901630787327708e-06,
+      "loss": 0.6184,
+      "step": 22184
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.900917990189466e-06,
+      "loss": 0.5851,
+      "step": 22185
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.900205193554686e-06,
+      "loss": 0.6363,
+      "step": 22186
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.899492397426995e-06,
+      "loss": 0.5788,
+      "step": 22187
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.898779601810013e-06,
+      "loss": 0.5843,
+      "step": 22188
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.89806680670736e-06,
+      "loss": 0.6581,
+      "step": 22189
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.897354012122665e-06,
+      "loss": 0.6163,
+      "step": 22190
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.896641218059538e-06,
+      "loss": 0.6427,
+      "step": 22191
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.895928424521608e-06,
+      "loss": 0.5996,
+      "step": 22192
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.8952156315125e-06,
+      "loss": 0.5699,
+      "step": 22193
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.894502839035828e-06,
+      "loss": 0.6154,
+      "step": 22194
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.893790047095221e-06,
+      "loss": 0.5936,
+      "step": 22195
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.893077255694297e-06,
+      "loss": 0.6106,
+      "step": 22196
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.892364464836677e-06,
+      "loss": 0.5612,
+      "step": 22197
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.891651674525985e-06,
+      "loss": 0.5472,
+      "step": 22198
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.890938884765842e-06,
+      "loss": 0.5898,
+      "step": 22199
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.890226095559872e-06,
+      "loss": 0.628,
+      "step": 22200
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.889513306911699e-06,
+      "loss": 0.6581,
+      "step": 22201
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.888800518824935e-06,
+      "loss": 0.5992,
+      "step": 22202
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.88808773130321e-06,
+      "loss": 0.5873,
+      "step": 22203
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.887374944350145e-06,
+      "loss": 0.5515,
+      "step": 22204
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.886662157969358e-06,
+      "loss": 0.5492,
+      "step": 22205
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.88594937216448e-06,
+      "loss": 0.6366,
+      "step": 22206
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.885236586939123e-06,
+      "loss": 0.6721,
+      "step": 22207
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.88452380229691e-06,
+      "loss": 0.6547,
+      "step": 22208
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.883811018241466e-06,
+      "loss": 0.5957,
+      "step": 22209
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.883098234776415e-06,
+      "loss": 0.6495,
+      "step": 22210
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.882385451905377e-06,
+      "loss": 0.6063,
+      "step": 22211
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.881672669631972e-06,
+      "loss": 0.6154,
+      "step": 22212
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.88095988795982e-06,
+      "loss": 0.6181,
+      "step": 22213
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.880247106892549e-06,
+      "loss": 0.5546,
+      "step": 22214
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.879534326433774e-06,
+      "loss": 0.5877,
+      "step": 22215
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.878821546587124e-06,
+      "loss": 0.6153,
+      "step": 22216
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.878108767356218e-06,
+      "loss": 0.6231,
+      "step": 22217
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.877395988744674e-06,
+      "loss": 0.6228,
+      "step": 22218
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.876683210756116e-06,
+      "loss": 0.6173,
+      "step": 22219
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.875970433394168e-06,
+      "loss": 0.6261,
+      "step": 22220
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.875257656662449e-06,
+      "loss": 0.633,
+      "step": 22221
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.874544880564588e-06,
+      "loss": 0.6482,
+      "step": 22222
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.873832105104198e-06,
+      "loss": 0.6087,
+      "step": 22223
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.873119330284901e-06,
+      "loss": 0.6364,
+      "step": 22224
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.872406556110324e-06,
+      "loss": 0.617,
+      "step": 22225
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.871693782584087e-06,
+      "loss": 0.6046,
+      "step": 22226
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.870981009709809e-06,
+      "loss": 0.5691,
+      "step": 22227
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.87026823749112e-06,
+      "loss": 0.642,
+      "step": 22228
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.869555465931633e-06,
+      "loss": 0.603,
+      "step": 22229
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.868842695034971e-06,
+      "loss": 0.5522,
+      "step": 22230
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.868129924804762e-06,
+      "loss": 0.6116,
+      "step": 22231
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.867417155244618e-06,
+      "loss": 0.5472,
+      "step": 22232
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.86670438635817e-06,
+      "loss": 0.597,
+      "step": 22233
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.865991618149036e-06,
+      "loss": 0.6446,
+      "step": 22234
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.865278850620838e-06,
+      "loss": 0.6183,
+      "step": 22235
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.864566083777195e-06,
+      "loss": 0.6255,
+      "step": 22236
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.863853317621733e-06,
+      "loss": 0.6518,
+      "step": 22237
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.863140552158071e-06,
+      "loss": 0.5476,
+      "step": 22238
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.862427787389837e-06,
+      "loss": 0.5955,
+      "step": 22239
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.861715023320642e-06,
+      "loss": 0.557,
+      "step": 22240
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.861002259954115e-06,
+      "loss": 0.6073,
+      "step": 22241
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.860289497293874e-06,
+      "loss": 0.5893,
+      "step": 22242
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.859576735343546e-06,
+      "loss": 0.636,
+      "step": 22243
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.858863974106751e-06,
+      "loss": 0.6259,
+      "step": 22244
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.858151213587109e-06,
+      "loss": 0.5519,
+      "step": 22245
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.857438453788239e-06,
+      "loss": 0.6314,
+      "step": 22246
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.856725694713767e-06,
+      "loss": 0.5523,
+      "step": 22247
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.856012936367313e-06,
+      "loss": 0.575,
+      "step": 22248
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.855300178752506e-06,
+      "loss": 0.5252,
+      "step": 22249
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.854587421872953e-06,
+      "loss": 0.5658,
+      "step": 22250
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.853874665732287e-06,
+      "loss": 0.5573,
+      "step": 22251
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.853161910334127e-06,
+      "loss": 0.565,
+      "step": 22252
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.852449155682094e-06,
+      "loss": 0.6631,
+      "step": 22253
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.851736401779808e-06,
+      "loss": 0.6121,
+      "step": 22254
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.851023648630898e-06,
+      "loss": 0.648,
+      "step": 22255
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.850310896238979e-06,
+      "loss": 0.5302,
+      "step": 22256
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.849598144607671e-06,
+      "loss": 0.6444,
+      "step": 22257
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.8488853937406e-06,
+      "loss": 0.6066,
+      "step": 22258
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.848172643641385e-06,
+      "loss": 0.67,
+      "step": 22259
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.847459894313655e-06,
+      "loss": 0.6882,
+      "step": 22260
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.846747145761021e-06,
+      "loss": 0.6542,
+      "step": 22261
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84603439798711e-06,
+      "loss": 0.616,
+      "step": 22262
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.845321650995543e-06,
+      "loss": 0.5456,
+      "step": 22263
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.844608904789941e-06,
+      "loss": 0.5925,
+      "step": 22264
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84389615937393e-06,
+      "loss": 0.6083,
+      "step": 22265
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84318341475113e-06,
+      "loss": 0.6431,
+      "step": 22266
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.842470670925157e-06,
+      "loss": 0.6139,
+      "step": 22267
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.841757927899637e-06,
+      "loss": 0.601,
+      "step": 22268
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.841045185678192e-06,
+      "loss": 0.6297,
+      "step": 22269
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.840332444264443e-06,
+      "loss": 0.5199,
+      "step": 22270
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.839619703662013e-06,
+      "loss": 0.6001,
+      "step": 22271
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83890696387452e-06,
+      "loss": 0.5519,
+      "step": 22272
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83819422490559e-06,
+      "loss": 0.6217,
+      "step": 22273
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83748148675884e-06,
+      "loss": 0.6033,
+      "step": 22274
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.836768749437895e-06,
+      "loss": 0.5876,
+      "step": 22275
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83605601294638e-06,
+      "loss": 0.6371,
+      "step": 22276
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.835343277287907e-06,
+      "loss": 0.62,
+      "step": 22277
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.834630542466104e-06,
+      "loss": 0.6481,
+      "step": 22278
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83391780848459e-06,
+      "loss": 0.5631,
+      "step": 22279
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83320507534699e-06,
+      "loss": 0.5805,
+      "step": 22280
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.832492343056923e-06,
+      "loss": 0.6634,
+      "step": 22281
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.831779611618018e-06,
+      "loss": 0.6018,
+      "step": 22282
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.831066881033883e-06,
+      "loss": 0.6419,
+      "step": 22283
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.830354151308149e-06,
+      "loss": 0.59,
+      "step": 22284
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.829641422444434e-06,
+      "loss": 0.6596,
+      "step": 22285
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.82892869444636e-06,
+      "loss": 0.5913,
+      "step": 22286
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.828215967317555e-06,
+      "loss": 0.6338,
+      "step": 22287
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.82750324106163e-06,
+      "loss": 0.5878,
+      "step": 22288
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.826790515682211e-06,
+      "loss": 0.5865,
+      "step": 22289
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.826077791182921e-06,
+      "loss": 0.6046,
+      "step": 22290
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.825365067567385e-06,
+      "loss": 0.6641,
+      "step": 22291
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.824652344839213e-06,
+      "loss": 0.6539,
+      "step": 22292
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.823939623002041e-06,
+      "loss": 0.5745,
+      "step": 22293
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.82322690205948e-06,
+      "loss": 0.5812,
+      "step": 22294
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.822514182015154e-06,
+      "loss": 0.6185,
+      "step": 22295
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.821801462872685e-06,
+      "loss": 0.5639,
+      "step": 22296
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.821088744635695e-06,
+      "loss": 0.6558,
+      "step": 22297
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.82037602730781e-06,
+      "loss": 0.5632,
+      "step": 22298
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.819663310892643e-06,
+      "loss": 0.5796,
+      "step": 22299
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.81895059539382e-06,
+      "loss": 0.5832,
+      "step": 22300
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.81823788081496e-06,
+      "loss": 0.5998,
+      "step": 22301
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.817525167159688e-06,
+      "loss": 0.6321,
+      "step": 22302
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.816812454431623e-06,
+      "loss": 0.6105,
+      "step": 22303
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.816099742634393e-06,
+      "loss": 0.5502,
+      "step": 22304
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.81538703177161e-06,
+      "loss": 0.6294,
+      "step": 22305
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.814674321846899e-06,
+      "loss": 0.5784,
+      "step": 22306
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.813961612863882e-06,
+      "loss": 0.594,
+      "step": 22307
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.813248904826182e-06,
+      "loss": 0.7132,
+      "step": 22308
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.81253619773742e-06,
+      "loss": 0.5675,
+      "step": 22309
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.811823491601215e-06,
+      "loss": 0.5683,
+      "step": 22310
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.811110786421191e-06,
+      "loss": 0.5224,
+      "step": 22311
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.810398082200966e-06,
+      "loss": 0.5995,
+      "step": 22312
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.809685378944164e-06,
+      "loss": 0.6047,
+      "step": 22313
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.808972676654408e-06,
+      "loss": 0.6612,
+      "step": 22314
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.80825997533532e-06,
+      "loss": 0.5882,
+      "step": 22315
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.807547274990516e-06,
+      "loss": 0.6408,
+      "step": 22316
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.80683457562362e-06,
+      "loss": 0.5766,
+      "step": 22317
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.806121877238254e-06,
+      "loss": 0.5999,
+      "step": 22318
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.80540917983804e-06,
+      "loss": 0.6102,
+      "step": 22319
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.804696483426603e-06,
+      "loss": 0.5719,
+      "step": 22320
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.803983788007556e-06,
+      "loss": 0.6003,
+      "step": 22321
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.803271093584525e-06,
+      "loss": 0.5123,
+      "step": 22322
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.802558400161133e-06,
+      "loss": 0.6508,
+      "step": 22323
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.801845707740996e-06,
+      "loss": 0.6128,
+      "step": 22324
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.801133016327745e-06,
+      "loss": 0.6535,
+      "step": 22325
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.800420325924991e-06,
+      "loss": 0.5709,
+      "step": 22326
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.79970763653636e-06,
+      "loss": 0.5961,
+      "step": 22327
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.798994948165473e-06,
+      "loss": 0.5726,
+      "step": 22328
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.798282260815954e-06,
+      "loss": 0.6018,
+      "step": 22329
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.79756957449142e-06,
+      "loss": 0.5514,
+      "step": 22330
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.796856889195497e-06,
+      "loss": 0.5783,
+      "step": 22331
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.796144204931801e-06,
+      "loss": 0.6337,
+      "step": 22332
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.795431521703956e-06,
+      "loss": 0.6693,
+      "step": 22333
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.794718839515583e-06,
+      "loss": 0.6361,
+      "step": 22334
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.794006158370304e-06,
+      "loss": 0.5786,
+      "step": 22335
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.793293478271742e-06,
+      "loss": 0.5858,
+      "step": 22336
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.792580799223515e-06,
+      "loss": 0.619,
+      "step": 22337
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.791868121229244e-06,
+      "loss": 0.5728,
+      "step": 22338
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.791155444292553e-06,
+      "loss": 0.5792,
+      "step": 22339
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.790442768417062e-06,
+      "loss": 0.5889,
+      "step": 22340
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.789730093606396e-06,
+      "loss": 0.6226,
+      "step": 22341
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.789017419864171e-06,
+      "loss": 0.6216,
+      "step": 22342
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.788304747194009e-06,
+      "loss": 0.6432,
+      "step": 22343
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.787592075599533e-06,
+      "loss": 0.57,
+      "step": 22344
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.786879405084361e-06,
+      "loss": 0.5961,
+      "step": 22345
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.786166735652121e-06,
+      "loss": 0.6359,
+      "step": 22346
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.785454067306433e-06,
+      "loss": 0.6374,
+      "step": 22347
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.784741400050912e-06,
+      "loss": 0.6044,
+      "step": 22348
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.784028733889183e-06,
+      "loss": 0.6217,
+      "step": 22349
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.783316068824868e-06,
+      "loss": 0.5625,
+      "step": 22350
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.782603404861588e-06,
+      "loss": 0.6496,
+      "step": 22351
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.781890742002964e-06,
+      "loss": 0.6187,
+      "step": 22352
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.781178080252616e-06,
+      "loss": 0.6362,
+      "step": 22353
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.780465419614166e-06,
+      "loss": 0.614,
+      "step": 22354
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.779752760091235e-06,
+      "loss": 0.585,
+      "step": 22355
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.779040101687446e-06,
+      "loss": 0.7067,
+      "step": 22356
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.778327444406415e-06,
+      "loss": 0.6107,
+      "step": 22357
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.777614788251776e-06,
+      "loss": 0.6143,
+      "step": 22358
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.776902133227136e-06,
+      "loss": 0.6003,
+      "step": 22359
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.776189479336119e-06,
+      "loss": 0.5624,
+      "step": 22360
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.77547682658235e-06,
+      "loss": 0.6268,
+      "step": 22361
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.774764174969451e-06,
+      "loss": 0.5976,
+      "step": 22362
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.774051524501044e-06,
+      "loss": 0.6055,
+      "step": 22363
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.773338875180744e-06,
+      "loss": 0.5727,
+      "step": 22364
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.772626227012173e-06,
+      "loss": 0.5751,
+      "step": 22365
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.771913579998957e-06,
+      "loss": 0.6053,
+      "step": 22366
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.771200934144716e-06,
+      "loss": 0.6347,
+      "step": 22367
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.770488289453072e-06,
+      "loss": 0.617,
+      "step": 22368
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.76977564592764e-06,
+      "loss": 0.5411,
+      "step": 22369
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.769063003572046e-06,
+      "loss": 0.618,
+      "step": 22370
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.768350362389912e-06,
+      "loss": 0.5864,
+      "step": 22371
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.767637722384857e-06,
+      "loss": 0.602,
+      "step": 22372
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.766925083560501e-06,
+      "loss": 0.597,
+      "step": 22373
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.766212445920474e-06,
+      "loss": 0.5577,
+      "step": 22374
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.765499809468383e-06,
+      "loss": 0.5629,
+      "step": 22375
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.764787174207856e-06,
+      "loss": 0.5893,
+      "step": 22376
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.764074540142516e-06,
+      "loss": 0.5847,
+      "step": 22377
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.76336190727598e-06,
+      "loss": 0.567,
+      "step": 22378
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.76264927561188e-06,
+      "loss": 0.597,
+      "step": 22379
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.76193664515382e-06,
+      "loss": 0.6375,
+      "step": 22380
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.761224015905432e-06,
+      "loss": 0.6255,
+      "step": 22381
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.760511387870335e-06,
+      "loss": 0.7037,
+      "step": 22382
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.759798761052148e-06,
+      "loss": 0.6614,
+      "step": 22383
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.759086135454495e-06,
+      "loss": 0.6029,
+      "step": 22384
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.758373511080999e-06,
+      "loss": 0.6016,
+      "step": 22385
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.757660887935274e-06,
+      "loss": 0.5271,
+      "step": 22386
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.756948266020946e-06,
+      "loss": 0.6188,
+      "step": 22387
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.756235645341637e-06,
+      "loss": 0.6069,
+      "step": 22388
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.755523025900963e-06,
+      "loss": 0.621,
+      "step": 22389
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.75481040770255e-06,
+      "loss": 0.6118,
+      "step": 22390
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.754097790750017e-06,
+      "loss": 0.6198,
+      "step": 22391
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.753385175046985e-06,
+      "loss": 0.6137,
+      "step": 22392
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.752672560597075e-06,
+      "loss": 0.578,
+      "step": 22393
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.751959947403907e-06,
+      "loss": 0.5697,
+      "step": 22394
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.751247335471103e-06,
+      "loss": 0.5622,
+      "step": 22395
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.75053472480229e-06,
+      "loss": 0.6082,
+      "step": 22396
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.749822115401076e-06,
+      "loss": 0.5699,
+      "step": 22397
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.749109507271092e-06,
+      "loss": 0.5519,
+      "step": 22398
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.748396900415954e-06,
+      "loss": 0.5834,
+      "step": 22399
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.747684294839285e-06,
+      "loss": 0.5529,
+      "step": 22400
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.746971690544712e-06,
+      "loss": 0.5621,
+      "step": 22401
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.746259087535845e-06,
+      "loss": 0.5755,
+      "step": 22402
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.74554648581631e-06,
+      "loss": 0.5409,
+      "step": 22403
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.744833885389727e-06,
+      "loss": 0.6422,
+      "step": 22404
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.744121286259719e-06,
+      "loss": 0.5878,
+      "step": 22405
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.743408688429908e-06,
+      "loss": 0.6467,
+      "step": 22406
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.74269609190391e-06,
+      "loss": 0.6052,
+      "step": 22407
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.741983496685349e-06,
+      "loss": 0.5571,
+      "step": 22408
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.741270902777846e-06,
+      "loss": 0.587,
+      "step": 22409
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.74055831018502e-06,
+      "loss": 0.6779,
+      "step": 22410
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.739845718910492e-06,
+      "loss": 0.5616,
+      "step": 22411
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.739133128957891e-06,
+      "loss": 0.6216,
+      "step": 22412
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.738420540330824e-06,
+      "loss": 0.593,
+      "step": 22413
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.73770795303292e-06,
+      "loss": 0.6456,
+      "step": 22414
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.736995367067799e-06,
+      "loss": 0.5801,
+      "step": 22415
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.73628278243908e-06,
+      "loss": 0.6365,
+      "step": 22416
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.735570199150392e-06,
+      "loss": 0.619,
+      "step": 22417
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.734857617205342e-06,
+      "loss": 0.6022,
+      "step": 22418
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.734145036607561e-06,
+      "loss": 0.6441,
+      "step": 22419
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.733432457360666e-06,
+      "loss": 0.5839,
+      "step": 22420
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.73271987946828e-06,
+      "loss": 0.6109,
+      "step": 22421
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.73200730293402e-06,
+      "loss": 0.6442,
+      "step": 22422
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.731294727761514e-06,
+      "loss": 0.6251,
+      "step": 22423
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.730582153954376e-06,
+      "loss": 0.5392,
+      "step": 22424
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.729869581516226e-06,
+      "loss": 0.6037,
+      "step": 22425
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.72915701045069e-06,
+      "loss": 0.6183,
+      "step": 22426
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.728444440761389e-06,
+      "loss": 0.6139,
+      "step": 22427
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.727731872451942e-06,
+      "loss": 0.5053,
+      "step": 22428
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.727019305525964e-06,
+      "loss": 0.6031,
+      "step": 22429
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.726306739987085e-06,
+      "loss": 0.576,
+      "step": 22430
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.725594175838918e-06,
+      "loss": 0.6774,
+      "step": 22431
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.724881613085089e-06,
+      "loss": 0.5888,
+      "step": 22432
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.724169051729221e-06,
+      "loss": 0.6001,
+      "step": 22433
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.723456491774926e-06,
+      "loss": 0.5684,
+      "step": 22434
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.72274393322583e-06,
+      "loss": 0.6175,
+      "step": 22435
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.722031376085552e-06,
+      "loss": 0.5775,
+      "step": 22436
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.721318820357715e-06,
+      "loss": 0.596,
+      "step": 22437
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.720606266045938e-06,
+      "loss": 0.6286,
+      "step": 22438
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.719893713153848e-06,
+      "loss": 0.6621,
+      "step": 22439
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.719181161685055e-06,
+      "loss": 0.6028,
+      "step": 22440
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.718468611643185e-06,
+      "loss": 0.604,
+      "step": 22441
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.717756063031859e-06,
+      "loss": 0.6014,
+      "step": 22442
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.717043515854696e-06,
+      "loss": 0.7199,
+      "step": 22443
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.716330970115323e-06,
+      "loss": 0.6398,
+      "step": 22444
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.71561842581735e-06,
+      "loss": 0.6125,
+      "step": 22445
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.714905882964404e-06,
+      "loss": 0.6622,
+      "step": 22446
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.714193341560103e-06,
+      "loss": 0.5829,
+      "step": 22447
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.713480801608074e-06,
+      "loss": 0.5632,
+      "step": 22448
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.71276826311193e-06,
+      "loss": 0.6117,
+      "step": 22449
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.712055726075297e-06,
+      "loss": 0.629,
+      "step": 22450
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.711343190501791e-06,
+      "loss": 0.5629,
+      "step": 22451
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.710630656395035e-06,
+      "loss": 0.584,
+      "step": 22452
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.70991812375865e-06,
+      "loss": 0.5043,
+      "step": 22453
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.709205592596253e-06,
+      "loss": 0.6601,
+      "step": 22454
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.708493062911474e-06,
+      "loss": 0.6009,
+      "step": 22455
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.707780534707923e-06,
+      "loss": 0.5924,
+      "step": 22456
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.707068007989223e-06,
+      "loss": 0.6449,
+      "step": 22457
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.706355482758997e-06,
+      "loss": 0.6233,
+      "step": 22458
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.705642959020866e-06,
+      "loss": 0.6657,
+      "step": 22459
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.704930436778453e-06,
+      "loss": 0.6352,
+      "step": 22460
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.704217916035369e-06,
+      "loss": 0.6332,
+      "step": 22461
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.703505396795242e-06,
+      "loss": 0.591,
+      "step": 22462
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.70279287906169e-06,
+      "loss": 0.5661,
+      "step": 22463
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.702080362838335e-06,
+      "loss": 0.6147,
+      "step": 22464
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.7013678481288e-06,
+      "loss": 0.6617,
+      "step": 22465
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.7006553349367e-06,
+      "loss": 0.5882,
+      "step": 22466
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.699942823265658e-06,
+      "loss": 0.5971,
+      "step": 22467
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.699230313119296e-06,
+      "loss": 0.6105,
+      "step": 22468
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.69851780450123e-06,
+      "loss": 0.527,
+      "step": 22469
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.697805297415083e-06,
+      "loss": 0.5448,
+      "step": 22470
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.697092791864479e-06,
+      "loss": 0.6307,
+      "step": 22471
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.696380287853035e-06,
+      "loss": 0.5059,
+      "step": 22472
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.695667785384369e-06,
+      "loss": 0.5581,
+      "step": 22473
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.694955284462105e-06,
+      "loss": 0.5594,
+      "step": 22474
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.694242785089863e-06,
+      "loss": 0.5645,
+      "step": 22475
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.693530287271261e-06,
+      "loss": 0.4599,
+      "step": 22476
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.692817791009927e-06,
+      "loss": 0.5582,
+      "step": 22477
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.69210529630947e-06,
+      "loss": 0.5919,
+      "step": 22478
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.691392803173518e-06,
+      "loss": 0.5729,
+      "step": 22479
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.690680311605688e-06,
+      "loss": 0.6017,
+      "step": 22480
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.689967821609604e-06,
+      "loss": 0.5979,
+      "step": 22481
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.689255333188888e-06,
+      "loss": 0.6846,
+      "step": 22482
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.68854284634715e-06,
+      "loss": 0.5229,
+      "step": 22483
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.68783036108802e-06,
+      "loss": 0.5973,
+      "step": 22484
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.687117877415113e-06,
+      "loss": 0.5839,
+      "step": 22485
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.686405395332055e-06,
+      "loss": 0.6394,
+      "step": 22486
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.685692914842462e-06,
+      "loss": 0.6519,
+      "step": 22487
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.684980435949957e-06,
+      "loss": 0.5777,
+      "step": 22488
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.684267958658157e-06,
+      "loss": 0.5508,
+      "step": 22489
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.683555482970684e-06,
+      "loss": 0.5843,
+      "step": 22490
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.682843008891157e-06,
+      "loss": 0.623,
+      "step": 22491
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.682130536423196e-06,
+      "loss": 0.6166,
+      "step": 22492
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.68141806557043e-06,
+      "loss": 0.617,
+      "step": 22493
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.680705596336466e-06,
+      "loss": 0.6344,
+      "step": 22494
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.679993128724932e-06,
+      "loss": 0.5599,
+      "step": 22495
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.679280662739445e-06,
+      "loss": 0.6087,
+      "step": 22496
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.678568198383627e-06,
+      "loss": 0.578,
+      "step": 22497
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.677855735661104e-06,
+      "loss": 0.5915,
+      "step": 22498
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.677143274575484e-06,
+      "loss": 0.6133,
+      "step": 22499
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.676430815130395e-06,
+      "loss": 0.5188,
+      "step": 22500
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.675718357329455e-06,
+      "loss": 0.5868,
+      "step": 22501
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.675005901176286e-06,
+      "loss": 0.584,
+      "step": 22502
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.674293446674508e-06,
+      "loss": 0.5539,
+      "step": 22503
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.673580993827742e-06,
+      "loss": 0.5141,
+      "step": 22504
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.672868542639604e-06,
+      "loss": 0.6268,
+      "step": 22505
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.672156093113715e-06,
+      "loss": 0.6137,
+      "step": 22506
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.6714436452537e-06,
+      "loss": 0.5881,
+      "step": 22507
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.670731199063173e-06,
+      "loss": 0.6137,
+      "step": 22508
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.670018754545762e-06,
+      "loss": 0.5922,
+      "step": 22509
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.66930631170508e-06,
+      "loss": 0.6275,
+      "step": 22510
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.668593870544749e-06,
+      "loss": 0.6051,
+      "step": 22511
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.667881431068388e-06,
+      "loss": 0.6216,
+      "step": 22512
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.66716899327962e-06,
+      "loss": 0.6139,
+      "step": 22513
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.666456557182063e-06,
+      "loss": 0.619,
+      "step": 22514
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.665744122779342e-06,
+      "loss": 0.5747,
+      "step": 22515
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.66503169007507e-06,
+      "loss": 0.6297,
+      "step": 22516
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.664319259072869e-06,
+      "loss": 0.6349,
+      "step": 22517
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.66360682977636e-06,
+      "loss": 0.5785,
+      "step": 22518
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.662894402189165e-06,
+      "loss": 0.6499,
+      "step": 22519
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.662181976314905e-06,
+      "loss": 0.5116,
+      "step": 22520
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.661469552157193e-06,
+      "loss": 0.5814,
+      "step": 22521
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.660757129719653e-06,
+      "loss": 0.661,
+      "step": 22522
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.660044709005907e-06,
+      "loss": 0.5799,
+      "step": 22523
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.659332290019573e-06,
+      "loss": 0.6429,
+      "step": 22524
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.658619872764275e-06,
+      "loss": 0.5537,
+      "step": 22525
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.657907457243625e-06,
+      "loss": 0.6152,
+      "step": 22526
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.657195043461248e-06,
+      "loss": 0.6014,
+      "step": 22527
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.656482631420765e-06,
+      "loss": 0.5989,
+      "step": 22528
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.655770221125792e-06,
+      "loss": 0.6584,
+      "step": 22529
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.65505781257995e-06,
+      "loss": 0.6135,
+      "step": 22530
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.654345405786868e-06,
+      "loss": 0.5562,
+      "step": 22531
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.653633000750151e-06,
+      "loss": 0.6216,
+      "step": 22532
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.65292059747343e-06,
+      "loss": 0.5195,
+      "step": 22533
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.652208195960316e-06,
+      "loss": 0.6323,
+      "step": 22534
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.651495796214437e-06,
+      "loss": 0.6599,
+      "step": 22535
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.650783398239414e-06,
+      "loss": 0.5981,
+      "step": 22536
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.650071002038857e-06,
+      "loss": 0.5402,
+      "step": 22537
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.649358607616394e-06,
+      "loss": 0.6231,
+      "step": 22538
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.648646214975641e-06,
+      "loss": 0.5401,
+      "step": 22539
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.64793382412022e-06,
+      "loss": 0.5763,
+      "step": 22540
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.64722143505375e-06,
+      "loss": 0.5731,
+      "step": 22541
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.646509047779856e-06,
+      "loss": 0.6755,
+      "step": 22542
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.645796662302149e-06,
+      "loss": 0.6233,
+      "step": 22543
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.645084278624252e-06,
+      "loss": 0.6049,
+      "step": 22544
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.64437189674979e-06,
+      "loss": 0.5472,
+      "step": 22545
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.643659516682373e-06,
+      "loss": 0.5877,
+      "step": 22546
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.64294713842563e-06,
+      "loss": 0.5615,
+      "step": 22547
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.642234761983177e-06,
+      "loss": 0.6419,
+      "step": 22548
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.641522387358632e-06,
+      "loss": 0.5971,
+      "step": 22549
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.640810014555619e-06,
+      "loss": 0.6004,
+      "step": 22550
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.640097643577752e-06,
+      "loss": 0.544,
+      "step": 22551
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.639385274428661e-06,
+      "loss": 0.6123,
+      "step": 22552
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.638672907111954e-06,
+      "loss": 0.6154,
+      "step": 22553
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.637960541631254e-06,
+      "loss": 0.5423,
+      "step": 22554
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.637248177990185e-06,
+      "loss": 0.5398,
+      "step": 22555
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.636535816192363e-06,
+      "loss": 0.6103,
+      "step": 22556
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.63582345624141e-06,
+      "loss": 0.6448,
+      "step": 22557
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.635111098140948e-06,
+      "loss": 0.6059,
+      "step": 22558
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.63439874189459e-06,
+      "loss": 0.593,
+      "step": 22559
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.633686387505958e-06,
+      "loss": 0.5991,
+      "step": 22560
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.632974034978672e-06,
+      "loss": 0.6019,
+      "step": 22561
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.632261684316354e-06,
+      "loss": 0.5749,
+      "step": 22562
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.631549335522627e-06,
+      "loss": 0.6124,
+      "step": 22563
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.630836988601099e-06,
+      "loss": 0.5163,
+      "step": 22564
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.630124643555398e-06,
+      "loss": 0.6548,
+      "step": 22565
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.629412300389145e-06,
+      "loss": 0.6371,
+      "step": 22566
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.628699959105952e-06,
+      "loss": 0.6261,
+      "step": 22567
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.627987619709444e-06,
+      "loss": 0.5809,
+      "step": 22568
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.627275282203245e-06,
+      "loss": 0.629,
+      "step": 22569
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.626562946590967e-06,
+      "loss": 0.5506,
+      "step": 22570
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.625850612876229e-06,
+      "loss": 0.5237,
+      "step": 22571
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.625138281062653e-06,
+      "loss": 0.6291,
+      "step": 22572
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.624425951153861e-06,
+      "loss": 0.5925,
+      "step": 22573
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.623713623153475e-06,
+      "loss": 0.6288,
+      "step": 22574
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.623001297065107e-06,
+      "loss": 0.6298,
+      "step": 22575
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.622288972892377e-06,
+      "loss": 0.6245,
+      "step": 22576
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.62157665063891e-06,
+      "loss": 0.6434,
+      "step": 22577
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.620864330308322e-06,
+      "loss": 0.6007,
+      "step": 22578
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.620152011904232e-06,
+      "loss": 0.6241,
+      "step": 22579
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.619439695430267e-06,
+      "loss": 0.6188,
+      "step": 22580
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.618727380890036e-06,
+      "loss": 0.632,
+      "step": 22581
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.618015068287162e-06,
+      "loss": 0.6056,
+      "step": 22582
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.617302757625266e-06,
+      "loss": 0.6037,
+      "step": 22583
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.616590448907968e-06,
+      "loss": 0.6052,
+      "step": 22584
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.615878142138887e-06,
+      "loss": 0.5711,
+      "step": 22585
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.615165837321638e-06,
+      "loss": 0.6608,
+      "step": 22586
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.614453534459848e-06,
+      "loss": 0.7243,
+      "step": 22587
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.613741233557129e-06,
+      "loss": 0.592,
+      "step": 22588
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.613028934617105e-06,
+      "loss": 0.613,
+      "step": 22589
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.612316637643399e-06,
+      "loss": 0.579,
+      "step": 22590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.61160434263962e-06,
+      "loss": 0.6383,
+      "step": 22591
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.610892049609394e-06,
+      "loss": 0.6499,
+      "step": 22592
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.61017975855634e-06,
+      "loss": 0.6168,
+      "step": 22593
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.609467469484076e-06,
+      "loss": 0.6021,
+      "step": 22594
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.608755182396221e-06,
+      "loss": 0.5809,
+      "step": 22595
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.6080428972964e-06,
+      "loss": 0.61,
+      "step": 22596
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.607330614188226e-06,
+      "loss": 0.6237,
+      "step": 22597
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.606618333075318e-06,
+      "loss": 0.6269,
+      "step": 22598
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.605906053961298e-06,
+      "loss": 0.5762,
+      "step": 22599
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.605193776849784e-06,
+      "loss": 0.5068,
+      "step": 22600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.604481501744402e-06,
+      "loss": 0.5527,
+      "step": 22601
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.603769228648761e-06,
+      "loss": 0.5761,
+      "step": 22602
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.603056957566484e-06,
+      "loss": 0.5581,
+      "step": 22603
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.602344688501193e-06,
+      "loss": 0.6475,
+      "step": 22604
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.601632421456502e-06,
+      "loss": 0.5781,
+      "step": 22605
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.600920156436034e-06,
+      "loss": 0.5689,
+      "step": 22606
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.600207893443411e-06,
+      "loss": 0.6808,
+      "step": 22607
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.599495632482246e-06,
+      "loss": 0.6153,
+      "step": 22608
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.59878337355616e-06,
+      "loss": 0.6123,
+      "step": 22609
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.598071116668774e-06,
+      "loss": 0.6135,
+      "step": 22610
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.597358861823707e-06,
+      "loss": 0.5768,
+      "step": 22611
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.59664660902458e-06,
+      "loss": 0.5961,
+      "step": 22612
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.595934358275008e-06,
+      "loss": 0.5663,
+      "step": 22613
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.595222109578609e-06,
+      "loss": 0.5778,
+      "step": 22614
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.594509862939007e-06,
+      "loss": 0.5782,
+      "step": 22615
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.593797618359816e-06,
+      "loss": 0.5934,
+      "step": 22616
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.593085375844667e-06,
+      "loss": 0.6329,
+      "step": 22617
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.592373135397162e-06,
+      "loss": 0.5579,
+      "step": 22618
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.591660897020931e-06,
+      "loss": 0.5825,
+      "step": 22619
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.59094866071959e-06,
+      "loss": 0.6572,
+      "step": 22620
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.59023642649676e-06,
+      "loss": 0.5911,
+      "step": 22621
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.589524194356058e-06,
+      "loss": 0.6257,
+      "step": 22622
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.588811964301105e-06,
+      "loss": 0.6018,
+      "step": 22623
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.588099736335517e-06,
+      "loss": 0.6593,
+      "step": 22624
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.587387510462918e-06,
+      "loss": 0.5746,
+      "step": 22625
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.58667528668692e-06,
+      "loss": 0.585,
+      "step": 22626
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.585963065011147e-06,
+      "loss": 0.5736,
+      "step": 22627
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.585250845439219e-06,
+      "loss": 0.5588,
+      "step": 22628
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.584538627974752e-06,
+      "loss": 0.5816,
+      "step": 22629
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.583826412621364e-06,
+      "loss": 0.6455,
+      "step": 22630
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.583114199382675e-06,
+      "loss": 0.562,
+      "step": 22631
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.582401988262307e-06,
+      "loss": 0.5281,
+      "step": 22632
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.581689779263876e-06,
+      "loss": 0.6745,
+      "step": 22633
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.580977572391005e-06,
+      "loss": 0.6238,
+      "step": 22634
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.580265367647308e-06,
+      "loss": 0.6334,
+      "step": 22635
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.579553165036403e-06,
+      "loss": 0.6426,
+      "step": 22636
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.578840964561913e-06,
+      "loss": 0.6461,
+      "step": 22637
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.578128766227455e-06,
+      "loss": 0.5962,
+      "step": 22638
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.577416570036653e-06,
+      "loss": 0.5924,
+      "step": 22639
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.576704375993116e-06,
+      "loss": 0.5435,
+      "step": 22640
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.575992184100467e-06,
+      "loss": 0.5815,
+      "step": 22641
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.575279994362328e-06,
+      "loss": 0.6079,
+      "step": 22642
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.574567806782317e-06,
+      "loss": 0.6051,
+      "step": 22643
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.573855621364051e-06,
+      "loss": 0.5842,
+      "step": 22644
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.573143438111147e-06,
+      "loss": 0.5805,
+      "step": 22645
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.57243125702723e-06,
+      "loss": 0.5993,
+      "step": 22646
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.571719078115912e-06,
+      "loss": 0.624,
+      "step": 22647
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.571006901380816e-06,
+      "loss": 0.6198,
+      "step": 22648
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.570294726825557e-06,
+      "loss": 0.5431,
+      "step": 22649
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.569582554453762e-06,
+      "loss": 0.5657,
+      "step": 22650
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.56887038426904e-06,
+      "loss": 0.6044,
+      "step": 22651
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.568158216275014e-06,
+      "loss": 0.6149,
+      "step": 22652
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.567446050475302e-06,
+      "loss": 0.6237,
+      "step": 22653
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.566733886873524e-06,
+      "loss": 0.5513,
+      "step": 22654
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.566021725473303e-06,
+      "loss": 0.5764,
+      "step": 22655
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.565309566278245e-06,
+      "loss": 0.5634,
+      "step": 22656
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.564597409291978e-06,
+      "loss": 0.6185,
+      "step": 22657
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.563885254518121e-06,
+      "loss": 0.6268,
+      "step": 22658
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.56317310196029e-06,
+      "loss": 0.585,
+      "step": 22659
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.562460951622105e-06,
+      "loss": 0.583,
+      "step": 22660
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.561748803507185e-06,
+      "loss": 0.6468,
+      "step": 22661
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.561036657619147e-06,
+      "loss": 0.6004,
+      "step": 22662
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.560324513961608e-06,
+      "loss": 0.5693,
+      "step": 22663
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.559612372538192e-06,
+      "loss": 0.5966,
+      "step": 22664
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.558900233352513e-06,
+      "loss": 0.5685,
+      "step": 22665
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.558188096408192e-06,
+      "loss": 0.6043,
+      "step": 22666
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.557475961708848e-06,
+      "loss": 0.6434,
+      "step": 22667
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.556763829258095e-06,
+      "loss": 0.5717,
+      "step": 22668
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.556051699059555e-06,
+      "loss": 0.6179,
+      "step": 22669
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.555339571116847e-06,
+      "loss": 0.6211,
+      "step": 22670
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.554627445433587e-06,
+      "loss": 0.6087,
+      "step": 22671
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.553915322013401e-06,
+      "loss": 0.5989,
+      "step": 22672
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.553203200859898e-06,
+      "loss": 0.5959,
+      "step": 22673
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.5524910819767e-06,
+      "loss": 0.6206,
+      "step": 22674
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.551778965367426e-06,
+      "loss": 0.5789,
+      "step": 22675
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.551066851035694e-06,
+      "loss": 0.6313,
+      "step": 22676
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.550354738985128e-06,
+      "loss": 0.6143,
+      "step": 22677
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.549642629219335e-06,
+      "loss": 0.5866,
+      "step": 22678
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.548930521741941e-06,
+      "loss": 0.5712,
+      "step": 22679
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.548218416556564e-06,
+      "loss": 0.5731,
+      "step": 22680
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.547506313666822e-06,
+      "loss": 0.6373,
+      "step": 22681
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.546794213076335e-06,
+      "loss": 0.6403,
+      "step": 22682
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.546082114788715e-06,
+      "loss": 0.5834,
+      "step": 22683
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.545370018807585e-06,
+      "loss": 0.5801,
+      "step": 22684
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.544657925136567e-06,
+      "loss": 0.572,
+      "step": 22685
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.543945833779272e-06,
+      "loss": 0.5841,
+      "step": 22686
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.543233744739322e-06,
+      "loss": 0.6109,
+      "step": 22687
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.54252165802034e-06,
+      "loss": 0.5273,
+      "step": 22688
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.541809573625933e-06,
+      "loss": 0.5816,
+      "step": 22689
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.541097491559727e-06,
+      "loss": 0.6106,
+      "step": 22690
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.54038541182534e-06,
+      "loss": 0.5896,
+      "step": 22691
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.539673334426387e-06,
+      "loss": 0.5918,
+      "step": 22692
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.538961259366496e-06,
+      "loss": 0.6284,
+      "step": 22693
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.538249186649272e-06,
+      "loss": 0.6552,
+      "step": 22694
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.53753711627834e-06,
+      "loss": 0.6008,
+      "step": 22695
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.536825048257315e-06,
+      "loss": 0.6162,
+      "step": 22696
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.53611298258982e-06,
+      "loss": 0.5739,
+      "step": 22697
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.535400919279469e-06,
+      "loss": 0.6965,
+      "step": 22698
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.534688858329888e-06,
+      "loss": 0.5654,
+      "step": 22699
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.533976799744683e-06,
+      "loss": 0.5923,
+      "step": 22700
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.53326474352748e-06,
+      "loss": 0.6002,
+      "step": 22701
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.532552689681896e-06,
+      "loss": 0.642,
+      "step": 22702
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.531840638211548e-06,
+      "loss": 0.6081,
+      "step": 22703
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.531128589120057e-06,
+      "loss": 0.6268,
+      "step": 22704
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.53041654241104e-06,
+      "loss": 0.5652,
+      "step": 22705
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.52970449808811e-06,
+      "loss": 0.6304,
+      "step": 22706
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.528992456154889e-06,
+      "loss": 0.529,
+      "step": 22707
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.528280416614995e-06,
+      "loss": 0.5832,
+      "step": 22708
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.527568379472052e-06,
+      "loss": 0.6076,
+      "step": 22709
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.526856344729668e-06,
+      "loss": 0.5732,
+      "step": 22710
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.526144312391464e-06,
+      "loss": 0.5913,
+      "step": 22711
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.52543228246106e-06,
+      "loss": 0.5961,
+      "step": 22712
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.524720254942076e-06,
+      "loss": 0.5929,
+      "step": 22713
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.524008229838125e-06,
+      "loss": 0.582,
+      "step": 22714
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.523296207152834e-06,
+      "loss": 0.6289,
+      "step": 22715
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.522584186889809e-06,
+      "loss": 0.6151,
+      "step": 22716
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.521872169052673e-06,
+      "loss": 0.5491,
+      "step": 22717
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.521160153645045e-06,
+      "loss": 0.6559,
+      "step": 22718
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.520448140670542e-06,
+      "loss": 0.5172,
+      "step": 22719
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.519736130132787e-06,
+      "loss": 0.5721,
+      "step": 22720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.51902412203539e-06,
+      "loss": 0.6116,
+      "step": 22721
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.51831211638197e-06,
+      "loss": 0.599,
+      "step": 22722
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.517600113176153e-06,
+      "loss": 0.6295,
+      "step": 22723
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.516888112421546e-06,
+      "loss": 0.6317,
+      "step": 22724
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.516176114121774e-06,
+      "loss": 0.5841,
+      "step": 22725
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.515464118280455e-06,
+      "loss": 0.5956,
+      "step": 22726
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.514752124901203e-06,
+      "loss": 0.618,
+      "step": 22727
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.514040133987636e-06,
+      "loss": 0.5883,
+      "step": 22728
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.513328145543372e-06,
+      "loss": 0.6272,
+      "step": 22729
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.512616159572032e-06,
+      "loss": 0.5762,
+      "step": 22730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.511904176077237e-06,
+      "loss": 0.6192,
+      "step": 22731
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.511192195062593e-06,
+      "loss": 0.5927,
+      "step": 22732
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.510480216531728e-06,
+      "loss": 0.6646,
+      "step": 22733
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.509768240488253e-06,
+      "loss": 0.5787,
+      "step": 22734
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.50905626693579e-06,
+      "loss": 0.5344,
+      "step": 22735
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.508344295877962e-06,
+      "loss": 0.5595,
+      "step": 22736
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.507632327318375e-06,
+      "loss": 0.5888,
+      "step": 22737
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.506920361260653e-06,
+      "loss": 0.6474,
+      "step": 22738
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.506208397708412e-06,
+      "loss": 0.5975,
+      "step": 22739
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.505496436665274e-06,
+      "loss": 0.6026,
+      "step": 22740
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.504784478134851e-06,
+      "loss": 0.6322,
+      "step": 22741
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.504072522120766e-06,
+      "loss": 0.6409,
+      "step": 22742
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.50336056862663e-06,
+      "loss": 0.6472,
+      "step": 22743
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.502648617656067e-06,
+      "loss": 0.647,
+      "step": 22744
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.50193666921269e-06,
+      "loss": 0.6345,
+      "step": 22745
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.501224723300118e-06,
+      "loss": 0.6436,
+      "step": 22746
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.500512779921976e-06,
+      "loss": 0.6123,
+      "step": 22747
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.499800839081868e-06,
+      "loss": 0.5664,
+      "step": 22748
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.49908890078342e-06,
+      "loss": 0.5901,
+      "step": 22749
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.498376965030246e-06,
+      "loss": 0.6451,
+      "step": 22750
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.497665031825967e-06,
+      "loss": 0.6146,
+      "step": 22751
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.496953101174198e-06,
+      "loss": 0.5725,
+      "step": 22752
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.496241173078565e-06,
+      "loss": 0.5792,
+      "step": 22753
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.49552924754267e-06,
+      "loss": 0.6247,
+      "step": 22754
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.49481732457014e-06,
+      "loss": 0.6108,
+      "step": 22755
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.494105404164591e-06,
+      "loss": 0.5345,
+      "step": 22756
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.493393486329642e-06,
+      "loss": 0.6587,
+      "step": 22757
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.492681571068913e-06,
+      "loss": 0.6038,
+      "step": 22758
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.491969658386011e-06,
+      "loss": 0.5741,
+      "step": 22759
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.491257748284563e-06,
+      "loss": 0.5885,
+      "step": 22760
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.490545840768185e-06,
+      "loss": 0.5672,
+      "step": 22761
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.489833935840488e-06,
+      "loss": 0.5688,
+      "step": 22762
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.489122033505095e-06,
+      "loss": 0.5808,
+      "step": 22763
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.488410133765626e-06,
+      "loss": 0.6257,
+      "step": 22764
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.487698236625695e-06,
+      "loss": 0.569,
+      "step": 22765
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.486986342088917e-06,
+      "loss": 0.6082,
+      "step": 22766
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.48627445015891e-06,
+      "loss": 0.604,
+      "step": 22767
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.485562560839295e-06,
+      "loss": 0.5696,
+      "step": 22768
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.484850674133691e-06,
+      "loss": 0.6098,
+      "step": 22769
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.484138790045707e-06,
+      "loss": 0.569,
+      "step": 22770
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.483426908578965e-06,
+      "loss": 0.5776,
+      "step": 22771
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.482715029737081e-06,
+      "loss": 0.6362,
+      "step": 22772
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.482003153523676e-06,
+      "loss": 0.6207,
+      "step": 22773
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.481291279942367e-06,
+      "loss": 0.5888,
+      "step": 22774
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.480579408996767e-06,
+      "loss": 0.6285,
+      "step": 22775
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.479867540690492e-06,
+      "loss": 0.5889,
+      "step": 22776
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.479155675027164e-06,
+      "loss": 0.6365,
+      "step": 22777
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.478443812010398e-06,
+      "loss": 0.5667,
+      "step": 22778
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.477731951643814e-06,
+      "loss": 0.584,
+      "step": 22779
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.477020093931028e-06,
+      "loss": 0.5878,
+      "step": 22780
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.476308238875652e-06,
+      "loss": 0.5851,
+      "step": 22781
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.47559638648131e-06,
+      "loss": 0.6126,
+      "step": 22782
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.474884536751615e-06,
+      "loss": 0.5914,
+      "step": 22783
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.474172689690184e-06,
+      "loss": 0.6082,
+      "step": 22784
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.473460845300642e-06,
+      "loss": 0.5793,
+      "step": 22785
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.472749003586594e-06,
+      "loss": 0.5988,
+      "step": 22786
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.472037164551662e-06,
+      "loss": 0.6371,
+      "step": 22787
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.471325328199465e-06,
+      "loss": 0.5814,
+      "step": 22788
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.47061349453362e-06,
+      "loss": 0.6131,
+      "step": 22789
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.46990166355774e-06,
+      "loss": 0.6585,
+      "step": 22790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.469189835275451e-06,
+      "loss": 0.6104,
+      "step": 22791
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.46847800969036e-06,
+      "loss": 0.5966,
+      "step": 22792
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.467766186806086e-06,
+      "loss": 0.566,
+      "step": 22793
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.46705436662625e-06,
+      "loss": 0.556,
+      "step": 22794
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.466342549154464e-06,
+      "loss": 0.5996,
+      "step": 22795
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.465630734394353e-06,
+      "loss": 0.6072,
+      "step": 22796
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.464918922349527e-06,
+      "loss": 0.5876,
+      "step": 22797
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.464207113023603e-06,
+      "loss": 0.5515,
+      "step": 22798
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.4634953064202e-06,
+      "loss": 0.601,
+      "step": 22799
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.462783502542935e-06,
+      "loss": 0.5955,
+      "step": 22800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.462071701395426e-06,
+      "loss": 0.5832,
+      "step": 22801
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.461359902981286e-06,
+      "loss": 0.5874,
+      "step": 22802
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.460648107304136e-06,
+      "loss": 0.5481,
+      "step": 22803
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.459936314367589e-06,
+      "loss": 0.5693,
+      "step": 22804
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.459224524175263e-06,
+      "loss": 0.5892,
+      "step": 22805
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.458512736730776e-06,
+      "loss": 0.5906,
+      "step": 22806
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.457800952037748e-06,
+      "loss": 0.5865,
+      "step": 22807
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.457089170099789e-06,
+      "loss": 0.6175,
+      "step": 22808
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.456377390920518e-06,
+      "loss": 0.6456,
+      "step": 22809
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.455665614503553e-06,
+      "loss": 0.616,
+      "step": 22810
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.45495384085251e-06,
+      "loss": 0.6132,
+      "step": 22811
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.454242069971012e-06,
+      "loss": 0.6163,
+      "step": 22812
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.453530301862665e-06,
+      "loss": 0.6121,
+      "step": 22813
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.45281853653109e-06,
+      "loss": 0.5901,
+      "step": 22814
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.452106773979903e-06,
+      "loss": 0.643,
+      "step": 22815
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.451395014212724e-06,
+      "loss": 0.5893,
+      "step": 22816
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.450683257233168e-06,
+      "loss": 0.5944,
+      "step": 22817
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.449971503044853e-06,
+      "loss": 0.5921,
+      "step": 22818
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.449259751651392e-06,
+      "loss": 0.5776,
+      "step": 22819
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.448548003056402e-06,
+      "loss": 0.6535,
+      "step": 22820
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.447836257263504e-06,
+      "loss": 0.6257,
+      "step": 22821
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.447124514276308e-06,
+      "loss": 0.5927,
+      "step": 22822
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.44641277409844e-06,
+      "loss": 0.5863,
+      "step": 22823
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.445701036733508e-06,
+      "loss": 0.6078,
+      "step": 22824
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.44498930218513e-06,
+      "loss": 0.6296,
+      "step": 22825
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.444277570456923e-06,
+      "loss": 0.5913,
+      "step": 22826
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.443565841552505e-06,
+      "loss": 0.6045,
+      "step": 22827
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.442854115475497e-06,
+      "loss": 0.6249,
+      "step": 22828
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.442142392229505e-06,
+      "loss": 0.6356,
+      "step": 22829
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.44143067181815e-06,
+      "loss": 0.6253,
+      "step": 22830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.440718954245052e-06,
+      "loss": 0.5678,
+      "step": 22831
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.440007239513823e-06,
+      "loss": 0.6148,
+      "step": 22832
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.439295527628083e-06,
+      "loss": 0.5908,
+      "step": 22833
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.438583818591448e-06,
+      "loss": 0.6168,
+      "step": 22834
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.43787211240753e-06,
+      "loss": 0.6331,
+      "step": 22835
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.437160409079948e-06,
+      "loss": 0.6304,
+      "step": 22836
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.43644870861232e-06,
+      "loss": 0.6285,
+      "step": 22837
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.435737011008263e-06,
+      "loss": 0.575,
+      "step": 22838
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.435025316271391e-06,
+      "loss": 0.6056,
+      "step": 22839
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.434313624405319e-06,
+      "loss": 0.6284,
+      "step": 22840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.433601935413668e-06,
+      "loss": 0.5952,
+      "step": 22841
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.432890249300048e-06,
+      "loss": 0.6114,
+      "step": 22842
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.432178566068079e-06,
+      "loss": 0.6041,
+      "step": 22843
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.431466885721377e-06,
+      "loss": 0.6099,
+      "step": 22844
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.430755208263565e-06,
+      "loss": 0.5978,
+      "step": 22845
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.430043533698245e-06,
+      "loss": 0.6115,
+      "step": 22846
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.429331862029042e-06,
+      "loss": 0.5863,
+      "step": 22847
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.42862019325957e-06,
+      "loss": 0.548,
+      "step": 22848
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.427908527393449e-06,
+      "loss": 0.6157,
+      "step": 22849
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.427196864434293e-06,
+      "loss": 0.638,
+      "step": 22850
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.426485204385715e-06,
+      "loss": 0.6595,
+      "step": 22851
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.425773547251333e-06,
+      "loss": 0.6776,
+      "step": 22852
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.425061893034763e-06,
+      "loss": 0.6355,
+      "step": 22853
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.424350241739624e-06,
+      "loss": 0.6108,
+      "step": 22854
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.423638593369529e-06,
+      "loss": 0.5767,
+      "step": 22855
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.422926947928099e-06,
+      "loss": 0.5684,
+      "step": 22856
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.422215305418943e-06,
+      "loss": 0.6326,
+      "step": 22857
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.421503665845679e-06,
+      "loss": 0.6164,
+      "step": 22858
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.420792029211926e-06,
+      "loss": 0.6022,
+      "step": 22859
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.420080395521298e-06,
+      "loss": 0.5726,
+      "step": 22860
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.419368764777413e-06,
+      "loss": 0.5753,
+      "step": 22861
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.418657136983885e-06,
+      "loss": 0.6151,
+      "step": 22862
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.417945512144328e-06,
+      "loss": 0.6211,
+      "step": 22863
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.41723389026236e-06,
+      "loss": 0.5955,
+      "step": 22864
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.416522271341598e-06,
+      "loss": 0.5775,
+      "step": 22865
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.415810655385662e-06,
+      "loss": 0.6268,
+      "step": 22866
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.415099042398158e-06,
+      "loss": 0.5906,
+      "step": 22867
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.414387432382707e-06,
+      "loss": 0.6241,
+      "step": 22868
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.413675825342925e-06,
+      "loss": 0.5956,
+      "step": 22869
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.412964221282427e-06,
+      "loss": 0.5852,
+      "step": 22870
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.412252620204832e-06,
+      "loss": 0.5905,
+      "step": 22871
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.411541022113755e-06,
+      "loss": 0.5628,
+      "step": 22872
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.410829427012808e-06,
+      "loss": 0.6207,
+      "step": 22873
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.41011783490561e-06,
+      "loss": 0.6061,
+      "step": 22874
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.409406245795774e-06,
+      "loss": 0.6468,
+      "step": 22875
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.40869465968692e-06,
+      "loss": 0.5709,
+      "step": 22876
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.407983076582663e-06,
+      "loss": 0.6258,
+      "step": 22877
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.407271496486617e-06,
+      "loss": 0.5684,
+      "step": 22878
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.406559919402396e-06,
+      "loss": 0.6202,
+      "step": 22879
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.405848345333618e-06,
+      "loss": 0.5952,
+      "step": 22880
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.405136774283899e-06,
+      "loss": 0.6162,
+      "step": 22881
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.404425206256854e-06,
+      "loss": 0.5847,
+      "step": 22882
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.403713641256103e-06,
+      "loss": 0.6408,
+      "step": 22883
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.403002079285252e-06,
+      "loss": 0.6561,
+      "step": 22884
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.402290520347924e-06,
+      "loss": 0.6376,
+      "step": 22885
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.401578964447733e-06,
+      "loss": 0.5817,
+      "step": 22886
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.400867411588293e-06,
+      "loss": 0.6111,
+      "step": 22887
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.400155861773225e-06,
+      "loss": 0.6489,
+      "step": 22888
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.39944431500614e-06,
+      "loss": 0.6139,
+      "step": 22889
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.39873277129065e-06,
+      "loss": 0.6221,
+      "step": 22890
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.398021230630378e-06,
+      "loss": 0.5989,
+      "step": 22891
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.397309693028936e-06,
+      "loss": 0.6508,
+      "step": 22892
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.396598158489943e-06,
+      "loss": 0.5594,
+      "step": 22893
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.395886627017008e-06,
+      "loss": 0.6182,
+      "step": 22894
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.395175098613749e-06,
+      "loss": 0.6368,
+      "step": 22895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.394463573283783e-06,
+      "loss": 0.5433,
+      "step": 22896
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.393752051030728e-06,
+      "loss": 0.6055,
+      "step": 22897
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.393040531858194e-06,
+      "loss": 0.6537,
+      "step": 22898
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.3923290157698e-06,
+      "loss": 0.5409,
+      "step": 22899
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.391617502769159e-06,
+      "loss": 0.6303,
+      "step": 22900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.390905992859888e-06,
+      "loss": 0.6212,
+      "step": 22901
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.390194486045602e-06,
+      "loss": 0.6544,
+      "step": 22902
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.389482982329914e-06,
+      "loss": 0.536,
+      "step": 22903
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.38877148171645e-06,
+      "loss": 0.6947,
+      "step": 22904
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.38805998420881e-06,
+      "loss": 0.6042,
+      "step": 22905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.387348489810617e-06,
+      "loss": 0.5629,
+      "step": 22906
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.386636998525485e-06,
+      "loss": 0.5357,
+      "step": 22907
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.385925510357032e-06,
+      "loss": 0.5578,
+      "step": 22908
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.385214025308868e-06,
+      "loss": 0.625,
+      "step": 22909
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.384502543384619e-06,
+      "loss": 0.5651,
+      "step": 22910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.383791064587887e-06,
+      "loss": 0.5927,
+      "step": 22911
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.383079588922294e-06,
+      "loss": 0.5462,
+      "step": 22912
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.382368116391454e-06,
+      "loss": 0.5931,
+      "step": 22913
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.381656646998983e-06,
+      "loss": 0.6362,
+      "step": 22914
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.380945180748498e-06,
+      "loss": 0.6177,
+      "step": 22915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.380233717643608e-06,
+      "loss": 0.6472,
+      "step": 22916
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.379522257687933e-06,
+      "loss": 0.6205,
+      "step": 22917
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.37881080088509e-06,
+      "loss": 0.6439,
+      "step": 22918
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.378099347238688e-06,
+      "loss": 0.6634,
+      "step": 22919
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.377387896752349e-06,
+      "loss": 0.5531,
+      "step": 22920
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.376676449429679e-06,
+      "loss": 0.6187,
+      "step": 22921
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.375965005274304e-06,
+      "loss": 0.6223,
+      "step": 22922
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.375253564289829e-06,
+      "loss": 0.6333,
+      "step": 22923
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.374542126479876e-06,
+      "loss": 0.5708,
+      "step": 22924
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.373830691848055e-06,
+      "loss": 0.593,
+      "step": 22925
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.37311926039799e-06,
+      "loss": 0.6334,
+      "step": 22926
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.372407832133283e-06,
+      "loss": 0.5621,
+      "step": 22927
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.371696407057557e-06,
+      "loss": 0.5673,
+      "step": 22928
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.370984985174426e-06,
+      "loss": 0.6435,
+      "step": 22929
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.370273566487504e-06,
+      "loss": 0.5546,
+      "step": 22930
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.369562151000411e-06,
+      "loss": 0.6021,
+      "step": 22931
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.368850738716753e-06,
+      "loss": 0.6522,
+      "step": 22932
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.36813932964015e-06,
+      "loss": 0.5957,
+      "step": 22933
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.367427923774215e-06,
+      "loss": 0.5993,
+      "step": 22934
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.366716521122564e-06,
+      "loss": 0.633,
+      "step": 22935
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.366005121688814e-06,
+      "loss": 0.6085,
+      "step": 22936
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.365293725476578e-06,
+      "loss": 0.6001,
+      "step": 22937
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.364582332489469e-06,
+      "loss": 0.5559,
+      "step": 22938
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.363870942731106e-06,
+      "loss": 0.5932,
+      "step": 22939
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.363159556205096e-06,
+      "loss": 0.5709,
+      "step": 22940
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.36244817291506e-06,
+      "loss": 0.6183,
+      "step": 22941
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.361736792864617e-06,
+      "loss": 0.5944,
+      "step": 22942
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.36102541605737e-06,
+      "loss": 0.5522,
+      "step": 22943
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.360314042496941e-06,
+      "loss": 0.5626,
+      "step": 22944
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.359602672186944e-06,
+      "loss": 0.6601,
+      "step": 22945
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.358891305130993e-06,
+      "loss": 0.6014,
+      "step": 22946
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.358179941332704e-06,
+      "loss": 0.6162,
+      "step": 22947
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.357468580795695e-06,
+      "loss": 0.6444,
+      "step": 22948
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.35675722352357e-06,
+      "loss": 0.6495,
+      "step": 22949
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.356045869519952e-06,
+      "loss": 0.6536,
+      "step": 22950
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.355334518788452e-06,
+      "loss": 0.5998,
+      "step": 22951
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.354623171332688e-06,
+      "loss": 0.6416,
+      "step": 22952
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.353911827156274e-06,
+      "loss": 0.6072,
+      "step": 22953
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.353200486262822e-06,
+      "loss": 0.6007,
+      "step": 22954
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.352489148655945e-06,
+      "loss": 0.6141,
+      "step": 22955
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.351777814339262e-06,
+      "loss": 0.5846,
+      "step": 22956
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.35106648331639e-06,
+      "loss": 0.4972,
+      "step": 22957
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.350355155590936e-06,
+      "loss": 0.6156,
+      "step": 22958
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.349643831166516e-06,
+      "loss": 0.6264,
+      "step": 22959
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.34893251004675e-06,
+      "loss": 0.5505,
+      "step": 22960
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.348221192235244e-06,
+      "loss": 0.6219,
+      "step": 22961
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.347509877735618e-06,
+      "loss": 0.5855,
+      "step": 22962
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.346798566551485e-06,
+      "loss": 0.6627,
+      "step": 22963
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.346087258686465e-06,
+      "loss": 0.6248,
+      "step": 22964
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.345375954144162e-06,
+      "loss": 0.63,
+      "step": 22965
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.344664652928197e-06,
+      "loss": 0.6202,
+      "step": 22966
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.343953355042182e-06,
+      "loss": 0.5946,
+      "step": 22967
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.343242060489732e-06,
+      "loss": 0.59,
+      "step": 22968
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.342530769274465e-06,
+      "loss": 0.6049,
+      "step": 22969
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.34181948139999e-06,
+      "loss": 0.5922,
+      "step": 22970
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.34110819686992e-06,
+      "loss": 0.6436,
+      "step": 22971
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.340396915687873e-06,
+      "loss": 0.574,
+      "step": 22972
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.339685637857463e-06,
+      "loss": 0.6463,
+      "step": 22973
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.338974363382305e-06,
+      "loss": 0.6158,
+      "step": 22974
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.338263092266012e-06,
+      "loss": 0.5874,
+      "step": 22975
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.337551824512197e-06,
+      "loss": 0.6808,
+      "step": 22976
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.336840560124476e-06,
+      "loss": 0.5842,
+      "step": 22977
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.336129299106461e-06,
+      "loss": 0.5849,
+      "step": 22978
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.335418041461768e-06,
+      "loss": 0.5691,
+      "step": 22979
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.334706787194012e-06,
+      "loss": 0.609,
+      "step": 22980
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.333995536306805e-06,
+      "loss": 0.6085,
+      "step": 22981
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.33328428880376e-06,
+      "loss": 0.6283,
+      "step": 22982
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.332573044688493e-06,
+      "loss": 0.64,
+      "step": 22983
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.331861803964618e-06,
+      "loss": 0.672,
+      "step": 22984
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.331150566635753e-06,
+      "loss": 0.5826,
+      "step": 22985
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.330439332705503e-06,
+      "loss": 0.5602,
+      "step": 22986
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.329728102177488e-06,
+      "loss": 0.601,
+      "step": 22987
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.32901687505532e-06,
+      "loss": 0.6071,
+      "step": 22988
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.328305651342614e-06,
+      "loss": 0.5728,
+      "step": 22989
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.327594431042984e-06,
+      "loss": 0.601,
+      "step": 22990
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.326883214160047e-06,
+      "loss": 0.5641,
+      "step": 22991
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.32617200069741e-06,
+      "loss": 0.6189,
+      "step": 22992
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.32546079065869e-06,
+      "loss": 0.5696,
+      "step": 22993
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.324749584047501e-06,
+      "loss": 0.6161,
+      "step": 22994
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.32403838086746e-06,
+      "loss": 0.6288,
+      "step": 22995
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.323327181122178e-06,
+      "loss": 0.56,
+      "step": 22996
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.322615984815266e-06,
+      "loss": 0.5879,
+      "step": 22997
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.321904791950343e-06,
+      "loss": 0.5808,
+      "step": 22998
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.32119360253102e-06,
+      "loss": 0.622,
+      "step": 22999
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.320482416560907e-06,
+      "loss": 0.6035,
+      "step": 23000
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.319771234043625e-06,
+      "loss": 0.6259,
+      "step": 23001
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.319060054982788e-06,
+      "loss": 0.588,
+      "step": 23002
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.318348879382003e-06,
+      "loss": 0.5317,
+      "step": 23003
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.317637707244886e-06,
+      "loss": 0.6,
+      "step": 23004
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.31692653857505e-06,
+      "loss": 0.5926,
+      "step": 23005
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.316215373376113e-06,
+      "loss": 0.5905,
+      "step": 23006
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.315504211651689e-06,
+      "loss": 0.5843,
+      "step": 23007
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.314793053405385e-06,
+      "loss": 0.6263,
+      "step": 23008
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.314081898640817e-06,
+      "loss": 0.6386,
+      "step": 23009
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.3133707473616e-06,
+      "loss": 0.6053,
+      "step": 23010
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.312659599571348e-06,
+      "loss": 0.6023,
+      "step": 23011
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.311948455273678e-06,
+      "loss": 0.6107,
+      "step": 23012
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.311237314472194e-06,
+      "loss": 0.5992,
+      "step": 23013
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.310526177170517e-06,
+      "loss": 0.6225,
+      "step": 23014
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.309815043372257e-06,
+      "loss": 0.6045,
+      "step": 23015
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.309103913081031e-06,
+      "loss": 0.5985,
+      "step": 23016
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.308392786300448e-06,
+      "loss": 0.5203,
+      "step": 23017
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.307681663034127e-06,
+      "loss": 0.5968,
+      "step": 23018
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.306970543285678e-06,
+      "loss": 0.6582,
+      "step": 23019
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.306259427058712e-06,
+      "loss": 0.6236,
+      "step": 23020
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.305548314356845e-06,
+      "loss": 0.6097,
+      "step": 23021
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.304837205183691e-06,
+      "loss": 0.6561,
+      "step": 23022
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.304126099542865e-06,
+      "loss": 0.6053,
+      "step": 23023
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.303414997437977e-06,
+      "loss": 0.6575,
+      "step": 23024
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.302703898872639e-06,
+      "loss": 0.6058,
+      "step": 23025
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.301992803850469e-06,
+      "loss": 0.6291,
+      "step": 23026
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.301281712375075e-06,
+      "loss": 0.6185,
+      "step": 23027
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.300570624450077e-06,
+      "loss": 0.571,
+      "step": 23028
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.299859540079085e-06,
+      "loss": 0.6045,
+      "step": 23029
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.29914845926571e-06,
+      "loss": 0.6268,
+      "step": 23030
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.298437382013565e-06,
+      "loss": 0.6176,
+      "step": 23031
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.297726308326266e-06,
+      "loss": 0.6102,
+      "step": 23032
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.297015238207427e-06,
+      "loss": 0.6443,
+      "step": 23033
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.296304171660661e-06,
+      "loss": 0.6077,
+      "step": 23034
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.295593108689576e-06,
+      "loss": 0.6129,
+      "step": 23035
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.294882049297789e-06,
+      "loss": 0.6444,
+      "step": 23036
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.294170993488916e-06,
+      "loss": 0.6499,
+      "step": 23037
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.293459941266563e-06,
+      "loss": 0.6011,
+      "step": 23038
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.292748892634348e-06,
+      "loss": 0.619,
+      "step": 23039
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.292037847595887e-06,
+      "loss": 0.5912,
+      "step": 23040
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.291326806154785e-06,
+      "loss": 0.6131,
+      "step": 23041
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.290615768314658e-06,
+      "loss": 0.6449,
+      "step": 23042
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.289904734079122e-06,
+      "loss": 0.5858,
+      "step": 23043
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.289193703451786e-06,
+      "loss": 0.5432,
+      "step": 23044
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.28848267643627e-06,
+      "loss": 0.5637,
+      "step": 23045
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.287771653036178e-06,
+      "loss": 0.6687,
+      "step": 23046
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.287060633255127e-06,
+      "loss": 0.5967,
+      "step": 23047
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.28634961709673e-06,
+      "loss": 0.6126,
+      "step": 23048
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.285638604564598e-06,
+      "loss": 0.5945,
+      "step": 23049
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.284927595662351e-06,
+      "loss": 0.6364,
+      "step": 23050
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.284216590393591e-06,
+      "loss": 0.6217,
+      "step": 23051
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.283505588761938e-06,
+      "loss": 0.6067,
+      "step": 23052
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.282794590771e-06,
+      "loss": 0.6074,
+      "step": 23053
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.282083596424396e-06,
+      "loss": 0.5841,
+      "step": 23054
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.281372605725735e-06,
+      "loss": 0.6094,
+      "step": 23055
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.280661618678632e-06,
+      "loss": 0.545,
+      "step": 23056
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.279950635286693e-06,
+      "loss": 0.5916,
+      "step": 23057
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.27923965555354e-06,
+      "loss": 0.5931,
+      "step": 23058
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.27852867948278e-06,
+      "loss": 0.6393,
+      "step": 23059
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.277817707078028e-06,
+      "loss": 0.6075,
+      "step": 23060
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.277106738342899e-06,
+      "loss": 0.6372,
+      "step": 23061
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.276395773280997e-06,
+      "loss": 0.6059,
+      "step": 23062
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.27568481189594e-06,
+      "loss": 0.6174,
+      "step": 23063
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.274973854191341e-06,
+      "loss": 0.6564,
+      "step": 23064
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.274262900170812e-06,
+      "loss": 0.5653,
+      "step": 23065
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.273551949837966e-06,
+      "loss": 0.5797,
+      "step": 23066
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.27284100319642e-06,
+      "loss": 0.5786,
+      "step": 23067
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.272130060249776e-06,
+      "loss": 0.6554,
+      "step": 23068
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.271419121001654e-06,
+      "loss": 0.6351,
+      "step": 23069
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.270708185455664e-06,
+      "loss": 0.5911,
+      "step": 23070
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.26999725361542e-06,
+      "loss": 0.5889,
+      "step": 23071
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.269286325484537e-06,
+      "loss": 0.6194,
+      "step": 23072
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.26857540106662e-06,
+      "loss": 0.6069,
+      "step": 23073
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.267864480365286e-06,
+      "loss": 0.6035,
+      "step": 23074
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.26715356338415e-06,
+      "loss": 0.5908,
+      "step": 23075
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.266442650126818e-06,
+      "loss": 0.5984,
+      "step": 23076
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.265731740596908e-06,
+      "loss": 0.5123,
+      "step": 23077
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.26502083479803e-06,
+      "loss": 0.6023,
+      "step": 23078
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.264309932733794e-06,
+      "loss": 0.5852,
+      "step": 23079
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.263599034407816e-06,
+      "loss": 0.607,
+      "step": 23080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.262888139823707e-06,
+      "loss": 0.6038,
+      "step": 23081
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.262177248985079e-06,
+      "loss": 0.5855,
+      "step": 23082
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.261466361895548e-06,
+      "loss": 0.6115,
+      "step": 23083
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.260755478558719e-06,
+      "loss": 0.652,
+      "step": 23084
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.260044598978207e-06,
+      "loss": 0.6207,
+      "step": 23085
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.259333723157624e-06,
+      "loss": 0.5895,
+      "step": 23086
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.258622851100586e-06,
+      "loss": 0.5276,
+      "step": 23087
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.257911982810706e-06,
+      "loss": 0.5718,
+      "step": 23088
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.257201118291588e-06,
+      "loss": 0.6256,
+      "step": 23089
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.256490257546849e-06,
+      "loss": 0.6432,
+      "step": 23090
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.2557794005801e-06,
+      "loss": 0.5771,
+      "step": 23091
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.255068547394954e-06,
+      "loss": 0.6301,
+      "step": 23092
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.254357697995025e-06,
+      "loss": 0.6741,
+      "step": 23093
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.253646852383924e-06,
+      "loss": 0.6162,
+      "step": 23094
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.252936010565258e-06,
+      "loss": 0.5619,
+      "step": 23095
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.252225172542645e-06,
+      "loss": 0.5953,
+      "step": 23096
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.251514338319695e-06,
+      "loss": 0.6155,
+      "step": 23097
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.250803507900018e-06,
+      "loss": 0.6123,
+      "step": 23098
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.250092681287232e-06,
+      "loss": 0.5835,
+      "step": 23099
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.24938185848494e-06,
+      "loss": 0.644,
+      "step": 23100
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.24867103949676e-06,
+      "loss": 0.6354,
+      "step": 23101
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.247960224326304e-06,
+      "loss": 0.5482,
+      "step": 23102
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.247249412977179e-06,
+      "loss": 0.6441,
+      "step": 23103
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.246538605453007e-06,
+      "loss": 0.6213,
+      "step": 23104
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.245827801757387e-06,
+      "loss": 0.5583,
+      "step": 23105
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.245117001893937e-06,
+      "loss": 0.6046,
+      "step": 23106
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.24440620586627e-06,
+      "loss": 0.5945,
+      "step": 23107
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.243695413677995e-06,
+      "loss": 0.5576,
+      "step": 23108
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.242984625332725e-06,
+      "loss": 0.591,
+      "step": 23109
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.242273840834078e-06,
+      "loss": 0.6258,
+      "step": 23110
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.241563060185654e-06,
+      "loss": 0.5998,
+      "step": 23111
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.240852283391068e-06,
+      "loss": 0.5847,
+      "step": 23112
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.240141510453937e-06,
+      "loss": 0.6037,
+      "step": 23113
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.239430741377871e-06,
+      "loss": 0.5601,
+      "step": 23114
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.23871997616648e-06,
+      "loss": 0.5873,
+      "step": 23115
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.238009214823373e-06,
+      "loss": 0.5604,
+      "step": 23116
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.237298457352167e-06,
+      "loss": 0.6629,
+      "step": 23117
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.236587703756467e-06,
+      "loss": 0.5933,
+      "step": 23118
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.23587695403989e-06,
+      "loss": 0.5871,
+      "step": 23119
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.235166208206046e-06,
+      "loss": 0.6159,
+      "step": 23120
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.234455466258552e-06,
+      "loss": 0.566,
+      "step": 23121
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.23374472820101e-06,
+      "loss": 0.5689,
+      "step": 23122
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.233033994037031e-06,
+      "loss": 0.6123,
+      "step": 23123
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.232323263770235e-06,
+      "loss": 0.5314,
+      "step": 23124
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.231612537404227e-06,
+      "loss": 0.5911,
+      "step": 23125
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.230901814942627e-06,
+      "loss": 0.5998,
+      "step": 23126
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.230191096389035e-06,
+      "loss": 0.5891,
+      "step": 23127
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.229480381747067e-06,
+      "loss": 0.619,
+      "step": 23128
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.228769671020335e-06,
+      "loss": 0.6423,
+      "step": 23129
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.228058964212452e-06,
+      "loss": 0.6073,
+      "step": 23130
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.227348261327027e-06,
+      "loss": 0.5725,
+      "step": 23131
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.226637562367673e-06,
+      "loss": 0.6129,
+      "step": 23132
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.225926867337998e-06,
+      "loss": 0.5489,
+      "step": 23133
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.225216176241617e-06,
+      "loss": 0.5644,
+      "step": 23134
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.224505489082138e-06,
+      "loss": 0.533,
+      "step": 23135
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.223794805863173e-06,
+      "loss": 0.6292,
+      "step": 23136
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.223084126588336e-06,
+      "loss": 0.5891,
+      "step": 23137
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.222373451261235e-06,
+      "loss": 0.5593,
+      "step": 23138
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.221662779885482e-06,
+      "loss": 0.5982,
+      "step": 23139
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.22095211246469e-06,
+      "loss": 0.5793,
+      "step": 23140
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.220241449002464e-06,
+      "loss": 0.6188,
+      "step": 23141
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.219530789502428e-06,
+      "loss": 0.6305,
+      "step": 23142
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.218820133968178e-06,
+      "loss": 0.6485,
+      "step": 23143
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.218109482403333e-06,
+      "loss": 0.6098,
+      "step": 23144
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.2173988348115e-06,
+      "loss": 0.5647,
+      "step": 23145
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.216688191196295e-06,
+      "loss": 0.6428,
+      "step": 23146
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.215977551561328e-06,
+      "loss": 0.5956,
+      "step": 23147
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.21526691591021e-06,
+      "loss": 0.5667,
+      "step": 23148
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.214556284246547e-06,
+      "loss": 0.6093,
+      "step": 23149
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.213845656573955e-06,
+      "loss": 0.5781,
+      "step": 23150
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.213135032896043e-06,
+      "loss": 0.632,
+      "step": 23151
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.212424413216423e-06,
+      "loss": 0.6011,
+      "step": 23152
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.211713797538707e-06,
+      "loss": 0.6428,
+      "step": 23153
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.211003185866502e-06,
+      "loss": 0.6413,
+      "step": 23154
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.210292578203421e-06,
+      "loss": 0.6132,
+      "step": 23155
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.209581974553075e-06,
+      "loss": 0.6051,
+      "step": 23156
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.208871374919073e-06,
+      "loss": 0.6031,
+      "step": 23157
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.208160779305026e-06,
+      "loss": 0.589,
+      "step": 23158
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.207450187714552e-06,
+      "loss": 0.5945,
+      "step": 23159
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.206739600151252e-06,
+      "loss": 0.5405,
+      "step": 23160
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.20602901661874e-06,
+      "loss": 0.5932,
+      "step": 23161
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.205318437120626e-06,
+      "loss": 0.5725,
+      "step": 23162
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.204607861660523e-06,
+      "loss": 0.5945,
+      "step": 23163
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.203897290242045e-06,
+      "loss": 0.5848,
+      "step": 23164
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.203186722868792e-06,
+      "loss": 0.5845,
+      "step": 23165
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.202476159544381e-06,
+      "loss": 0.5815,
+      "step": 23166
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.201765600272424e-06,
+      "loss": 0.6278,
+      "step": 23167
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.201055045056528e-06,
+      "loss": 0.6102,
+      "step": 23168
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.20034449390031e-06,
+      "loss": 0.56,
+      "step": 23169
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.19963394680737e-06,
+      "loss": 0.6538,
+      "step": 23170
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.198923403781326e-06,
+      "loss": 0.5885,
+      "step": 23171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.198212864825788e-06,
+      "loss": 0.694,
+      "step": 23172
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.197502329944366e-06,
+      "loss": 0.5892,
+      "step": 23173
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.196791799140668e-06,
+      "loss": 0.6246,
+      "step": 23174
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.196081272418306e-06,
+      "loss": 0.5739,
+      "step": 23175
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.195370749780893e-06,
+      "loss": 0.5994,
+      "step": 23176
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.194660231232035e-06,
+      "loss": 0.5835,
+      "step": 23177
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.193949716775342e-06,
+      "loss": 0.5632,
+      "step": 23178
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.193239206414428e-06,
+      "loss": 0.6167,
+      "step": 23179
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.192528700152906e-06,
+      "loss": 0.6137,
+      "step": 23180
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.191818197994378e-06,
+      "loss": 0.5898,
+      "step": 23181
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.191107699942459e-06,
+      "loss": 0.5869,
+      "step": 23182
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.190397206000757e-06,
+      "loss": 0.6279,
+      "step": 23183
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.189686716172885e-06,
+      "loss": 0.5669,
+      "step": 23184
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.188976230462451e-06,
+      "loss": 0.6092,
+      "step": 23185
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.188265748873071e-06,
+      "loss": 0.6328,
+      "step": 23186
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.187555271408349e-06,
+      "loss": 0.5788,
+      "step": 23187
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.186844798071894e-06,
+      "loss": 0.6021,
+      "step": 23188
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.186134328867318e-06,
+      "loss": 0.579,
+      "step": 23189
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.185423863798235e-06,
+      "loss": 0.5587,
+      "step": 23190
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.184713402868252e-06,
+      "loss": 0.606,
+      "step": 23191
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.184002946080976e-06,
+      "loss": 0.5283,
+      "step": 23192
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.183292493440021e-06,
+      "loss": 0.6516,
+      "step": 23193
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.182582044949e-06,
+      "loss": 0.5452,
+      "step": 23194
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.181871600611513e-06,
+      "loss": 0.5646,
+      "step": 23195
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.18116116043118e-06,
+      "loss": 0.5938,
+      "step": 23196
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.180450724411607e-06,
+      "loss": 0.5466,
+      "step": 23197
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.179740292556402e-06,
+      "loss": 0.6382,
+      "step": 23198
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.179029864869175e-06,
+      "loss": 0.6608,
+      "step": 23199
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.178319441353541e-06,
+      "loss": 0.6019,
+      "step": 23200
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.177609022013104e-06,
+      "loss": 0.574,
+      "step": 23201
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.176898606851481e-06,
+      "loss": 0.6099,
+      "step": 23202
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.176188195872274e-06,
+      "loss": 0.5846,
+      "step": 23203
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.175477789079095e-06,
+      "loss": 0.5987,
+      "step": 23204
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.174767386475556e-06,
+      "loss": 0.5596,
+      "step": 23205
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.174056988065263e-06,
+      "loss": 0.5927,
+      "step": 23206
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.173346593851835e-06,
+      "loss": 0.6401,
+      "step": 23207
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.17263620383887e-06,
+      "loss": 0.5631,
+      "step": 23208
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.171925818029983e-06,
+      "loss": 0.5293,
+      "step": 23209
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.171215436428783e-06,
+      "loss": 0.6038,
+      "step": 23210
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.170505059038883e-06,
+      "loss": 0.5833,
+      "step": 23211
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.169794685863887e-06,
+      "loss": 0.595,
+      "step": 23212
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.169084316907409e-06,
+      "loss": 0.5908,
+      "step": 23213
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.168373952173058e-06,
+      "loss": 0.5851,
+      "step": 23214
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.167663591664439e-06,
+      "loss": 0.5665,
+      "step": 23215
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.166953235385166e-06,
+      "loss": 0.6546,
+      "step": 23216
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.166242883338848e-06,
+      "loss": 0.6184,
+      "step": 23217
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.165532535529098e-06,
+      "loss": 0.5887,
+      "step": 23218
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.164822191959517e-06,
+      "loss": 0.6454,
+      "step": 23219
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.16411185263372e-06,
+      "loss": 0.5655,
+      "step": 23220
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.163401517555315e-06,
+      "loss": 0.6452,
+      "step": 23221
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.162691186727912e-06,
+      "loss": 0.5383,
+      "step": 23222
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.161980860155119e-06,
+      "loss": 0.5631,
+      "step": 23223
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.161270537840552e-06,
+      "loss": 0.631,
+      "step": 23224
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.16056021978781e-06,
+      "loss": 0.632,
+      "step": 23225
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.159849906000509e-06,
+      "loss": 0.6159,
+      "step": 23226
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.159139596482255e-06,
+      "loss": 0.5684,
+      "step": 23227
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.15842929123666e-06,
+      "loss": 0.592,
+      "step": 23228
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.157718990267336e-06,
+      "loss": 0.6176,
+      "step": 23229
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.157008693577883e-06,
+      "loss": 0.6568,
+      "step": 23230
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.156298401171918e-06,
+      "loss": 0.5788,
+      "step": 23231
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.155588113053048e-06,
+      "loss": 0.5674,
+      "step": 23232
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.154877829224882e-06,
+      "loss": 0.583,
+      "step": 23233
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.154167549691031e-06,
+      "loss": 0.5615,
+      "step": 23234
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.153457274455103e-06,
+      "loss": 0.5448,
+      "step": 23235
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.1527470035207e-06,
+      "loss": 0.5624,
+      "step": 23236
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.152036736891442e-06,
+      "loss": 0.5862,
+      "step": 23237
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.151326474570933e-06,
+      "loss": 0.6337,
+      "step": 23238
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.150616216562783e-06,
+      "loss": 0.6371,
+      "step": 23239
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.149905962870604e-06,
+      "loss": 0.5966,
+      "step": 23240
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.149195713497997e-06,
+      "loss": 0.5932,
+      "step": 23241
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.148485468448578e-06,
+      "loss": 0.5964,
+      "step": 23242
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.14777522772595e-06,
+      "loss": 0.5923,
+      "step": 23243
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.147064991333728e-06,
+      "loss": 0.6128,
+      "step": 23244
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.146354759275523e-06,
+      "loss": 0.6054,
+      "step": 23245
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.145644531554936e-06,
+      "loss": 0.6211,
+      "step": 23246
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.144934308175576e-06,
+      "loss": 0.6115,
+      "step": 23247
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.144224089141058e-06,
+      "loss": 0.6112,
+      "step": 23248
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.143513874454987e-06,
+      "loss": 0.6529,
+      "step": 23249
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.142803664120978e-06,
+      "loss": 0.6205,
+      "step": 23250
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.142093458142631e-06,
+      "loss": 0.5732,
+      "step": 23251
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.141383256523558e-06,
+      "loss": 0.5684,
+      "step": 23252
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.140673059267368e-06,
+      "loss": 0.6952,
+      "step": 23253
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.13996286637767e-06,
+      "loss": 0.6035,
+      "step": 23254
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.139252677858072e-06,
+      "loss": 0.5453,
+      "step": 23255
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.138542493712187e-06,
+      "loss": 0.6133,
+      "step": 23256
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.137832313943616e-06,
+      "loss": 0.573,
+      "step": 23257
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.137122138555972e-06,
+      "loss": 0.6419,
+      "step": 23258
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.136411967552864e-06,
+      "loss": 0.5801,
+      "step": 23259
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.135701800937896e-06,
+      "loss": 0.5283,
+      "step": 23260
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.134991638714688e-06,
+      "loss": 0.5911,
+      "step": 23261
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.134281480886835e-06,
+      "loss": 0.6227,
+      "step": 23262
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.133571327457953e-06,
+      "loss": 0.5915,
+      "step": 23263
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.132861178431646e-06,
+      "loss": 0.595,
+      "step": 23264
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.132151033811527e-06,
+      "loss": 0.5159,
+      "step": 23265
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.131440893601202e-06,
+      "loss": 0.6399,
+      "step": 23266
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.130730757804285e-06,
+      "loss": 0.5932,
+      "step": 23267
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.130020626424377e-06,
+      "loss": 0.5804,
+      "step": 23268
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.129310499465086e-06,
+      "loss": 0.5553,
+      "step": 23269
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.128600376930027e-06,
+      "loss": 0.591,
+      "step": 23270
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.127890258822803e-06,
+      "loss": 0.6406,
+      "step": 23271
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.127180145147025e-06,
+      "loss": 0.6417,
+      "step": 23272
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.126470035906299e-06,
+      "loss": 0.5826,
+      "step": 23273
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.125759931104236e-06,
+      "loss": 0.5831,
+      "step": 23274
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.125049830744442e-06,
+      "loss": 0.5931,
+      "step": 23275
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.124339734830525e-06,
+      "loss": 0.672,
+      "step": 23276
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.123629643366094e-06,
+      "loss": 0.6568,
+      "step": 23277
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.122919556354763e-06,
+      "loss": 0.6053,
+      "step": 23278
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.122209473800129e-06,
+      "loss": 0.5943,
+      "step": 23279
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.121499395705807e-06,
+      "loss": 0.5848,
+      "step": 23280
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.120789322075402e-06,
+      "loss": 0.5892,
+      "step": 23281
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.120079252912525e-06,
+      "loss": 0.5887,
+      "step": 23282
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.119369188220787e-06,
+      "loss": 0.612,
+      "step": 23283
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.11865912800379e-06,
+      "loss": 0.5412,
+      "step": 23284
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.11794907226514e-06,
+      "loss": 0.6635,
+      "step": 23285
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.117239021008451e-06,
+      "loss": 0.5929,
+      "step": 23286
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.11652897423733e-06,
+      "loss": 0.5976,
+      "step": 23287
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.115818931955387e-06,
+      "loss": 0.6252,
+      "step": 23288
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.115108894166223e-06,
+      "loss": 0.5767,
+      "step": 23289
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.11439886087345e-06,
+      "loss": 0.6212,
+      "step": 23290
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.113688832080679e-06,
+      "loss": 0.6008,
+      "step": 23291
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.11297880779151e-06,
+      "loss": 0.6122,
+      "step": 23292
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.112268788009559e-06,
+      "loss": 0.5724,
+      "step": 23293
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.11155877273843e-06,
+      "loss": 0.5411,
+      "step": 23294
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.110848761981731e-06,
+      "loss": 0.6433,
+      "step": 23295
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.110138755743068e-06,
+      "loss": 0.5905,
+      "step": 23296
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.10942875402605e-06,
+      "loss": 0.5353,
+      "step": 23297
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.108718756834286e-06,
+      "loss": 0.648,
+      "step": 23298
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.108008764171387e-06,
+      "loss": 0.5878,
+      "step": 23299
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.107298776040954e-06,
+      "loss": 0.6632,
+      "step": 23300
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.106588792446596e-06,
+      "loss": 0.6313,
+      "step": 23301
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.105878813391923e-06,
+      "loss": 0.5999,
+      "step": 23302
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.10516883888054e-06,
+      "loss": 0.6027,
+      "step": 23303
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.104458868916056e-06,
+      "loss": 0.6364,
+      "step": 23304
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.103748903502085e-06,
+      "loss": 0.6294,
+      "step": 23305
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.103038942642221e-06,
+      "loss": 0.5205,
+      "step": 23306
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.102328986340083e-06,
+      "loss": 0.6123,
+      "step": 23307
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.101619034599271e-06,
+      "loss": 0.6393,
+      "step": 23308
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.1009090874234e-06,
+      "loss": 0.594,
+      "step": 23309
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.100199144816071e-06,
+      "loss": 0.6556,
+      "step": 23310
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.099489206780893e-06,
+      "loss": 0.5549,
+      "step": 23311
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.098779273321477e-06,
+      "loss": 0.5605,
+      "step": 23312
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.098069344441424e-06,
+      "loss": 0.6645,
+      "step": 23313
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.097359420144346e-06,
+      "loss": 0.6439,
+      "step": 23314
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.096649500433846e-06,
+      "loss": 0.6286,
+      "step": 23315
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.095939585313541e-06,
+      "loss": 0.6032,
+      "step": 23316
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.095229674787029e-06,
+      "loss": 0.6001,
+      "step": 23317
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.094519768857916e-06,
+      "loss": 0.6376,
+      "step": 23318
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.093809867529817e-06,
+      "loss": 0.6183,
+      "step": 23319
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.093099970806335e-06,
+      "loss": 0.6006,
+      "step": 23320
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.09239007869108e-06,
+      "loss": 0.5596,
+      "step": 23321
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.091680191187654e-06,
+      "loss": 0.626,
+      "step": 23322
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.090970308299665e-06,
+      "loss": 0.5567,
+      "step": 23323
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.090260430030725e-06,
+      "loss": 0.6164,
+      "step": 23324
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.089550556384437e-06,
+      "loss": 0.6156,
+      "step": 23325
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.088840687364412e-06,
+      "loss": 0.5986,
+      "step": 23326
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.088130822974253e-06,
+      "loss": 0.5841,
+      "step": 23327
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.087420963217565e-06,
+      "loss": 0.6149,
+      "step": 23328
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.086711108097961e-06,
+      "loss": 0.6541,
+      "step": 23329
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.086001257619045e-06,
+      "loss": 0.5648,
+      "step": 23330
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.085291411784425e-06,
+      "loss": 0.579,
+      "step": 23331
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.084581570597708e-06,
+      "loss": 0.5717,
+      "step": 23332
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.0838717340625e-06,
+      "loss": 0.6208,
+      "step": 23333
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.083161902182407e-06,
+      "loss": 0.5915,
+      "step": 23334
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.082452074961035e-06,
+      "loss": 0.6091,
+      "step": 23335
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.081742252401994e-06,
+      "loss": 0.6502,
+      "step": 23336
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.081032434508894e-06,
+      "loss": 0.5699,
+      "step": 23337
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.080322621285333e-06,
+      "loss": 0.5977,
+      "step": 23338
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.079612812734921e-06,
+      "loss": 0.6129,
+      "step": 23339
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.078903008861266e-06,
+      "loss": 0.5853,
+      "step": 23340
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.078193209667974e-06,
+      "loss": 0.5953,
+      "step": 23341
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.077483415158655e-06,
+      "loss": 0.6548,
+      "step": 23342
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.076773625336914e-06,
+      "loss": 0.5975,
+      "step": 23343
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.076063840206355e-06,
+      "loss": 0.6183,
+      "step": 23344
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.075354059770586e-06,
+      "loss": 0.5985,
+      "step": 23345
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.074644284033212e-06,
+      "loss": 0.6577,
+      "step": 23346
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.073934512997843e-06,
+      "loss": 0.533,
+      "step": 23347
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.073224746668087e-06,
+      "loss": 0.5289,
+      "step": 23348
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.072514985047542e-06,
+      "loss": 0.5543,
+      "step": 23349
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.071805228139822e-06,
+      "loss": 0.604,
+      "step": 23350
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.071095475948533e-06,
+      "loss": 0.5772,
+      "step": 23351
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.070385728477279e-06,
+      "loss": 0.585,
+      "step": 23352
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.069675985729668e-06,
+      "loss": 0.6565,
+      "step": 23353
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.068966247709306e-06,
+      "loss": 0.6113,
+      "step": 23354
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.0682565144198e-06,
+      "loss": 0.5778,
+      "step": 23355
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.067546785864751e-06,
+      "loss": 0.5609,
+      "step": 23356
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.066837062047771e-06,
+      "loss": 0.5855,
+      "step": 23357
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.066127342972466e-06,
+      "loss": 0.6452,
+      "step": 23358
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.065417628642447e-06,
+      "loss": 0.6147,
+      "step": 23359
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.06470791906131e-06,
+      "loss": 0.5902,
+      "step": 23360
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.063998214232663e-06,
+      "loss": 0.6472,
+      "step": 23361
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.063288514160118e-06,
+      "loss": 0.5569,
+      "step": 23362
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.062578818847277e-06,
+      "loss": 0.5943,
+      "step": 23363
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.061869128297754e-06,
+      "loss": 0.6105,
+      "step": 23364
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.061159442515142e-06,
+      "loss": 0.6241,
+      "step": 23365
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.060449761503055e-06,
+      "loss": 0.5813,
+      "step": 23366
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.059740085265098e-06,
+      "loss": 0.6067,
+      "step": 23367
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.05903041380488e-06,
+      "loss": 0.6257,
+      "step": 23368
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.058320747126e-06,
+      "loss": 0.6237,
+      "step": 23369
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.05761108523207e-06,
+      "loss": 0.551,
+      "step": 23370
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.056901428126695e-06,
+      "loss": 0.6464,
+      "step": 23371
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.056191775813479e-06,
+      "loss": 0.5918,
+      "step": 23372
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.055482128296027e-06,
+      "loss": 0.592,
+      "step": 23373
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.054772485577948e-06,
+      "loss": 0.5965,
+      "step": 23374
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.054062847662851e-06,
+      "loss": 0.6237,
+      "step": 23375
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.053353214554333e-06,
+      "loss": 0.6596,
+      "step": 23376
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.052643586256005e-06,
+      "loss": 0.6955,
+      "step": 23377
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.051933962771472e-06,
+      "loss": 0.6295,
+      "step": 23378
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.051224344104341e-06,
+      "loss": 0.627,
+      "step": 23379
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.050514730258221e-06,
+      "loss": 0.6039,
+      "step": 23380
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.049805121236708e-06,
+      "loss": 0.5708,
+      "step": 23381
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.049095517043416e-06,
+      "loss": 0.5821,
+      "step": 23382
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.048385917681946e-06,
+      "loss": 0.6588,
+      "step": 23383
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.047676323155908e-06,
+      "loss": 0.5995,
+      "step": 23384
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.046966733468902e-06,
+      "loss": 0.6203,
+      "step": 23385
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.046257148624544e-06,
+      "loss": 0.5994,
+      "step": 23386
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.045547568626428e-06,
+      "loss": 0.5483,
+      "step": 23387
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.044837993478166e-06,
+      "loss": 0.5866,
+      "step": 23388
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.044128423183361e-06,
+      "loss": 0.5444,
+      "step": 23389
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.04341885774562e-06,
+      "loss": 0.6618,
+      "step": 23390
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.042709297168549e-06,
+      "loss": 0.6671,
+      "step": 23391
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.041999741455752e-06,
+      "loss": 0.5708,
+      "step": 23392
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.041290190610831e-06,
+      "loss": 0.5939,
+      "step": 23393
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.040580644637397e-06,
+      "loss": 0.6599,
+      "step": 23394
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.039871103539055e-06,
+      "loss": 0.5862,
+      "step": 23395
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.039161567319408e-06,
+      "loss": 0.6199,
+      "step": 23396
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.038452035982067e-06,
+      "loss": 0.6125,
+      "step": 23397
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.037742509530628e-06,
+      "loss": 0.5604,
+      "step": 23398
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.037032987968702e-06,
+      "loss": 0.586,
+      "step": 23399
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.036323471299892e-06,
+      "loss": 0.6029,
+      "step": 23400
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.035613959527806e-06,
+      "loss": 0.61,
+      "step": 23401
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.034904452656052e-06,
+      "loss": 0.6247,
+      "step": 23402
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.034194950688227e-06,
+      "loss": 0.6712,
+      "step": 23403
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.033485453627939e-06,
+      "loss": 0.566,
+      "step": 23404
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.032775961478795e-06,
+      "loss": 0.6003,
+      "step": 23405
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.032066474244402e-06,
+      "loss": 0.6353,
+      "step": 23406
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.03135699192836e-06,
+      "loss": 0.5513,
+      "step": 23407
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.03064751453428e-06,
+      "loss": 0.5283,
+      "step": 23408
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.029938042065762e-06,
+      "loss": 0.6823,
+      "step": 23409
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.029228574526414e-06,
+      "loss": 0.6526,
+      "step": 23410
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.028519111919838e-06,
+      "loss": 0.6394,
+      "step": 23411
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.027809654249642e-06,
+      "loss": 0.5548,
+      "step": 23412
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.027100201519432e-06,
+      "loss": 0.6442,
+      "step": 23413
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.026390753732808e-06,
+      "loss": 0.5988,
+      "step": 23414
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.025681310893378e-06,
+      "loss": 0.6206,
+      "step": 23415
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.024971873004746e-06,
+      "loss": 0.6011,
+      "step": 23416
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.024262440070517e-06,
+      "loss": 0.615,
+      "step": 23417
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.023553012094302e-06,
+      "loss": 0.5805,
+      "step": 23418
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.022843589079694e-06,
+      "loss": 0.621,
+      "step": 23419
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.022134171030305e-06,
+      "loss": 0.6133,
+      "step": 23420
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.021424757949739e-06,
+      "loss": 0.6091,
+      "step": 23421
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.0207153498416e-06,
+      "loss": 0.6262,
+      "step": 23422
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.020005946709491e-06,
+      "loss": 0.6095,
+      "step": 23423
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.019296548557027e-06,
+      "loss": 0.6336,
+      "step": 23424
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.018587155387798e-06,
+      "loss": 0.542,
+      "step": 23425
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.017877767205414e-06,
+      "loss": 0.5846,
+      "step": 23426
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.017168384013484e-06,
+      "loss": 0.604,
+      "step": 23427
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.016459005815608e-06,
+      "loss": 0.5515,
+      "step": 23428
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.015749632615393e-06,
+      "loss": 0.6327,
+      "step": 23429
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.01504026441644e-06,
+      "loss": 0.5485,
+      "step": 23430
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.01433090122236e-06,
+      "loss": 0.5922,
+      "step": 23431
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.013621543036748e-06,
+      "loss": 0.5766,
+      "step": 23432
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.012912189863215e-06,
+      "loss": 0.5608,
+      "step": 23433
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.012202841705364e-06,
+      "loss": 0.6019,
+      "step": 23434
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.011493498566807e-06,
+      "loss": 0.5157,
+      "step": 23435
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.010784160451133e-06,
+      "loss": 0.61,
+      "step": 23436
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.010074827361956e-06,
+      "loss": 0.5869,
+      "step": 23437
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.00936549930288e-06,
+      "loss": 0.6259,
+      "step": 23438
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.008656176277506e-06,
+      "loss": 0.638,
+      "step": 23439
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.007946858289444e-06,
+      "loss": 0.5619,
+      "step": 23440
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.007237545342292e-06,
+      "loss": 0.5556,
+      "step": 23441
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.006528237439655e-06,
+      "loss": 0.6162,
+      "step": 23442
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.005818934585141e-06,
+      "loss": 0.5929,
+      "step": 23443
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.005109636782352e-06,
+      "loss": 0.5349,
+      "step": 23444
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.004400344034896e-06,
+      "loss": 0.6152,
+      "step": 23445
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.003691056346368e-06,
+      "loss": 0.6546,
+      "step": 23446
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.002981773720379e-06,
+      "loss": 0.6403,
+      "step": 23447
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.002272496160533e-06,
+      "loss": 0.5599,
+      "step": 23448
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.001563223670431e-06,
+      "loss": 0.5767,
+      "step": 23449
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.000853956253679e-06,
+      "loss": 0.5674,
+      "step": 23450
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.000144693913884e-06,
+      "loss": 0.5925,
+      "step": 23451
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.999435436654642e-06,
+      "loss": 0.6235,
+      "step": 23452
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.998726184479563e-06,
+      "loss": 0.6558,
+      "step": 23453
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.998016937392248e-06,
+      "loss": 0.585,
+      "step": 23454
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.997307695396303e-06,
+      "loss": 0.6271,
+      "step": 23455
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.996598458495337e-06,
+      "loss": 0.6554,
+      "step": 23456
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.995889226692942e-06,
+      "loss": 0.6814,
+      "step": 23457
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.99517999999273e-06,
+      "loss": 0.5376,
+      "step": 23458
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.994470778398299e-06,
+      "loss": 0.5471,
+      "step": 23459
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.993761561913258e-06,
+      "loss": 0.5529,
+      "step": 23460
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.993052350541211e-06,
+      "loss": 0.5673,
+      "step": 23461
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.992343144285764e-06,
+      "loss": 0.5609,
+      "step": 23462
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.991633943150511e-06,
+      "loss": 0.6654,
+      "step": 23463
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.99092474713906e-06,
+      "loss": 0.5926,
+      "step": 23464
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.990215556255019e-06,
+      "loss": 0.6002,
+      "step": 23465
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.989506370501988e-06,
+      "loss": 0.5884,
+      "step": 23466
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.988797189883573e-06,
+      "loss": 0.5964,
+      "step": 23467
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.988088014403372e-06,
+      "loss": 0.5778,
+      "step": 23468
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.987378844064994e-06,
+      "loss": 0.6029,
+      "step": 23469
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.98666967887204e-06,
+      "loss": 0.6395,
+      "step": 23470
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.985960518828115e-06,
+      "loss": 0.5516,
+      "step": 23471
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.985251363936823e-06,
+      "loss": 0.6265,
+      "step": 23472
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.984542214201764e-06,
+      "loss": 0.5971,
+      "step": 23473
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.983833069626543e-06,
+      "loss": 0.5543,
+      "step": 23474
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.983123930214762e-06,
+      "loss": 0.6504,
+      "step": 23475
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.982414795970027e-06,
+      "loss": 0.6097,
+      "step": 23476
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.98170566689594e-06,
+      "loss": 0.6237,
+      "step": 23477
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.98099654299611e-06,
+      "loss": 0.6347,
+      "step": 23478
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.980287424274132e-06,
+      "loss": 0.5998,
+      "step": 23479
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.97957831073361e-06,
+      "loss": 0.6133,
+      "step": 23480
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.978869202378152e-06,
+      "loss": 0.6622,
+      "step": 23481
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.978160099211357e-06,
+      "loss": 0.6055,
+      "step": 23482
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.977451001236832e-06,
+      "loss": 0.6187,
+      "step": 23483
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.976741908458176e-06,
+      "loss": 0.5981,
+      "step": 23484
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.976032820878993e-06,
+      "loss": 0.6162,
+      "step": 23485
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.975323738502888e-06,
+      "loss": 0.6221,
+      "step": 23486
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.974614661333464e-06,
+      "loss": 0.6869,
+      "step": 23487
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.973905589374323e-06,
+      "loss": 0.5711,
+      "step": 23488
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.97319652262907e-06,
+      "loss": 0.5922,
+      "step": 23489
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.972487461101306e-06,
+      "loss": 0.6338,
+      "step": 23490
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.97177840479463e-06,
+      "loss": 0.5972,
+      "step": 23491
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.971069353712653e-06,
+      "loss": 0.6041,
+      "step": 23492
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.97036030785897e-06,
+      "loss": 0.6268,
+      "step": 23493
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.969651267237195e-06,
+      "loss": 0.6302,
+      "step": 23494
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.968942231850918e-06,
+      "loss": 0.6088,
+      "step": 23495
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.968233201703748e-06,
+      "loss": 0.591,
+      "step": 23496
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.967524176799285e-06,
+      "loss": 0.5496,
+      "step": 23497
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.966815157141137e-06,
+      "loss": 0.6952,
+      "step": 23498
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.966106142732902e-06,
+      "loss": 0.6092,
+      "step": 23499
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.965397133578191e-06,
+      "loss": 0.6141,
+      "step": 23500
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.964688129680593e-06,
+      "loss": 0.6244,
+      "step": 23501
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.96397913104372e-06,
+      "loss": 0.5593,
+      "step": 23502
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.963270137671171e-06,
+      "loss": 0.6198,
+      "step": 23503
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.962561149566553e-06,
+      "loss": 0.5786,
+      "step": 23504
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.961852166733465e-06,
+      "loss": 0.6448,
+      "step": 23505
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.961143189175509e-06,
+      "loss": 0.6015,
+      "step": 23506
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.96043421689629e-06,
+      "loss": 0.5665,
+      "step": 23507
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.959725249899409e-06,
+      "loss": 0.5803,
+      "step": 23508
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.959016288188466e-06,
+      "loss": 0.6346,
+      "step": 23509
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.958307331767069e-06,
+      "loss": 0.6508,
+      "step": 23510
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.957598380638817e-06,
+      "loss": 0.6344,
+      "step": 23511
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.956889434807312e-06,
+      "loss": 0.6274,
+      "step": 23512
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.956180494276157e-06,
+      "loss": 0.559,
+      "step": 23513
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.955471559048953e-06,
+      "loss": 0.6321,
+      "step": 23514
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.954762629129307e-06,
+      "loss": 0.6125,
+      "step": 23515
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.95405370452082e-06,
+      "loss": 0.5652,
+      "step": 23516
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.953344785227089e-06,
+      "loss": 0.6241,
+      "step": 23517
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.95263587125172e-06,
+      "loss": 0.5868,
+      "step": 23518
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.951926962598314e-06,
+      "loss": 0.5983,
+      "step": 23519
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.951218059270476e-06,
+      "loss": 0.584,
+      "step": 23520
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.950509161271808e-06,
+      "loss": 0.6041,
+      "step": 23521
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.949800268605906e-06,
+      "loss": 0.6586,
+      "step": 23522
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.94909138127638e-06,
+      "loss": 0.5849,
+      "step": 23523
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.948382499286825e-06,
+      "loss": 0.5403,
+      "step": 23524
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.947673622640848e-06,
+      "loss": 0.5859,
+      "step": 23525
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.94696475134205e-06,
+      "loss": 0.5128,
+      "step": 23526
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.946255885394034e-06,
+      "loss": 0.6093,
+      "step": 23527
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.9455470248004e-06,
+      "loss": 0.5925,
+      "step": 23528
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.944838169564748e-06,
+      "loss": 0.6252,
+      "step": 23529
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.944129319690682e-06,
+      "loss": 0.6224,
+      "step": 23530
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.943420475181805e-06,
+      "loss": 0.6462,
+      "step": 23531
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.942711636041723e-06,
+      "loss": 0.5732,
+      "step": 23532
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.942002802274026e-06,
+      "loss": 0.62,
+      "step": 23533
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.941293973882324e-06,
+      "loss": 0.5767,
+      "step": 23534
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.940585150870218e-06,
+      "loss": 0.6634,
+      "step": 23535
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.939876333241308e-06,
+      "loss": 0.5873,
+      "step": 23536
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.939167520999202e-06,
+      "loss": 0.6324,
+      "step": 23537
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.938458714147494e-06,
+      "loss": 0.5698,
+      "step": 23538
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.937749912689784e-06,
+      "loss": 0.6184,
+      "step": 23539
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.93704111662968e-06,
+      "loss": 0.5912,
+      "step": 23540
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.936332325970782e-06,
+      "loss": 0.6188,
+      "step": 23541
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.93562354071669e-06,
+      "loss": 0.626,
+      "step": 23542
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.93491476087101e-06,
+      "loss": 0.6113,
+      "step": 23543
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.934205986437337e-06,
+      "loss": 0.5715,
+      "step": 23544
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.933497217419275e-06,
+      "loss": 0.5717,
+      "step": 23545
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.932788453820428e-06,
+      "loss": 0.5468,
+      "step": 23546
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.932079695644395e-06,
+      "loss": 0.5781,
+      "step": 23547
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.93137094289478e-06,
+      "loss": 0.63,
+      "step": 23548
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.93066219557518e-06,
+      "loss": 0.5846,
+      "step": 23549
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.929953453689197e-06,
+      "loss": 0.6419,
+      "step": 23550
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.929244717240433e-06,
+      "loss": 0.6383,
+      "step": 23551
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.928535986232492e-06,
+      "loss": 0.6333,
+      "step": 23552
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.927827260668974e-06,
+      "loss": 0.6456,
+      "step": 23553
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.927118540553481e-06,
+      "loss": 0.614,
+      "step": 23554
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.926409825889612e-06,
+      "loss": 0.6216,
+      "step": 23555
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.925701116680966e-06,
+      "loss": 0.5821,
+      "step": 23556
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.92499241293115e-06,
+      "loss": 0.6113,
+      "step": 23557
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.92428371464376e-06,
+      "loss": 0.5593,
+      "step": 23558
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.923575021822405e-06,
+      "loss": 0.631,
+      "step": 23559
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.922866334470677e-06,
+      "loss": 0.567,
+      "step": 23560
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.922157652592179e-06,
+      "loss": 0.5762,
+      "step": 23561
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.921448976190515e-06,
+      "loss": 0.6935,
+      "step": 23562
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.920740305269285e-06,
+      "loss": 0.6233,
+      "step": 23563
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.92003163983209e-06,
+      "loss": 0.612,
+      "step": 23564
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.91932297988253e-06,
+      "loss": 0.6434,
+      "step": 23565
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.918614325424207e-06,
+      "loss": 0.6188,
+      "step": 23566
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.91790567646072e-06,
+      "loss": 0.5997,
+      "step": 23567
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.917197032995672e-06,
+      "loss": 0.6105,
+      "step": 23568
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.916488395032662e-06,
+      "loss": 0.5844,
+      "step": 23569
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.915779762575297e-06,
+      "loss": 0.6189,
+      "step": 23570
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.915071135627166e-06,
+      "loss": 0.5569,
+      "step": 23571
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.914362514191877e-06,
+      "loss": 0.6478,
+      "step": 23572
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.913653898273032e-06,
+      "loss": 0.6468,
+      "step": 23573
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.912945287874227e-06,
+      "loss": 0.5377,
+      "step": 23574
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.912236682999072e-06,
+      "loss": 0.6782,
+      "step": 23575
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.911528083651155e-06,
+      "loss": 0.6323,
+      "step": 23576
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.910819489834085e-06,
+      "loss": 0.5975,
+      "step": 23577
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.910110901551456e-06,
+      "loss": 0.5889,
+      "step": 23578
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.909402318806876e-06,
+      "loss": 0.6221,
+      "step": 23579
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.908693741603941e-06,
+      "loss": 0.5907,
+      "step": 23580
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.907985169946257e-06,
+      "loss": 0.5903,
+      "step": 23581
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.907276603837416e-06,
+      "loss": 0.568,
+      "step": 23582
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.906568043281023e-06,
+      "loss": 0.5837,
+      "step": 23583
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.90585948828068e-06,
+      "loss": 0.5841,
+      "step": 23584
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.905150938839984e-06,
+      "loss": 0.5964,
+      "step": 23585
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.904442394962536e-06,
+      "loss": 0.6153,
+      "step": 23586
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.903733856651938e-06,
+      "loss": 0.6116,
+      "step": 23587
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.90302532391179e-06,
+      "loss": 0.6587,
+      "step": 23588
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.90231679674569e-06,
+      "loss": 0.6039,
+      "step": 23589
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.90160827515724e-06,
+      "loss": 0.5744,
+      "step": 23590
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.900899759150039e-06,
+      "loss": 0.6383,
+      "step": 23591
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.900191248727695e-06,
+      "loss": 0.5828,
+      "step": 23592
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.899482743893793e-06,
+      "loss": 0.6082,
+      "step": 23593
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.898774244651944e-06,
+      "loss": 0.5904,
+      "step": 23594
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.898065751005745e-06,
+      "loss": 0.6486,
+      "step": 23595
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.897357262958798e-06,
+      "loss": 0.6246,
+      "step": 23596
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.896648780514705e-06,
+      "loss": 0.6349,
+      "step": 23597
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.895940303677058e-06,
+      "loss": 0.5646,
+      "step": 23598
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.895231832449461e-06,
+      "loss": 0.6187,
+      "step": 23599
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.894523366835516e-06,
+      "loss": 0.5109,
+      "step": 23600
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.893814906838821e-06,
+      "loss": 0.6126,
+      "step": 23601
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.89310645246298e-06,
+      "loss": 0.6055,
+      "step": 23602
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.892398003711586e-06,
+      "loss": 0.5755,
+      "step": 23603
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.891689560588243e-06,
+      "loss": 0.59,
+      "step": 23604
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.89098112309655e-06,
+      "loss": 0.6261,
+      "step": 23605
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.890272691240105e-06,
+      "loss": 0.6361,
+      "step": 23606
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.889564265022511e-06,
+      "loss": 0.6839,
+      "step": 23607
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.88885584444737e-06,
+      "loss": 0.6367,
+      "step": 23608
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.888147429518272e-06,
+      "loss": 0.5526,
+      "step": 23609
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.887439020238825e-06,
+      "loss": 0.5434,
+      "step": 23610
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.886730616612624e-06,
+      "loss": 0.6919,
+      "step": 23611
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.88602221864327e-06,
+      "loss": 0.6896,
+      "step": 23612
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.88531382633437e-06,
+      "loss": 0.6255,
+      "step": 23613
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.884605439689512e-06,
+      "loss": 0.6201,
+      "step": 23614
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.8838970587123e-06,
+      "loss": 0.5718,
+      "step": 23615
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.883188683406335e-06,
+      "loss": 0.5452,
+      "step": 23616
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.882480313775214e-06,
+      "loss": 0.5716,
+      "step": 23617
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.881771949822538e-06,
+      "loss": 0.6225,
+      "step": 23618
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.88106359155191e-06,
+      "loss": 0.6439,
+      "step": 23619
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.880355238966923e-06,
+      "loss": 0.6229,
+      "step": 23620
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.879646892071178e-06,
+      "loss": 0.5883,
+      "step": 23621
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.878938550868275e-06,
+      "loss": 0.6076,
+      "step": 23622
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.878230215361816e-06,
+      "loss": 0.5413,
+      "step": 23623
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.877521885555398e-06,
+      "loss": 0.6509,
+      "step": 23624
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.876813561452618e-06,
+      "loss": 0.5468,
+      "step": 23625
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.876105243057078e-06,
+      "loss": 0.5722,
+      "step": 23626
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.875396930372375e-06,
+      "loss": 0.5521,
+      "step": 23627
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.874688623402111e-06,
+      "loss": 0.6287,
+      "step": 23628
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.873980322149886e-06,
+      "loss": 0.6531,
+      "step": 23629
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.873272026619293e-06,
+      "loss": 0.5531,
+      "step": 23630
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.872563736813934e-06,
+      "loss": 0.5938,
+      "step": 23631
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.871855452737412e-06,
+      "loss": 0.5854,
+      "step": 23632
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.87114717439332e-06,
+      "loss": 0.5591,
+      "step": 23633
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.87043890178526e-06,
+      "loss": 0.604,
+      "step": 23634
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.869730634916835e-06,
+      "loss": 0.5614,
+      "step": 23635
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.869022373791638e-06,
+      "loss": 0.5627,
+      "step": 23636
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.868314118413266e-06,
+      "loss": 0.5962,
+      "step": 23637
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.867605868785322e-06,
+      "loss": 0.6108,
+      "step": 23638
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.866897624911406e-06,
+      "loss": 0.6114,
+      "step": 23639
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.866189386795118e-06,
+      "loss": 0.6097,
+      "step": 23640
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.86548115444005e-06,
+      "loss": 0.6101,
+      "step": 23641
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.864772927849804e-06,
+      "loss": 0.5884,
+      "step": 23642
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.864064707027983e-06,
+      "loss": 0.5593,
+      "step": 23643
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.863356491978178e-06,
+      "loss": 0.5195,
+      "step": 23644
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.862648282703992e-06,
+      "loss": 0.6471,
+      "step": 23645
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.861940079209024e-06,
+      "loss": 0.6272,
+      "step": 23646
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.861231881496875e-06,
+      "loss": 0.5903,
+      "step": 23647
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.860523689571136e-06,
+      "loss": 0.59,
+      "step": 23648
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.85981550343541e-06,
+      "loss": 0.616,
+      "step": 23649
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.859107323093295e-06,
+      "loss": 0.5639,
+      "step": 23650
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.858399148548395e-06,
+      "loss": 0.538,
+      "step": 23651
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.857690979804299e-06,
+      "loss": 0.6211,
+      "step": 23652
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.85698281686461e-06,
+      "loss": 0.6871,
+      "step": 23653
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.856274659732925e-06,
+      "loss": 0.6148,
+      "step": 23654
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.855566508412843e-06,
+      "loss": 0.6063,
+      "step": 23655
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.854858362907967e-06,
+      "loss": 0.6399,
+      "step": 23656
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.854150223221888e-06,
+      "loss": 0.533,
+      "step": 23657
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.853442089358205e-06,
+      "loss": 0.5873,
+      "step": 23658
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.85273396132052e-06,
+      "loss": 0.5937,
+      "step": 23659
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.852025839112431e-06,
+      "loss": 0.5637,
+      "step": 23660
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.851317722737536e-06,
+      "loss": 0.7167,
+      "step": 23661
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.850609612199432e-06,
+      "loss": 0.6214,
+      "step": 23662
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.849901507501716e-06,
+      "loss": 0.6778,
+      "step": 23663
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.849193408647988e-06,
+      "loss": 0.5449,
+      "step": 23664
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.848485315641845e-06,
+      "loss": 0.6521,
+      "step": 23665
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.847777228486882e-06,
+      "loss": 0.588,
+      "step": 23666
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.847069147186705e-06,
+      "loss": 0.6004,
+      "step": 23667
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.846361071744908e-06,
+      "loss": 0.6172,
+      "step": 23668
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.845653002165084e-06,
+      "loss": 0.6137,
+      "step": 23669
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.844944938450837e-06,
+      "loss": 0.6165,
+      "step": 23670
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.844236880605761e-06,
+      "loss": 0.563,
+      "step": 23671
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.843528828633458e-06,
+      "loss": 0.5539,
+      "step": 23672
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.842820782537527e-06,
+      "loss": 0.6273,
+      "step": 23673
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.842112742321558e-06,
+      "loss": 0.5659,
+      "step": 23674
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.841404707989154e-06,
+      "loss": 0.5611,
+      "step": 23675
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.84069667954391e-06,
+      "loss": 0.5607,
+      "step": 23676
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.839988656989427e-06,
+      "loss": 0.5994,
+      "step": 23677
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.839280640329307e-06,
+      "loss": 0.5937,
+      "step": 23678
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.838572629567137e-06,
+      "loss": 0.6146,
+      "step": 23679
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.83786462470652e-06,
+      "loss": 0.6479,
+      "step": 23680
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.837156625751052e-06,
+      "loss": 0.6109,
+      "step": 23681
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.836448632704333e-06,
+      "loss": 0.6647,
+      "step": 23682
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.835740645569959e-06,
+      "loss": 0.5724,
+      "step": 23683
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.835032664351529e-06,
+      "loss": 0.5855,
+      "step": 23684
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.834324689052641e-06,
+      "loss": 0.6045,
+      "step": 23685
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.833616719676885e-06,
+      "loss": 0.6519,
+      "step": 23686
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.832908756227868e-06,
+      "loss": 0.603,
+      "step": 23687
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.832200798709179e-06,
+      "loss": 0.5484,
+      "step": 23688
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.831492847124428e-06,
+      "loss": 0.5816,
+      "step": 23689
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.830784901477198e-06,
+      "loss": 0.5999,
+      "step": 23690
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.830076961771093e-06,
+      "loss": 0.5942,
+      "step": 23691
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.829369028009708e-06,
+      "loss": 0.5528,
+      "step": 23692
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.828661100196643e-06,
+      "loss": 0.5602,
+      "step": 23693
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.827953178335498e-06,
+      "loss": 0.5807,
+      "step": 23694
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.827245262429862e-06,
+      "loss": 0.5808,
+      "step": 23695
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.826537352483337e-06,
+      "loss": 0.5958,
+      "step": 23696
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.82582944849952e-06,
+      "loss": 0.5596,
+      "step": 23697
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.825121550482005e-06,
+      "loss": 0.6737,
+      "step": 23698
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.824413658434394e-06,
+      "loss": 0.6145,
+      "step": 23699
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.823705772360283e-06,
+      "loss": 0.6461,
+      "step": 23700
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.822997892263263e-06,
+      "loss": 0.6395,
+      "step": 23701
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.822290018146938e-06,
+      "loss": 0.6418,
+      "step": 23702
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.821582150014905e-06,
+      "loss": 0.5864,
+      "step": 23703
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.820874287870753e-06,
+      "loss": 0.5898,
+      "step": 23704
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.82016643171809e-06,
+      "loss": 0.5881,
+      "step": 23705
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.819458581560503e-06,
+      "loss": 0.5686,
+      "step": 23706
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.818750737401592e-06,
+      "loss": 0.6785,
+      "step": 23707
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.818042899244955e-06,
+      "loss": 0.6577,
+      "step": 23708
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.817335067094188e-06,
+      "loss": 0.6424,
+      "step": 23709
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.816627240952888e-06,
+      "loss": 0.5589,
+      "step": 23710
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.815919420824654e-06,
+      "loss": 0.589,
+      "step": 23711
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.81521160671308e-06,
+      "loss": 0.6125,
+      "step": 23712
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.814503798621758e-06,
+      "loss": 0.5985,
+      "step": 23713
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.813795996554293e-06,
+      "loss": 0.6137,
+      "step": 23714
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.813088200514275e-06,
+      "loss": 0.5784,
+      "step": 23715
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.81238041050531e-06,
+      "loss": 0.5453,
+      "step": 23716
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.811672626530983e-06,
+      "loss": 0.5613,
+      "step": 23717
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.810964848594897e-06,
+      "loss": 0.5721,
+      "step": 23718
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.810257076700644e-06,
+      "loss": 0.5501,
+      "step": 23719
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.809549310851827e-06,
+      "loss": 0.615,
+      "step": 23720
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.808841551052037e-06,
+      "loss": 0.6148,
+      "step": 23721
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.808133797304873e-06,
+      "loss": 0.6077,
+      "step": 23722
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.807426049613927e-06,
+      "loss": 0.6167,
+      "step": 23723
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.806718307982802e-06,
+      "loss": 0.6251,
+      "step": 23724
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.806010572415088e-06,
+      "loss": 0.6007,
+      "step": 23725
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.805302842914386e-06,
+      "loss": 0.637,
+      "step": 23726
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.804595119484292e-06,
+      "loss": 0.5343,
+      "step": 23727
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.803887402128397e-06,
+      "loss": 0.5942,
+      "step": 23728
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.8031796908503e-06,
+      "loss": 0.5808,
+      "step": 23729
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.802471985653597e-06,
+      "loss": 0.5648,
+      "step": 23730
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.801764286541886e-06,
+      "loss": 0.5799,
+      "step": 23731
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.801056593518765e-06,
+      "loss": 0.5788,
+      "step": 23732
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.800348906587823e-06,
+      "loss": 0.6035,
+      "step": 23733
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.79964122575266e-06,
+      "loss": 0.5497,
+      "step": 23734
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.798933551016869e-06,
+      "loss": 0.5851,
+      "step": 23735
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.79822588238405e-06,
+      "loss": 0.5893,
+      "step": 23736
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.797518219857797e-06,
+      "loss": 0.6081,
+      "step": 23737
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.79681056344171e-06,
+      "loss": 0.6689,
+      "step": 23738
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.796102913139377e-06,
+      "loss": 0.5886,
+      "step": 23739
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.7953952689544e-06,
+      "loss": 0.5914,
+      "step": 23740
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.79468763089037e-06,
+      "loss": 0.5834,
+      "step": 23741
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.793979998950885e-06,
+      "loss": 0.602,
+      "step": 23742
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.793272373139544e-06,
+      "loss": 0.5994,
+      "step": 23743
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.79256475345994e-06,
+      "loss": 0.6293,
+      "step": 23744
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.791857139915664e-06,
+      "loss": 0.567,
+      "step": 23745
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.791149532510315e-06,
+      "loss": 0.6231,
+      "step": 23746
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.790441931247492e-06,
+      "loss": 0.5947,
+      "step": 23747
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.789734336130791e-06,
+      "loss": 0.5252,
+      "step": 23748
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.7890267471638e-06,
+      "loss": 0.5615,
+      "step": 23749
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.788319164350118e-06,
+      "loss": 0.597,
+      "step": 23750
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.787611587693342e-06,
+      "loss": 0.583,
+      "step": 23751
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.786904017197066e-06,
+      "loss": 0.5392,
+      "step": 23752
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.786196452864886e-06,
+      "loss": 0.6011,
+      "step": 23753
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.785488894700402e-06,
+      "loss": 0.5923,
+      "step": 23754
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.7847813427072e-06,
+      "loss": 0.651,
+      "step": 23755
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.78407379688888e-06,
+      "loss": 0.5893,
+      "step": 23756
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.783366257249036e-06,
+      "loss": 0.5692,
+      "step": 23757
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.782658723791266e-06,
+      "loss": 0.646,
+      "step": 23758
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.781951196519168e-06,
+      "loss": 0.6193,
+      "step": 23759
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.781243675436328e-06,
+      "loss": 0.5358,
+      "step": 23760
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.780536160546344e-06,
+      "loss": 0.6552,
+      "step": 23761
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.779828651852817e-06,
+      "loss": 0.5324,
+      "step": 23762
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.779121149359337e-06,
+      "loss": 0.5637,
+      "step": 23763
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.778413653069498e-06,
+      "loss": 0.6547,
+      "step": 23764
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.777706162986902e-06,
+      "loss": 0.6352,
+      "step": 23765
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.776998679115134e-06,
+      "loss": 0.5809,
+      "step": 23766
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.776291201457794e-06,
+      "loss": 0.5885,
+      "step": 23767
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.775583730018476e-06,
+      "loss": 0.5791,
+      "step": 23768
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.774876264800777e-06,
+      "loss": 0.6273,
+      "step": 23769
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.774168805808295e-06,
+      "loss": 0.6093,
+      "step": 23770
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.773461353044615e-06,
+      "loss": 0.6391,
+      "step": 23771
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.772753906513339e-06,
+      "loss": 0.6181,
+      "step": 23772
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.772046466218057e-06,
+      "loss": 0.5732,
+      "step": 23773
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.771339032162368e-06,
+      "loss": 0.675,
+      "step": 23774
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.770631604349865e-06,
+      "loss": 0.6046,
+      "step": 23775
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.769924182784148e-06,
+      "loss": 0.6716,
+      "step": 23776
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.7692167674688e-06,
+      "loss": 0.544,
+      "step": 23777
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.768509358407425e-06,
+      "loss": 0.6178,
+      "step": 23778
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.767801955603611e-06,
+      "loss": 0.6032,
+      "step": 23779
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.767094559060962e-06,
+      "loss": 0.6146,
+      "step": 23780
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.766387168783066e-06,
+      "loss": 0.5856,
+      "step": 23781
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.765679784773512e-06,
+      "loss": 0.5721,
+      "step": 23782
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.764972407035908e-06,
+      "loss": 0.5967,
+      "step": 23783
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.764265035573836e-06,
+      "loss": 0.6229,
+      "step": 23784
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.763557670390895e-06,
+      "loss": 0.5416,
+      "step": 23785
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.762850311490685e-06,
+      "loss": 0.6415,
+      "step": 23786
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.762142958876789e-06,
+      "loss": 0.6016,
+      "step": 23787
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.761435612552809e-06,
+      "loss": 0.613,
+      "step": 23788
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.760728272522336e-06,
+      "loss": 0.6088,
+      "step": 23789
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.760020938788967e-06,
+      "loss": 0.5324,
+      "step": 23790
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.759313611356294e-06,
+      "loss": 0.5805,
+      "step": 23791
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.758606290227915e-06,
+      "loss": 0.5214,
+      "step": 23792
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.757898975407418e-06,
+      "loss": 0.588,
+      "step": 23793
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.757191666898401e-06,
+      "loss": 0.6169,
+      "step": 23794
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.756484364704455e-06,
+      "loss": 0.6878,
+      "step": 23795
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.755777068829176e-06,
+      "loss": 0.5848,
+      "step": 23796
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.755069779276165e-06,
+      "loss": 0.6213,
+      "step": 23797
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.754362496049002e-06,
+      "loss": 0.6302,
+      "step": 23798
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.75365521915129e-06,
+      "loss": 0.5442,
+      "step": 23799
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.752947948586622e-06,
+      "loss": 0.5779,
+      "step": 23800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.75224068435859e-06,
+      "loss": 0.5421,
+      "step": 23801
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.751533426470787e-06,
+      "loss": 0.5652,
+      "step": 23802
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.75082617492681e-06,
+      "loss": 0.5735,
+      "step": 23803
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.750118929730251e-06,
+      "loss": 0.6253,
+      "step": 23804
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.749411690884703e-06,
+      "loss": 0.5903,
+      "step": 23805
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.748704458393759e-06,
+      "loss": 0.6326,
+      "step": 23806
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.747997232261017e-06,
+      "loss": 0.5792,
+      "step": 23807
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.747290012490068e-06,
+      "loss": 0.5927,
+      "step": 23808
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.746582799084504e-06,
+      "loss": 0.6555,
+      "step": 23809
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.74587559204792e-06,
+      "loss": 0.5529,
+      "step": 23810
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.745168391383907e-06,
+      "loss": 0.5406,
+      "step": 23811
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.744461197096062e-06,
+      "loss": 0.5814,
+      "step": 23812
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.743754009187982e-06,
+      "loss": 0.6296,
+      "step": 23813
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.74304682766325e-06,
+      "loss": 0.5559,
+      "step": 23814
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.742339652525467e-06,
+      "loss": 0.6099,
+      "step": 23815
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.741632483778225e-06,
+      "loss": 0.5771,
+      "step": 23816
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.740925321425115e-06,
+      "loss": 0.6053,
+      "step": 23817
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.740218165469735e-06,
+      "loss": 0.6227,
+      "step": 23818
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.739511015915676e-06,
+      "loss": 0.5626,
+      "step": 23819
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.738803872766528e-06,
+      "loss": 0.6113,
+      "step": 23820
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.738096736025889e-06,
+      "loss": 0.6159,
+      "step": 23821
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.737389605697349e-06,
+      "loss": 0.6443,
+      "step": 23822
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.736682481784502e-06,
+      "loss": 0.5631,
+      "step": 23823
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.735975364290941e-06,
+      "loss": 0.6141,
+      "step": 23824
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.73526825322026e-06,
+      "loss": 0.5987,
+      "step": 23825
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.734561148576051e-06,
+      "loss": 0.6358,
+      "step": 23826
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.733854050361908e-06,
+      "loss": 0.6256,
+      "step": 23827
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.733146958581422e-06,
+      "loss": 0.6445,
+      "step": 23828
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.732439873238188e-06,
+      "loss": 0.5828,
+      "step": 23829
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.731732794335801e-06,
+      "loss": 0.5545,
+      "step": 23830
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.731025721877848e-06,
+      "loss": 0.5319,
+      "step": 23831
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.730318655867924e-06,
+      "loss": 0.5706,
+      "step": 23832
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.729611596309625e-06,
+      "loss": 0.5743,
+      "step": 23833
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.72890454320654e-06,
+      "loss": 0.5507,
+      "step": 23834
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.728197496562268e-06,
+      "loss": 0.6063,
+      "step": 23835
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.727490456380391e-06,
+      "loss": 0.5942,
+      "step": 23836
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.726783422664509e-06,
+      "loss": 0.6066,
+      "step": 23837
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.726076395418214e-06,
+      "loss": 0.5962,
+      "step": 23838
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.725369374645098e-06,
+      "loss": 0.6155,
+      "step": 23839
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.724662360348755e-06,
+      "loss": 0.6288,
+      "step": 23840
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.723955352532773e-06,
+      "loss": 0.5514,
+      "step": 23841
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.723248351200752e-06,
+      "loss": 0.6431,
+      "step": 23842
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.722541356356275e-06,
+      "loss": 0.6185,
+      "step": 23843
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.721834368002942e-06,
+      "loss": 0.5979,
+      "step": 23844
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.72112738614434e-06,
+      "loss": 0.6531,
+      "step": 23845
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.720420410784071e-06,
+      "loss": 0.5679,
+      "step": 23846
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.719713441925716e-06,
+      "loss": 0.6071,
+      "step": 23847
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.719006479572872e-06,
+      "loss": 0.5856,
+      "step": 23848
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.71829952372913e-06,
+      "loss": 0.5918,
+      "step": 23849
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.717592574398086e-06,
+      "loss": 0.5729,
+      "step": 23850
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.71688563158333e-06,
+      "loss": 0.6703,
+      "step": 23851
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.716178695288453e-06,
+      "loss": 0.5655,
+      "step": 23852
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.715471765517046e-06,
+      "loss": 0.5804,
+      "step": 23853
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.714764842272705e-06,
+      "loss": 0.5396,
+      "step": 23854
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.71405792555902e-06,
+      "loss": 0.6799,
+      "step": 23855
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.713351015379584e-06,
+      "loss": 0.596,
+      "step": 23856
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.712644111737988e-06,
+      "loss": 0.5943,
+      "step": 23857
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.711937214637825e-06,
+      "loss": 0.6848,
+      "step": 23858
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.711230324082684e-06,
+      "loss": 0.5978,
+      "step": 23859
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.710523440076161e-06,
+      "loss": 0.5679,
+      "step": 23860
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.709816562621845e-06,
+      "loss": 0.6094,
+      "step": 23861
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.709109691723331e-06,
+      "loss": 0.5963,
+      "step": 23862
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.708402827384209e-06,
+      "loss": 0.5779,
+      "step": 23863
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.707695969608067e-06,
+      "loss": 0.5833,
+      "step": 23864
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.7069891183985e-06,
+      "loss": 0.6434,
+      "step": 23865
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.706282273759101e-06,
+      "loss": 0.5585,
+      "step": 23866
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.70557543569346e-06,
+      "loss": 0.6201,
+      "step": 23867
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.704868604205175e-06,
+      "loss": 0.5858,
+      "step": 23868
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.704161779297827e-06,
+      "loss": 0.6032,
+      "step": 23869
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.703454960975013e-06,
+      "loss": 0.559,
+      "step": 23870
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.702748149240322e-06,
+      "loss": 0.5638,
+      "step": 23871
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.70204134409735e-06,
+      "loss": 0.6324,
+      "step": 23872
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.701334545549688e-06,
+      "loss": 0.6158,
+      "step": 23873
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.700627753600924e-06,
+      "loss": 0.6027,
+      "step": 23874
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.699920968254649e-06,
+      "loss": 0.6556,
+      "step": 23875
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.699214189514457e-06,
+      "loss": 0.5929,
+      "step": 23876
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.69850741738394e-06,
+      "loss": 0.5683,
+      "step": 23877
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.69780065186669e-06,
+      "loss": 0.6342,
+      "step": 23878
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.697093892966291e-06,
+      "loss": 0.5996,
+      "step": 23879
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.696387140686345e-06,
+      "loss": 0.5799,
+      "step": 23880
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.695680395030433e-06,
+      "loss": 0.5959,
+      "step": 23881
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.69497365600215e-06,
+      "loss": 0.5457,
+      "step": 23882
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.694266923605091e-06,
+      "loss": 0.5968,
+      "step": 23883
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.693560197842847e-06,
+      "loss": 0.6416,
+      "step": 23884
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.692853478719002e-06,
+      "loss": 0.5985,
+      "step": 23885
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.692146766237151e-06,
+      "loss": 0.6331,
+      "step": 23886
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.691440060400887e-06,
+      "loss": 0.6433,
+      "step": 23887
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.690733361213799e-06,
+      "loss": 0.6192,
+      "step": 23888
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.69002666867948e-06,
+      "loss": 0.5936,
+      "step": 23889
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.689319982801516e-06,
+      "loss": 0.6251,
+      "step": 23890
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.688613303583502e-06,
+      "loss": 0.6061,
+      "step": 23891
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.687906631029027e-06,
+      "loss": 0.6565,
+      "step": 23892
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.687199965141684e-06,
+      "loss": 0.5645,
+      "step": 23893
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.686493305925061e-06,
+      "loss": 0.5986,
+      "step": 23894
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.685786653382754e-06,
+      "loss": 0.5219,
+      "step": 23895
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.685080007518347e-06,
+      "loss": 0.619,
+      "step": 23896
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.684373368335432e-06,
+      "loss": 0.5807,
+      "step": 23897
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.683666735837606e-06,
+      "loss": 0.6015,
+      "step": 23898
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.682960110028452e-06,
+      "loss": 0.614,
+      "step": 23899
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.682253490911563e-06,
+      "loss": 0.6116,
+      "step": 23900
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.681546878490532e-06,
+      "loss": 0.5251,
+      "step": 23901
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.680840272768944e-06,
+      "loss": 0.5789,
+      "step": 23902
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.680133673750395e-06,
+      "loss": 0.5956,
+      "step": 23903
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.679427081438471e-06,
+      "loss": 0.5909,
+      "step": 23904
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.67872049583677e-06,
+      "loss": 0.5759,
+      "step": 23905
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.678013916948874e-06,
+      "loss": 0.5332,
+      "step": 23906
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.677307344778374e-06,
+      "loss": 0.5976,
+      "step": 23907
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.676600779328866e-06,
+      "loss": 0.5432,
+      "step": 23908
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.675894220603935e-06,
+      "loss": 0.6122,
+      "step": 23909
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.675187668607174e-06,
+      "loss": 0.6103,
+      "step": 23910
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.674481123342176e-06,
+      "loss": 0.5328,
+      "step": 23911
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.673774584812523e-06,
+      "loss": 0.6296,
+      "step": 23912
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.67306805302181e-06,
+      "loss": 0.6066,
+      "step": 23913
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.672361527973627e-06,
+      "loss": 0.6365,
+      "step": 23914
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.671655009671565e-06,
+      "loss": 0.5607,
+      "step": 23915
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.670948498119215e-06,
+      "loss": 0.5939,
+      "step": 23916
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.670241993320161e-06,
+      "loss": 0.5777,
+      "step": 23917
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.669535495277999e-06,
+      "loss": 0.5412,
+      "step": 23918
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.668829003996317e-06,
+      "loss": 0.6361,
+      "step": 23919
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.668122519478702e-06,
+      "loss": 0.5431,
+      "step": 23920
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.66741604172875e-06,
+      "loss": 0.5906,
+      "step": 23921
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.666709570750049e-06,
+      "loss": 0.566,
+      "step": 23922
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.666003106546184e-06,
+      "loss": 0.6102,
+      "step": 23923
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.665296649120745e-06,
+      "loss": 0.6365,
+      "step": 23924
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.664590198477327e-06,
+      "loss": 0.6548,
+      "step": 23925
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.663883754619517e-06,
+      "loss": 0.579,
+      "step": 23926
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.663177317550911e-06,
+      "loss": 0.5684,
+      "step": 23927
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.662470887275085e-06,
+      "loss": 0.5802,
+      "step": 23928
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.661764463795639e-06,
+      "loss": 0.6038,
+      "step": 23929
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.661058047116158e-06,
+      "loss": 0.5686,
+      "step": 23930
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.660351637240234e-06,
+      "loss": 0.6608,
+      "step": 23931
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.65964523417146e-06,
+      "loss": 0.6175,
+      "step": 23932
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.658938837913414e-06,
+      "loss": 0.5947,
+      "step": 23933
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.658232448469695e-06,
+      "loss": 0.6465,
+      "step": 23934
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.65752606584389e-06,
+      "loss": 0.6139,
+      "step": 23935
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.65681969003959e-06,
+      "loss": 0.594,
+      "step": 23936
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.65611332106038e-06,
+      "loss": 0.5261,
+      "step": 23937
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.655406958909855e-06,
+      "loss": 0.6213,
+      "step": 23938
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.654700603591598e-06,
+      "loss": 0.6366,
+      "step": 23939
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.653994255109204e-06,
+      "loss": 0.5836,
+      "step": 23940
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.653287913466256e-06,
+      "loss": 0.6199,
+      "step": 23941
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.652581578666349e-06,
+      "loss": 0.5916,
+      "step": 23942
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.651875250713071e-06,
+      "loss": 0.6033,
+      "step": 23943
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.651168929610007e-06,
+      "loss": 0.616,
+      "step": 23944
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.650462615360748e-06,
+      "loss": 0.5746,
+      "step": 23945
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.649756307968883e-06,
+      "loss": 0.5828,
+      "step": 23946
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.649050007438003e-06,
+      "loss": 0.5776,
+      "step": 23947
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.648343713771696e-06,
+      "loss": 0.5992,
+      "step": 23948
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.647637426973553e-06,
+      "loss": 0.6045,
+      "step": 23949
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.646931147047157e-06,
+      "loss": 0.5642,
+      "step": 23950
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.6462248739961e-06,
+      "loss": 0.5911,
+      "step": 23951
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.645518607823972e-06,
+      "loss": 0.6465,
+      "step": 23952
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.64481234853436e-06,
+      "loss": 0.6108,
+      "step": 23953
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.644106096130857e-06,
+      "loss": 0.6039,
+      "step": 23954
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.643399850617043e-06,
+      "loss": 0.6124,
+      "step": 23955
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.642693611996514e-06,
+      "loss": 0.5764,
+      "step": 23956
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.641987380272856e-06,
+      "loss": 0.6638,
+      "step": 23957
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.641281155449656e-06,
+      "loss": 0.5701,
+      "step": 23958
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.640574937530509e-06,
+      "loss": 0.5984,
+      "step": 23959
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.639868726518994e-06,
+      "loss": 0.5768,
+      "step": 23960
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.639162522418707e-06,
+      "loss": 0.6466,
+      "step": 23961
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.638456325233232e-06,
+      "loss": 0.6168,
+      "step": 23962
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.637750134966159e-06,
+      "loss": 0.5773,
+      "step": 23963
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.637043951621077e-06,
+      "loss": 0.5377,
+      "step": 23964
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.636337775201577e-06,
+      "loss": 0.567,
+      "step": 23965
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.635631605711242e-06,
+      "loss": 0.5654,
+      "step": 23966
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.63492544315366e-06,
+      "loss": 0.6634,
+      "step": 23967
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.634219287532423e-06,
+      "loss": 0.547,
+      "step": 23968
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.633513138851118e-06,
+      "loss": 0.6324,
+      "step": 23969
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.632806997113336e-06,
+      "loss": 0.5974,
+      "step": 23970
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.63210086232266e-06,
+      "loss": 0.5718,
+      "step": 23971
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.631394734482677e-06,
+      "loss": 0.5746,
+      "step": 23972
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.63068861359698e-06,
+      "loss": 0.6124,
+      "step": 23973
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.629982499669153e-06,
+      "loss": 0.6049,
+      "step": 23974
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.629276392702792e-06,
+      "loss": 0.56,
+      "step": 23975
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.628570292701476e-06,
+      "loss": 0.6053,
+      "step": 23976
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.627864199668795e-06,
+      "loss": 0.6058,
+      "step": 23977
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.627158113608341e-06,
+      "loss": 0.6068,
+      "step": 23978
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.626452034523696e-06,
+      "loss": 0.523,
+      "step": 23979
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.625745962418451e-06,
+      "loss": 0.6331,
+      "step": 23980
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.625039897296195e-06,
+      "loss": 0.5735,
+      "step": 23981
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.624333839160511e-06,
+      "loss": 0.5281,
+      "step": 23982
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.62362778801499e-06,
+      "loss": 0.638,
+      "step": 23983
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.622921743863219e-06,
+      "loss": 0.5315,
+      "step": 23984
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.622215706708788e-06,
+      "loss": 0.6614,
+      "step": 23985
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.62150967655528e-06,
+      "loss": 0.5787,
+      "step": 23986
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.62080365340629e-06,
+      "loss": 0.5476,
+      "step": 23987
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.620097637265396e-06,
+      "loss": 0.6039,
+      "step": 23988
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.61939162813619e-06,
+      "loss": 0.6091,
+      "step": 23989
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.61868562602226e-06,
+      "loss": 0.5997,
+      "step": 23990
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.617979630927192e-06,
+      "loss": 0.6246,
+      "step": 23991
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.61727364285458e-06,
+      "loss": 0.6329,
+      "step": 23992
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.616567661808e-06,
+      "loss": 0.5562,
+      "step": 23993
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.615861687791044e-06,
+      "loss": 0.5932,
+      "step": 23994
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.615155720807302e-06,
+      "loss": 0.6124,
+      "step": 23995
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.614449760860361e-06,
+      "loss": 0.5612,
+      "step": 23996
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.613743807953808e-06,
+      "loss": 0.6067,
+      "step": 23997
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.613037862091224e-06,
+      "loss": 0.6439,
+      "step": 23998
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.612331923276204e-06,
+      "loss": 0.649,
+      "step": 23999
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.61162599151233e-06,
+      "loss": 0.6094,
+      "step": 24000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.61092006680319e-06,
+      "loss": 0.6023,
+      "step": 24001
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.610214149152372e-06,
+      "loss": 0.6482,
+      "step": 24002
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.609508238563468e-06,
+      "loss": 0.6784,
+      "step": 24003
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.608802335040056e-06,
+      "loss": 0.6901,
+      "step": 24004
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.608096438585727e-06,
+      "loss": 0.6245,
+      "step": 24005
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.607390549204067e-06,
+      "loss": 0.5751,
+      "step": 24006
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.606684666898663e-06,
+      "loss": 0.6576,
+      "step": 24007
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.605978791673107e-06,
+      "loss": 0.5784,
+      "step": 24008
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.605272923530977e-06,
+      "loss": 0.6282,
+      "step": 24009
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.604567062475861e-06,
+      "loss": 0.5856,
+      "step": 24010
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.603861208511352e-06,
+      "loss": 0.6199,
+      "step": 24011
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.603155361641031e-06,
+      "loss": 0.599,
+      "step": 24012
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.60244952186849e-06,
+      "loss": 0.5455,
+      "step": 24013
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.601743689197313e-06,
+      "loss": 0.5275,
+      "step": 24014
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.601037863631082e-06,
+      "loss": 0.6124,
+      "step": 24015
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.600332045173387e-06,
+      "loss": 0.5857,
+      "step": 24016
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.599626233827818e-06,
+      "loss": 0.5718,
+      "step": 24017
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.598920429597956e-06,
+      "loss": 0.5886,
+      "step": 24018
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.598214632487391e-06,
+      "loss": 0.5853,
+      "step": 24019
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.59750884249971e-06,
+      "loss": 0.594,
+      "step": 24020
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.596803059638494e-06,
+      "loss": 0.6464,
+      "step": 24021
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.596097283907332e-06,
+      "loss": 0.6548,
+      "step": 24022
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.595391515309811e-06,
+      "loss": 0.5704,
+      "step": 24023
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.594685753849523e-06,
+      "loss": 0.5946,
+      "step": 24024
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.59397999953004e-06,
+      "loss": 0.6071,
+      "step": 24025
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.59327425235496e-06,
+      "loss": 0.6519,
+      "step": 24026
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.592568512327865e-06,
+      "loss": 0.5818,
+      "step": 24027
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.59186277945234e-06,
+      "loss": 0.608,
+      "step": 24028
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.591157053731974e-06,
+      "loss": 0.5989,
+      "step": 24029
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.590451335170356e-06,
+      "loss": 0.6266,
+      "step": 24030
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.589745623771063e-06,
+      "loss": 0.5717,
+      "step": 24031
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.589039919537685e-06,
+      "loss": 0.6223,
+      "step": 24032
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.588334222473808e-06,
+      "loss": 0.6694,
+      "step": 24033
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.587628532583022e-06,
+      "loss": 0.6142,
+      "step": 24034
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.586922849868908e-06,
+      "loss": 0.5625,
+      "step": 24035
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.586217174335051e-06,
+      "loss": 0.6367,
+      "step": 24036
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.58551150598504e-06,
+      "loss": 0.5644,
+      "step": 24037
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.584805844822459e-06,
+      "loss": 0.6776,
+      "step": 24038
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.584100190850892e-06,
+      "loss": 0.5948,
+      "step": 24039
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.583394544073928e-06,
+      "loss": 0.5427,
+      "step": 24040
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.582688904495155e-06,
+      "loss": 0.6129,
+      "step": 24041
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.581983272118149e-06,
+      "loss": 0.582,
+      "step": 24042
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.581277646946503e-06,
+      "loss": 0.6202,
+      "step": 24043
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.5805720289838e-06,
+      "loss": 0.6686,
+      "step": 24044
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.579866418233627e-06,
+      "loss": 0.6103,
+      "step": 24045
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.579160814699572e-06,
+      "loss": 0.5824,
+      "step": 24046
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.578455218385214e-06,
+      "loss": 0.6004,
+      "step": 24047
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.57774962929414e-06,
+      "loss": 0.5798,
+      "step": 24048
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.577044047429937e-06,
+      "loss": 0.514,
+      "step": 24049
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.57633847279619e-06,
+      "loss": 0.5754,
+      "step": 24050
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.57563290539649e-06,
+      "loss": 0.573,
+      "step": 24051
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.574927345234409e-06,
+      "loss": 0.6237,
+      "step": 24052
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.574221792313542e-06,
+      "loss": 0.6218,
+      "step": 24053
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.57351624663747e-06,
+      "loss": 0.6258,
+      "step": 24054
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.572810708209783e-06,
+      "loss": 0.6071,
+      "step": 24055
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.572105177034061e-06,
+      "loss": 0.5575,
+      "step": 24056
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.571399653113891e-06,
+      "loss": 0.5952,
+      "step": 24057
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.57069413645286e-06,
+      "loss": 0.5627,
+      "step": 24058
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.569988627054547e-06,
+      "loss": 0.5298,
+      "step": 24059
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.569283124922541e-06,
+      "loss": 0.6657,
+      "step": 24060
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.568577630060426e-06,
+      "loss": 0.587,
+      "step": 24061
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.567872142471792e-06,
+      "loss": 0.5479,
+      "step": 24062
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.567166662160214e-06,
+      "loss": 0.5867,
+      "step": 24063
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.566461189129284e-06,
+      "loss": 0.5545,
+      "step": 24064
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.565755723382583e-06,
+      "loss": 0.6057,
+      "step": 24065
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.565050264923697e-06,
+      "loss": 0.6243,
+      "step": 24066
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.564344813756212e-06,
+      "loss": 0.5967,
+      "step": 24067
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.563639369883713e-06,
+      "loss": 0.5792,
+      "step": 24068
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.562933933309781e-06,
+      "loss": 0.6533,
+      "step": 24069
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.562228504038003e-06,
+      "loss": 0.5931,
+      "step": 24070
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.561523082071962e-06,
+      "loss": 0.6461,
+      "step": 24071
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.560817667415246e-06,
+      "loss": 0.5971,
+      "step": 24072
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.560112260071437e-06,
+      "loss": 0.6108,
+      "step": 24073
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.559406860044116e-06,
+      "loss": 0.6106,
+      "step": 24074
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.558701467336874e-06,
+      "loss": 0.5885,
+      "step": 24075
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.55799608195329e-06,
+      "loss": 0.5931,
+      "step": 24076
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.55729070389695e-06,
+      "loss": 0.563,
+      "step": 24077
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.556585333171439e-06,
+      "loss": 0.6383,
+      "step": 24078
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.555879969780344e-06,
+      "loss": 0.6205,
+      "step": 24079
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.555174613727242e-06,
+      "loss": 0.5924,
+      "step": 24080
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.554469265015719e-06,
+      "loss": 0.573,
+      "step": 24081
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.553763923649362e-06,
+      "loss": 0.6138,
+      "step": 24082
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.553058589631757e-06,
+      "loss": 0.6134,
+      "step": 24083
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.552353262966485e-06,
+      "loss": 0.5552,
+      "step": 24084
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.55164794365713e-06,
+      "loss": 0.5666,
+      "step": 24085
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.550942631707273e-06,
+      "loss": 0.576,
+      "step": 24086
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.5502373271205e-06,
+      "loss": 0.5869,
+      "step": 24087
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.549532029900398e-06,
+      "loss": 0.535,
+      "step": 24088
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.548826740050553e-06,
+      "loss": 0.6307,
+      "step": 24089
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.548121457574538e-06,
+      "loss": 0.597,
+      "step": 24090
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.547416182475945e-06,
+      "loss": 0.6203,
+      "step": 24091
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.546710914758356e-06,
+      "loss": 0.6159,
+      "step": 24092
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.546005654425354e-06,
+      "loss": 0.6067,
+      "step": 24093
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.545300401480523e-06,
+      "loss": 0.6051,
+      "step": 24094
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.544595155927448e-06,
+      "loss": 0.5903,
+      "step": 24095
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.54388991776971e-06,
+      "loss": 0.5694,
+      "step": 24096
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.543184687010896e-06,
+      "loss": 0.6042,
+      "step": 24097
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.542479463654584e-06,
+      "loss": 0.6306,
+      "step": 24098
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.541774247704362e-06,
+      "loss": 0.6072,
+      "step": 24099
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.541069039163816e-06,
+      "loss": 0.5465,
+      "step": 24100
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.540363838036521e-06,
+      "loss": 0.5705,
+      "step": 24101
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.539658644326065e-06,
+      "loss": 0.61,
+      "step": 24102
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.53895345803603e-06,
+      "loss": 0.6066,
+      "step": 24103
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.538248279170001e-06,
+      "loss": 0.5819,
+      "step": 24104
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.537543107731561e-06,
+      "loss": 0.5651,
+      "step": 24105
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.536837943724296e-06,
+      "loss": 0.5349,
+      "step": 24106
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.536132787151782e-06,
+      "loss": 0.548,
+      "step": 24107
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.535427638017607e-06,
+      "loss": 0.6278,
+      "step": 24108
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.534722496325352e-06,
+      "loss": 0.5739,
+      "step": 24109
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.534017362078602e-06,
+      "loss": 0.595,
+      "step": 24110
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.533312235280944e-06,
+      "loss": 0.5816,
+      "step": 24111
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.53260711593595e-06,
+      "loss": 0.626,
+      "step": 24112
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.53190200404721e-06,
+      "loss": 0.5577,
+      "step": 24113
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.531196899618307e-06,
+      "loss": 0.634,
+      "step": 24114
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.530491802652823e-06,
+      "loss": 0.546,
+      "step": 24115
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.52978671315434e-06,
+      "loss": 0.6107,
+      "step": 24116
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.529081631126444e-06,
+      "loss": 0.5678,
+      "step": 24117
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.52837655657271e-06,
+      "loss": 0.575,
+      "step": 24118
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.527671489496727e-06,
+      "loss": 0.6385,
+      "step": 24119
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.526966429902076e-06,
+      "loss": 0.5999,
+      "step": 24120
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.526261377792341e-06,
+      "loss": 0.6205,
+      "step": 24121
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.525556333171107e-06,
+      "loss": 0.6079,
+      "step": 24122
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.524851296041948e-06,
+      "loss": 0.5719,
+      "step": 24123
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.524146266408453e-06,
+      "loss": 0.5925,
+      "step": 24124
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.523441244274203e-06,
+      "loss": 0.5732,
+      "step": 24125
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.52273622964278e-06,
+      "loss": 0.5517,
+      "step": 24126
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.52203122251777e-06,
+      "loss": 0.6363,
+      "step": 24127
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.52132622290275e-06,
+      "loss": 0.6349,
+      "step": 24128
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.520621230801303e-06,
+      "loss": 0.5667,
+      "step": 24129
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.519916246217014e-06,
+      "loss": 0.58,
+      "step": 24130
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.519211269153462e-06,
+      "loss": 0.5909,
+      "step": 24131
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.518506299614235e-06,
+      "loss": 0.6013,
+      "step": 24132
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.51780133760291e-06,
+      "loss": 0.5525,
+      "step": 24133
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.51709638312307e-06,
+      "loss": 0.689,
+      "step": 24134
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.516391436178299e-06,
+      "loss": 0.6154,
+      "step": 24135
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.515686496772176e-06,
+      "loss": 0.5892,
+      "step": 24136
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.514981564908284e-06,
+      "loss": 0.6109,
+      "step": 24137
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.514276640590212e-06,
+      "loss": 0.618,
+      "step": 24138
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.513571723821527e-06,
+      "loss": 0.5904,
+      "step": 24139
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.512866814605822e-06,
+      "loss": 0.5524,
+      "step": 24140
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.512161912946677e-06,
+      "loss": 0.6155,
+      "step": 24141
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.511457018847671e-06,
+      "loss": 0.5845,
+      "step": 24142
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.510752132312395e-06,
+      "loss": 0.5754,
+      "step": 24143
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.510047253344415e-06,
+      "loss": 0.5931,
+      "step": 24144
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.509342381947324e-06,
+      "loss": 0.5531,
+      "step": 24145
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.5086375181247e-06,
+      "loss": 0.6054,
+      "step": 24146
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.507932661880127e-06,
+      "loss": 0.6463,
+      "step": 24147
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.507227813217186e-06,
+      "loss": 0.5456,
+      "step": 24148
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.506522972139459e-06,
+      "loss": 0.6015,
+      "step": 24149
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.505818138650522e-06,
+      "loss": 0.6352,
+      "step": 24150
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.505113312753963e-06,
+      "loss": 0.5735,
+      "step": 24151
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.504408494453358e-06,
+      "loss": 0.5992,
+      "step": 24152
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.503703683752296e-06,
+      "loss": 0.5865,
+      "step": 24153
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.502998880654352e-06,
+      "loss": 0.6245,
+      "step": 24154
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.502294085163108e-06,
+      "loss": 0.5396,
+      "step": 24155
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.501589297282149e-06,
+      "loss": 0.6202,
+      "step": 24156
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.500884517015052e-06,
+      "loss": 0.6562,
+      "step": 24157
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.500179744365398e-06,
+      "loss": 0.6259,
+      "step": 24158
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.49947497933677e-06,
+      "loss": 0.5844,
+      "step": 24159
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.498770221932755e-06,
+      "loss": 0.6106,
+      "step": 24160
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.498065472156924e-06,
+      "loss": 0.6355,
+      "step": 24161
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.49736073001286e-06,
+      "loss": 0.5912,
+      "step": 24162
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.496655995504148e-06,
+      "loss": 0.5724,
+      "step": 24163
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.495951268634365e-06,
+      "loss": 0.5756,
+      "step": 24164
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.495246549407102e-06,
+      "loss": 0.5631,
+      "step": 24165
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.494541837825925e-06,
+      "loss": 0.5807,
+      "step": 24166
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.493837133894423e-06,
+      "loss": 0.5932,
+      "step": 24167
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.493132437616173e-06,
+      "loss": 0.5517,
+      "step": 24168
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.492427748994761e-06,
+      "loss": 0.606,
+      "step": 24169
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.491723068033768e-06,
+      "loss": 0.5802,
+      "step": 24170
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.491018394736769e-06,
+      "loss": 0.6502,
+      "step": 24171
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.490313729107348e-06,
+      "loss": 0.6136,
+      "step": 24172
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.489609071149086e-06,
+      "loss": 0.5433,
+      "step": 24173
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.48890442086556e-06,
+      "loss": 0.622,
+      "step": 24174
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.488199778260354e-06,
+      "loss": 0.5967,
+      "step": 24175
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.48749514333705e-06,
+      "loss": 0.6102,
+      "step": 24176
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.486790516099225e-06,
+      "loss": 0.602,
+      "step": 24177
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.48608589655046e-06,
+      "loss": 0.5565,
+      "step": 24178
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.485381284694334e-06,
+      "loss": 0.6105,
+      "step": 24179
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.484676680534432e-06,
+      "loss": 0.648,
+      "step": 24180
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.483972084074333e-06,
+      "loss": 0.6431,
+      "step": 24181
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.483267495317614e-06,
+      "loss": 0.6013,
+      "step": 24182
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.482562914267855e-06,
+      "loss": 0.6607,
+      "step": 24183
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.48185834092864e-06,
+      "loss": 0.6013,
+      "step": 24184
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.481153775303545e-06,
+      "loss": 0.615,
+      "step": 24185
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.480449217396154e-06,
+      "loss": 0.6533,
+      "step": 24186
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.47974466721005e-06,
+      "loss": 0.6276,
+      "step": 24187
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.479040124748802e-06,
+      "loss": 0.6426,
+      "step": 24188
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.478335590016e-06,
+      "loss": 0.6779,
+      "step": 24189
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.477631063015219e-06,
+      "loss": 0.5837,
+      "step": 24190
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.476926543750042e-06,
+      "loss": 0.5926,
+      "step": 24191
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.47622203222405e-06,
+      "loss": 0.5929,
+      "step": 24192
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.475517528440815e-06,
+      "loss": 0.571,
+      "step": 24193
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.474813032403924e-06,
+      "loss": 0.624,
+      "step": 24194
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.474108544116953e-06,
+      "loss": 0.5659,
+      "step": 24195
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.473404063583485e-06,
+      "loss": 0.5906,
+      "step": 24196
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.472699590807097e-06,
+      "loss": 0.5946,
+      "step": 24197
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.471995125791373e-06,
+      "loss": 0.6273,
+      "step": 24198
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.471290668539887e-06,
+      "loss": 0.6258,
+      "step": 24199
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.470586219056218e-06,
+      "loss": 0.5792,
+      "step": 24200
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.46988177734395e-06,
+      "loss": 0.6497,
+      "step": 24201
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.469177343406661e-06,
+      "loss": 0.6185,
+      "step": 24202
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.468472917247934e-06,
+      "loss": 0.5564,
+      "step": 24203
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.46776849887134e-06,
+      "loss": 0.6633,
+      "step": 24204
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.467064088280464e-06,
+      "loss": 0.6026,
+      "step": 24205
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.466359685478884e-06,
+      "loss": 0.6107,
+      "step": 24206
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.46565529047018e-06,
+      "loss": 0.6016,
+      "step": 24207
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.464950903257935e-06,
+      "loss": 0.591,
+      "step": 24208
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.464246523845719e-06,
+      "loss": 0.5866,
+      "step": 24209
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.463542152237116e-06,
+      "loss": 0.577,
+      "step": 24210
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.462837788435705e-06,
+      "loss": 0.5533,
+      "step": 24211
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.46213343244507e-06,
+      "loss": 0.5957,
+      "step": 24212
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.461429084268781e-06,
+      "loss": 0.6035,
+      "step": 24213
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.460724743910424e-06,
+      "loss": 0.6544,
+      "step": 24214
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.460020411373575e-06,
+      "loss": 0.6264,
+      "step": 24215
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.459316086661814e-06,
+      "loss": 0.6593,
+      "step": 24216
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.458611769778715e-06,
+      "loss": 0.5908,
+      "step": 24217
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.457907460727863e-06,
+      "loss": 0.59,
+      "step": 24218
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.457203159512839e-06,
+      "loss": 0.6234,
+      "step": 24219
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.456498866137212e-06,
+      "loss": 0.6069,
+      "step": 24220
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.455794580604568e-06,
+      "loss": 0.5466,
+      "step": 24221
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.455090302918483e-06,
+      "loss": 0.631,
+      "step": 24222
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.454386033082539e-06,
+      "loss": 0.5723,
+      "step": 24223
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.453681771100309e-06,
+      "loss": 0.5415,
+      "step": 24224
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.452977516975381e-06,
+      "loss": 0.5592,
+      "step": 24225
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.452273270711322e-06,
+      "loss": 0.5538,
+      "step": 24226
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.451569032311717e-06,
+      "loss": 0.5342,
+      "step": 24227
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.450864801780141e-06,
+      "loss": 0.6043,
+      "step": 24228
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.45016057912018e-06,
+      "loss": 0.5893,
+      "step": 24229
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.449456364335404e-06,
+      "loss": 0.535,
+      "step": 24230
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.448752157429394e-06,
+      "loss": 0.5971,
+      "step": 24231
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.448047958405728e-06,
+      "loss": 0.6337,
+      "step": 24232
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.447343767267987e-06,
+      "loss": 0.5687,
+      "step": 24233
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.446639584019745e-06,
+      "loss": 0.617,
+      "step": 24234
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.445935408664583e-06,
+      "loss": 0.6433,
+      "step": 24235
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.44523124120608e-06,
+      "loss": 0.5904,
+      "step": 24236
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.44452708164781e-06,
+      "loss": 0.6254,
+      "step": 24237
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.443822929993352e-06,
+      "loss": 0.5903,
+      "step": 24238
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.443118786246288e-06,
+      "loss": 0.5425,
+      "step": 24239
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.442414650410192e-06,
+      "loss": 0.6237,
+      "step": 24240
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.441710522488648e-06,
+      "loss": 0.6109,
+      "step": 24241
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.441006402485224e-06,
+      "loss": 0.635,
+      "step": 24242
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.440302290403504e-06,
+      "loss": 0.6419,
+      "step": 24243
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.439598186247064e-06,
+      "loss": 0.6208,
+      "step": 24244
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.438894090019482e-06,
+      "loss": 0.598,
+      "step": 24245
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.438190001724341e-06,
+      "loss": 0.6265,
+      "step": 24246
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.437485921365212e-06,
+      "loss": 0.6447,
+      "step": 24247
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.436781848945673e-06,
+      "loss": 0.6179,
+      "step": 24248
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.436077784469302e-06,
+      "loss": 0.575,
+      "step": 24249
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.435373727939681e-06,
+      "loss": 0.5852,
+      "step": 24250
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.434669679360383e-06,
+      "loss": 0.5733,
+      "step": 24251
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.43396563873499e-06,
+      "loss": 0.6487,
+      "step": 24252
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.43326160606707e-06,
+      "loss": 0.6124,
+      "step": 24253
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.432557581360211e-06,
+      "loss": 0.5927,
+      "step": 24254
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.431853564617986e-06,
+      "loss": 0.5935,
+      "step": 24255
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.431149555843969e-06,
+      "loss": 0.6033,
+      "step": 24256
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.430445555041747e-06,
+      "loss": 0.5909,
+      "step": 24257
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.429741562214885e-06,
+      "loss": 0.6403,
+      "step": 24258
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.429037577366967e-06,
+      "loss": 0.6107,
+      "step": 24259
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.428333600501567e-06,
+      "loss": 0.571,
+      "step": 24260
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.427629631622267e-06,
+      "loss": 0.6179,
+      "step": 24261
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.42692567073264e-06,
+      "loss": 0.5477,
+      "step": 24262
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.426221717836269e-06,
+      "loss": 0.642,
+      "step": 24263
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.425517772936723e-06,
+      "loss": 0.6166,
+      "step": 24264
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.424813836037582e-06,
+      "loss": 0.5834,
+      "step": 24265
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.424109907142421e-06,
+      "loss": 0.6179,
+      "step": 24266
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.423405986254821e-06,
+      "loss": 0.6484,
+      "step": 24267
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.422702073378361e-06,
+      "loss": 0.694,
+      "step": 24268
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.42199816851661e-06,
+      "loss": 0.6241,
+      "step": 24269
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.421294271673147e-06,
+      "loss": 0.5647,
+      "step": 24270
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.420590382851553e-06,
+      "loss": 0.5439,
+      "step": 24271
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.4198865020554e-06,
+      "loss": 0.6336,
+      "step": 24272
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.419182629288268e-06,
+      "loss": 0.63,
+      "step": 24273
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.418478764553733e-06,
+      "loss": 0.5186,
+      "step": 24274
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.417774907855368e-06,
+      "loss": 0.5675,
+      "step": 24275
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.417071059196751e-06,
+      "loss": 0.6161,
+      "step": 24276
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.416367218581461e-06,
+      "loss": 0.5601,
+      "step": 24277
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.41566338601307e-06,
+      "loss": 0.5728,
+      "step": 24278
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.414959561495165e-06,
+      "loss": 0.5945,
+      "step": 24279
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.41425574503131e-06,
+      "loss": 0.5672,
+      "step": 24280
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.413551936625085e-06,
+      "loss": 0.6067,
+      "step": 24281
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.412848136280067e-06,
+      "loss": 0.5767,
+      "step": 24282
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.412144343999833e-06,
+      "loss": 0.5574,
+      "step": 24283
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.411440559787962e-06,
+      "loss": 0.6145,
+      "step": 24284
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.410736783648024e-06,
+      "loss": 0.5793,
+      "step": 24285
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.410033015583598e-06,
+      "loss": 0.5789,
+      "step": 24286
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.409329255598257e-06,
+      "loss": 0.5967,
+      "step": 24287
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.408625503695584e-06,
+      "loss": 0.5891,
+      "step": 24288
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.40792175987915e-06,
+      "loss": 0.5775,
+      "step": 24289
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.407218024152532e-06,
+      "loss": 0.6121,
+      "step": 24290
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.406514296519304e-06,
+      "loss": 0.6543,
+      "step": 24291
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.405810576983047e-06,
+      "loss": 0.6084,
+      "step": 24292
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.40510686554733e-06,
+      "loss": 0.5784,
+      "step": 24293
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.404403162215732e-06,
+      "loss": 0.6284,
+      "step": 24294
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.403699466991833e-06,
+      "loss": 0.6125,
+      "step": 24295
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.4029957798792e-06,
+      "loss": 0.705,
+      "step": 24296
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.402292100881415e-06,
+      "loss": 0.6171,
+      "step": 24297
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.40158843000205e-06,
+      "loss": 0.5662,
+      "step": 24298
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.400884767244682e-06,
+      "loss": 0.5555,
+      "step": 24299
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.400181112612892e-06,
+      "loss": 0.6068,
+      "step": 24300
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.399477466110246e-06,
+      "loss": 0.6215,
+      "step": 24301
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.398773827740325e-06,
+      "loss": 0.6034,
+      "step": 24302
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.3980701975067e-06,
+      "loss": 0.5722,
+      "step": 24303
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.397366575412951e-06,
+      "loss": 0.6522,
+      "step": 24304
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.396662961462653e-06,
+      "loss": 0.6158,
+      "step": 24305
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.395959355659384e-06,
+      "loss": 0.5889,
+      "step": 24306
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.395255758006709e-06,
+      "loss": 0.5608,
+      "step": 24307
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.39455216850821e-06,
+      "loss": 0.5514,
+      "step": 24308
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.393848587167466e-06,
+      "loss": 0.548,
+      "step": 24309
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.393145013988043e-06,
+      "loss": 0.5699,
+      "step": 24310
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.392441448973525e-06,
+      "loss": 0.5827,
+      "step": 24311
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.391737892127478e-06,
+      "loss": 0.6432,
+      "step": 24312
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.391034343453487e-06,
+      "loss": 0.5993,
+      "step": 24313
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.390330802955118e-06,
+      "loss": 0.6038,
+      "step": 24314
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.38962727063595e-06,
+      "loss": 0.6134,
+      "step": 24315
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.388923746499557e-06,
+      "loss": 0.5757,
+      "step": 24316
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.388220230549519e-06,
+      "loss": 0.6105,
+      "step": 24317
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.3875167227894e-06,
+      "loss": 0.6414,
+      "step": 24318
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.386813223222783e-06,
+      "loss": 0.5764,
+      "step": 24319
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.38610973185324e-06,
+      "loss": 0.6097,
+      "step": 24320
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.385406248684346e-06,
+      "loss": 0.5812,
+      "step": 24321
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.38470277371968e-06,
+      "loss": 0.6056,
+      "step": 24322
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.383999306962807e-06,
+      "loss": 0.6041,
+      "step": 24323
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.383295848417307e-06,
+      "loss": 0.5911,
+      "step": 24324
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.382592398086755e-06,
+      "loss": 0.5829,
+      "step": 24325
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.381888955974726e-06,
+      "loss": 0.6118,
+      "step": 24326
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.381185522084794e-06,
+      "loss": 0.593,
+      "step": 24327
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.380482096420531e-06,
+      "loss": 0.5732,
+      "step": 24328
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.379778678985512e-06,
+      "loss": 0.5839,
+      "step": 24329
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.379075269783315e-06,
+      "loss": 0.6,
+      "step": 24330
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.378371868817508e-06,
+      "loss": 0.5973,
+      "step": 24331
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.377668476091667e-06,
+      "loss": 0.5903,
+      "step": 24332
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.376965091609372e-06,
+      "loss": 0.571,
+      "step": 24333
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.376261715374193e-06,
+      "loss": 0.553,
+      "step": 24334
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.3755583473897e-06,
+      "loss": 0.6133,
+      "step": 24335
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.37485498765947e-06,
+      "loss": 0.6272,
+      "step": 24336
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.37415163618708e-06,
+      "loss": 0.5954,
+      "step": 24337
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.373448292976106e-06,
+      "loss": 0.569,
+      "step": 24338
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.372744958030112e-06,
+      "loss": 0.6,
+      "step": 24339
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.372041631352678e-06,
+      "loss": 0.5924,
+      "step": 24340
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.371338312947377e-06,
+      "loss": 0.6391,
+      "step": 24341
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.370635002817784e-06,
+      "loss": 0.6129,
+      "step": 24342
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.369931700967471e-06,
+      "loss": 0.5782,
+      "step": 24343
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.369228407400016e-06,
+      "loss": 0.6374,
+      "step": 24344
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.368525122118984e-06,
+      "loss": 0.6002,
+      "step": 24345
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.367821845127955e-06,
+      "loss": 0.6001,
+      "step": 24346
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.367118576430502e-06,
+      "loss": 0.5946,
+      "step": 24347
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.3664153160302e-06,
+      "loss": 0.5724,
+      "step": 24348
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.36571206393062e-06,
+      "loss": 0.5895,
+      "step": 24349
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.365008820135332e-06,
+      "loss": 0.689,
+      "step": 24350
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.364305584647916e-06,
+      "loss": 0.6049,
+      "step": 24351
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.363602357471939e-06,
+      "loss": 0.6317,
+      "step": 24352
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.362899138610978e-06,
+      "loss": 0.6054,
+      "step": 24353
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.362195928068609e-06,
+      "loss": 0.6315,
+      "step": 24354
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.361492725848404e-06,
+      "loss": 0.637,
+      "step": 24355
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.36078953195393e-06,
+      "loss": 0.5357,
+      "step": 24356
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.360086346388764e-06,
+      "loss": 0.5541,
+      "step": 24357
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.359383169156481e-06,
+      "loss": 0.5952,
+      "step": 24358
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.358680000260651e-06,
+      "loss": 0.5987,
+      "step": 24359
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.357976839704853e-06,
+      "loss": 0.5721,
+      "step": 24360
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.357273687492653e-06,
+      "loss": 0.6144,
+      "step": 24361
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.356570543627624e-06,
+      "loss": 0.5639,
+      "step": 24362
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.355867408113344e-06,
+      "loss": 0.5221,
+      "step": 24363
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.355164280953382e-06,
+      "loss": 0.5744,
+      "step": 24364
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.354461162151316e-06,
+      "loss": 0.6369,
+      "step": 24365
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.35375805171071e-06,
+      "loss": 0.5671,
+      "step": 24366
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.353054949635143e-06,
+      "loss": 0.5851,
+      "step": 24367
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.352351855928187e-06,
+      "loss": 0.5915,
+      "step": 24368
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.351648770593412e-06,
+      "loss": 0.6275,
+      "step": 24369
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.350945693634394e-06,
+      "loss": 0.6202,
+      "step": 24370
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.350242625054702e-06,
+      "loss": 0.5678,
+      "step": 24371
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.349539564857914e-06,
+      "loss": 0.6417,
+      "step": 24372
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.348836513047596e-06,
+      "loss": 0.6249,
+      "step": 24373
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.348133469627324e-06,
+      "loss": 0.52,
+      "step": 24374
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.347430434600667e-06,
+      "loss": 0.5717,
+      "step": 24375
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.346727407971206e-06,
+      "loss": 0.6064,
+      "step": 24376
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.346024389742502e-06,
+      "loss": 0.6387,
+      "step": 24377
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.345321379918133e-06,
+      "loss": 0.5726,
+      "step": 24378
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.34461837850167e-06,
+      "loss": 0.5683,
+      "step": 24379
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.343915385496687e-06,
+      "loss": 0.6429,
+      "step": 24380
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.343212400906755e-06,
+      "loss": 0.5664,
+      "step": 24381
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.34250942473545e-06,
+      "loss": 0.6002,
+      "step": 24382
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.341806456986335e-06,
+      "loss": 0.6066,
+      "step": 24383
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.341103497662989e-06,
+      "loss": 0.6142,
+      "step": 24384
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.340400546768979e-06,
+      "loss": 0.64,
+      "step": 24385
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.339697604307883e-06,
+      "loss": 0.5805,
+      "step": 24386
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.338994670283271e-06,
+      "loss": 0.598,
+      "step": 24387
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.33829174469871e-06,
+      "loss": 0.5231,
+      "step": 24388
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.337588827557776e-06,
+      "loss": 0.6392,
+      "step": 24389
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.336885918864042e-06,
+      "loss": 0.6516,
+      "step": 24390
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.336183018621076e-06,
+      "loss": 0.6389,
+      "step": 24391
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.335480126832452e-06,
+      "loss": 0.5915,
+      "step": 24392
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.334777243501742e-06,
+      "loss": 0.5901,
+      "step": 24393
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.334074368632516e-06,
+      "loss": 0.6058,
+      "step": 24394
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.333371502228343e-06,
+      "loss": 0.585,
+      "step": 24395
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.332668644292799e-06,
+      "loss": 0.6216,
+      "step": 24396
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.331965794829455e-06,
+      "loss": 0.6403,
+      "step": 24397
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.331262953841885e-06,
+      "loss": 0.577,
+      "step": 24398
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.330560121333651e-06,
+      "loss": 0.5892,
+      "step": 24399
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.329857297308331e-06,
+      "loss": 0.5978,
+      "step": 24400
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.329154481769495e-06,
+      "loss": 0.592,
+      "step": 24401
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.328451674720714e-06,
+      "loss": 0.6261,
+      "step": 24402
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.327748876165564e-06,
+      "loss": 0.5774,
+      "step": 24403
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.327046086107607e-06,
+      "loss": 0.622,
+      "step": 24404
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.32634330455042e-06,
+      "loss": 0.6471,
+      "step": 24405
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.325640531497572e-06,
+      "loss": 0.5651,
+      "step": 24406
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.324937766952638e-06,
+      "loss": 0.5867,
+      "step": 24407
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.324235010919181e-06,
+      "loss": 0.5944,
+      "step": 24408
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.323532263400782e-06,
+      "loss": 0.64,
+      "step": 24409
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.322829524401005e-06,
+      "loss": 0.6167,
+      "step": 24410
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.32212679392342e-06,
+      "loss": 0.5668,
+      "step": 24411
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.3214240719716e-06,
+      "loss": 0.5739,
+      "step": 24412
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.320721358549118e-06,
+      "loss": 0.6429,
+      "step": 24413
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.320018653659545e-06,
+      "loss": 0.5945,
+      "step": 24414
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.319315957306445e-06,
+      "loss": 0.5875,
+      "step": 24415
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.318613269493392e-06,
+      "loss": 0.608,
+      "step": 24416
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.317910590223958e-06,
+      "loss": 0.5409,
+      "step": 24417
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.317207919501715e-06,
+      "loss": 0.6029,
+      "step": 24418
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.316505257330233e-06,
+      "loss": 0.6207,
+      "step": 24419
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.315802603713078e-06,
+      "loss": 0.6031,
+      "step": 24420
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.315099958653822e-06,
+      "loss": 0.6323,
+      "step": 24421
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.314397322156038e-06,
+      "loss": 0.559,
+      "step": 24422
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.313694694223294e-06,
+      "loss": 0.5696,
+      "step": 24423
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.312992074859162e-06,
+      "loss": 0.5432,
+      "step": 24424
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.312289464067215e-06,
+      "loss": 0.5622,
+      "step": 24425
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.311586861851015e-06,
+      "loss": 0.6196,
+      "step": 24426
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.310884268214136e-06,
+      "loss": 0.59,
+      "step": 24427
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.310181683160153e-06,
+      "loss": 0.6376,
+      "step": 24428
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.309479106692629e-06,
+      "loss": 0.5667,
+      "step": 24429
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.308776538815137e-06,
+      "loss": 0.6404,
+      "step": 24430
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.308073979531249e-06,
+      "loss": 0.5852,
+      "step": 24431
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.307371428844528e-06,
+      "loss": 0.5792,
+      "step": 24432
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.30666888675855e-06,
+      "loss": 0.5927,
+      "step": 24433
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.305966353276883e-06,
+      "loss": 0.5859,
+      "step": 24434
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.305263828403097e-06,
+      "loss": 0.6231,
+      "step": 24435
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.304561312140767e-06,
+      "loss": 0.6681,
+      "step": 24436
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.303858804493452e-06,
+      "loss": 0.6112,
+      "step": 24437
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.303156305464728e-06,
+      "loss": 0.6213,
+      "step": 24438
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.302453815058162e-06,
+      "loss": 0.6016,
+      "step": 24439
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.301751333277327e-06,
+      "loss": 0.5788,
+      "step": 24440
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.301048860125795e-06,
+      "loss": 0.6252,
+      "step": 24441
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.300346395607128e-06,
+      "loss": 0.6366,
+      "step": 24442
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.299643939724898e-06,
+      "loss": 0.6111,
+      "step": 24443
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.298941492482674e-06,
+      "loss": 0.6201,
+      "step": 24444
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.298239053884028e-06,
+      "loss": 0.5447,
+      "step": 24445
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.297536623932528e-06,
+      "loss": 0.5855,
+      "step": 24446
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.296834202631744e-06,
+      "loss": 0.5942,
+      "step": 24447
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.296131789985243e-06,
+      "loss": 0.5992,
+      "step": 24448
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.295429385996597e-06,
+      "loss": 0.58,
+      "step": 24449
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.294726990669369e-06,
+      "loss": 0.5999,
+      "step": 24450
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.294024604007136e-06,
+      "loss": 0.5585,
+      "step": 24451
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.293322226013467e-06,
+      "loss": 0.6332,
+      "step": 24452
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.292619856691921e-06,
+      "loss": 0.6047,
+      "step": 24453
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.291917496046076e-06,
+      "loss": 0.5978,
+      "step": 24454
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.291215144079497e-06,
+      "loss": 0.5812,
+      "step": 24455
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.290512800795756e-06,
+      "loss": 0.6625,
+      "step": 24456
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.289810466198422e-06,
+      "loss": 0.6301,
+      "step": 24457
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.28910814029106e-06,
+      "loss": 0.5989,
+      "step": 24458
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.28840582307724e-06,
+      "loss": 0.5465,
+      "step": 24459
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.28770351456053e-06,
+      "loss": 0.6531,
+      "step": 24460
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.2870012147445e-06,
+      "loss": 0.5658,
+      "step": 24461
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.28629892363272e-06,
+      "loss": 0.6501,
+      "step": 24462
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.28559664122876e-06,
+      "loss": 0.614,
+      "step": 24463
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.284894367536183e-06,
+      "loss": 0.6036,
+      "step": 24464
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.284192102558559e-06,
+      "loss": 0.6104,
+      "step": 24465
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.28348984629946e-06,
+      "loss": 0.6204,
+      "step": 24466
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.282787598762448e-06,
+      "loss": 0.6163,
+      "step": 24467
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.2820853599511e-06,
+      "loss": 0.5965,
+      "step": 24468
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.281383129868975e-06,
+      "loss": 0.6323,
+      "step": 24469
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.280680908519649e-06,
+      "loss": 0.5539,
+      "step": 24470
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.279978695906683e-06,
+      "loss": 0.6015,
+      "step": 24471
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.27927649203365e-06,
+      "loss": 0.6327,
+      "step": 24472
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.278574296904118e-06,
+      "loss": 0.6112,
+      "step": 24473
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.277872110521658e-06,
+      "loss": 0.639,
+      "step": 24474
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.27716993288983e-06,
+      "loss": 0.6318,
+      "step": 24475
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.276467764012203e-06,
+      "loss": 0.6294,
+      "step": 24476
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.275765603892352e-06,
+      "loss": 0.6034,
+      "step": 24477
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.27506345253384e-06,
+      "loss": 0.5277,
+      "step": 24478
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.27436130994024e-06,
+      "loss": 0.5593,
+      "step": 24479
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.27365917611511e-06,
+      "loss": 0.5883,
+      "step": 24480
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.272957051062025e-06,
+      "loss": 0.5715,
+      "step": 24481
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.272254934784549e-06,
+      "loss": 0.6088,
+      "step": 24482
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.271552827286254e-06,
+      "loss": 0.6076,
+      "step": 24483
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.270850728570708e-06,
+      "loss": 0.5429,
+      "step": 24484
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.270148638641472e-06,
+      "loss": 0.5541,
+      "step": 24485
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.269446557502117e-06,
+      "loss": 0.6398,
+      "step": 24486
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.268744485156215e-06,
+      "loss": 0.6265,
+      "step": 24487
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.268042421607325e-06,
+      "loss": 0.5589,
+      "step": 24488
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.26734036685902e-06,
+      "loss": 0.5782,
+      "step": 24489
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.266638320914868e-06,
+      "loss": 0.6249,
+      "step": 24490
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.265936283778431e-06,
+      "loss": 0.6108,
+      "step": 24491
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.26523425545328e-06,
+      "loss": 0.5801,
+      "step": 24492
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.264532235942984e-06,
+      "loss": 0.6004,
+      "step": 24493
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.263830225251105e-06,
+      "loss": 0.6284,
+      "step": 24494
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.263128223381218e-06,
+      "loss": 0.6544,
+      "step": 24495
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.262426230336881e-06,
+      "loss": 0.6547,
+      "step": 24496
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.261724246121667e-06,
+      "loss": 0.5943,
+      "step": 24497
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.261022270739138e-06,
+      "loss": 0.6034,
+      "step": 24498
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.260320304192866e-06,
+      "loss": 0.6219,
+      "step": 24499
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.259618346486415e-06,
+      "loss": 0.6065,
+      "step": 24500
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25891639762336e-06,
+      "loss": 0.5434,
+      "step": 24501
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.258214457607253e-06,
+      "loss": 0.6151,
+      "step": 24502
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25751252644167e-06,
+      "loss": 0.6144,
+      "step": 24503
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.256810604130177e-06,
+      "loss": 0.5978,
+      "step": 24504
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25610869067634e-06,
+      "loss": 0.5327,
+      "step": 24505
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.255406786083726e-06,
+      "loss": 0.5999,
+      "step": 24506
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.2547048903559e-06,
+      "loss": 0.602,
+      "step": 24507
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25400300349643e-06,
+      "loss": 0.6587,
+      "step": 24508
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.253301125508882e-06,
+      "loss": 0.5805,
+      "step": 24509
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25259925639682e-06,
+      "loss": 0.5404,
+      "step": 24510
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.25189739616382e-06,
+      "loss": 0.6381,
+      "step": 24511
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.251195544813434e-06,
+      "loss": 0.4934,
+      "step": 24512
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.250493702349238e-06,
+      "loss": 0.587,
+      "step": 24513
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.249791868774796e-06,
+      "loss": 0.5919,
+      "step": 24514
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.249090044093672e-06,
+      "loss": 0.6003,
+      "step": 24515
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.248388228309436e-06,
+      "loss": 0.6001,
+      "step": 24516
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.247686421425657e-06,
+      "loss": 0.6485,
+      "step": 24517
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.24698462344589e-06,
+      "loss": 0.5845,
+      "step": 24518
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.24628283437371e-06,
+      "loss": 0.6094,
+      "step": 24519
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.24558105421268e-06,
+      "loss": 0.5736,
+      "step": 24520
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.244879282966364e-06,
+      "loss": 0.6132,
+      "step": 24521
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.244177520638337e-06,
+      "loss": 0.6058,
+      "step": 24522
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.243475767232154e-06,
+      "loss": 0.5929,
+      "step": 24523
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.242774022751384e-06,
+      "loss": 0.6443,
+      "step": 24524
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.242072287199596e-06,
+      "loss": 0.646,
+      "step": 24525
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.241370560580353e-06,
+      "loss": 0.6686,
+      "step": 24526
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.240668842897219e-06,
+      "loss": 0.6058,
+      "step": 24527
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.239967134153766e-06,
+      "loss": 0.6121,
+      "step": 24528
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.239265434353554e-06,
+      "loss": 0.5801,
+      "step": 24529
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.238563743500149e-06,
+      "loss": 0.6116,
+      "step": 24530
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.237862061597117e-06,
+      "loss": 0.6031,
+      "step": 24531
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.237160388648024e-06,
+      "loss": 0.5771,
+      "step": 24532
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.236458724656439e-06,
+      "loss": 0.6055,
+      "step": 24533
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.235757069625921e-06,
+      "loss": 0.5835,
+      "step": 24534
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.235055423560035e-06,
+      "loss": 0.6197,
+      "step": 24535
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.234353786462351e-06,
+      "loss": 0.6643,
+      "step": 24536
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.233652158336434e-06,
+      "loss": 0.6262,
+      "step": 24537
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.232950539185845e-06,
+      "loss": 0.5866,
+      "step": 24538
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.232248929014158e-06,
+      "loss": 0.5486,
+      "step": 24539
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.231547327824926e-06,
+      "loss": 0.6227,
+      "step": 24540
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.230845735621722e-06,
+      "loss": 0.6047,
+      "step": 24541
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.230144152408108e-06,
+      "loss": 0.6183,
+      "step": 24542
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.229442578187652e-06,
+      "loss": 0.6308,
+      "step": 24543
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.228741012963917e-06,
+      "loss": 0.5984,
+      "step": 24544
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.228039456740465e-06,
+      "loss": 0.5795,
+      "step": 24545
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.227337909520864e-06,
+      "loss": 0.5759,
+      "step": 24546
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.22663637130868e-06,
+      "loss": 0.6454,
+      "step": 24547
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.225934842107474e-06,
+      "loss": 0.6136,
+      "step": 24548
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.225233321920813e-06,
+      "loss": 0.5518,
+      "step": 24549
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.224531810752265e-06,
+      "loss": 0.6431,
+      "step": 24550
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.223830308605385e-06,
+      "loss": 0.5835,
+      "step": 24551
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.223128815483746e-06,
+      "loss": 0.5925,
+      "step": 24552
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.222427331390909e-06,
+      "loss": 0.5691,
+      "step": 24553
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.221725856330438e-06,
+      "loss": 0.5839,
+      "step": 24554
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.221024390305905e-06,
+      "loss": 0.5782,
+      "step": 24555
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.220322933320863e-06,
+      "loss": 0.6016,
+      "step": 24556
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.219621485378882e-06,
+      "loss": 0.5886,
+      "step": 24557
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.218920046483523e-06,
+      "loss": 0.6426,
+      "step": 24558
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.218218616638355e-06,
+      "loss": 0.6202,
+      "step": 24559
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.217517195846945e-06,
+      "loss": 0.5757,
+      "step": 24560
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.216815784112846e-06,
+      "loss": 0.6151,
+      "step": 24561
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.21611438143963e-06,
+      "loss": 0.611,
+      "step": 24562
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.215412987830859e-06,
+      "loss": 0.6082,
+      "step": 24563
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.214711603290099e-06,
+      "loss": 0.5593,
+      "step": 24564
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.21401022782091e-06,
+      "loss": 0.5754,
+      "step": 24565
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.21330886142686e-06,
+      "loss": 0.5764,
+      "step": 24566
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.21260750411151e-06,
+      "loss": 0.6058,
+      "step": 24567
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.211906155878424e-06,
+      "loss": 0.5805,
+      "step": 24568
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.211204816731166e-06,
+      "loss": 0.5537,
+      "step": 24569
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.2105034866733e-06,
+      "loss": 0.6191,
+      "step": 24570
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.209802165708396e-06,
+      "loss": 0.6169,
+      "step": 24571
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.209100853840005e-06,
+      "loss": 0.6007,
+      "step": 24572
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.208399551071697e-06,
+      "loss": 0.6123,
+      "step": 24573
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.207698257407037e-06,
+      "loss": 0.634,
+      "step": 24574
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.206996972849587e-06,
+      "loss": 0.557,
+      "step": 24575
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.206295697402915e-06,
+      "loss": 0.6338,
+      "step": 24576
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.205594431070574e-06,
+      "loss": 0.6146,
+      "step": 24577
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.204893173856134e-06,
+      "loss": 0.6071,
+      "step": 24578
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.204191925763156e-06,
+      "loss": 0.6445,
+      "step": 24579
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.203490686795206e-06,
+      "loss": 0.6288,
+      "step": 24580
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.202789456955847e-06,
+      "loss": 0.5741,
+      "step": 24581
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.202088236248644e-06,
+      "loss": 0.6296,
+      "step": 24582
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.201387024677155e-06,
+      "loss": 0.6216,
+      "step": 24583
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.200685822244943e-06,
+      "loss": 0.595,
+      "step": 24584
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.199984628955576e-06,
+      "loss": 0.6109,
+      "step": 24585
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.199283444812613e-06,
+      "loss": 0.6109,
+      "step": 24586
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.19858226981962e-06,
+      "loss": 0.6309,
+      "step": 24587
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.197881103980157e-06,
+      "loss": 0.5974,
+      "step": 24588
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.197179947297787e-06,
+      "loss": 0.646,
+      "step": 24589
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.196478799776073e-06,
+      "loss": 0.5796,
+      "step": 24590
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.195777661418578e-06,
+      "loss": 0.6027,
+      "step": 24591
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.195076532228865e-06,
+      "loss": 0.5999,
+      "step": 24592
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.194375412210502e-06,
+      "loss": 0.5679,
+      "step": 24593
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.193674301367041e-06,
+      "loss": 0.5992,
+      "step": 24594
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.192973199702051e-06,
+      "loss": 0.6322,
+      "step": 24595
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.192272107219094e-06,
+      "loss": 0.6192,
+      "step": 24596
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.19157102392173e-06,
+      "loss": 0.5556,
+      "step": 24597
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.19086994981353e-06,
+      "loss": 0.6087,
+      "step": 24598
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.190168884898041e-06,
+      "loss": 0.577,
+      "step": 24599
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.189467829178838e-06,
+      "loss": 0.6158,
+      "step": 24600
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.188766782659477e-06,
+      "loss": 0.5958,
+      "step": 24601
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.188065745343525e-06,
+      "loss": 0.6348,
+      "step": 24602
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.187364717234542e-06,
+      "loss": 0.6088,
+      "step": 24603
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.186663698336088e-06,
+      "loss": 0.5899,
+      "step": 24604
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.185962688651727e-06,
+      "loss": 0.5692,
+      "step": 24605
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.185261688185023e-06,
+      "loss": 0.6361,
+      "step": 24606
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.184560696939532e-06,
+      "loss": 0.633,
+      "step": 24607
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.18385971491882e-06,
+      "loss": 0.6035,
+      "step": 24608
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.183158742126455e-06,
+      "loss": 0.6647,
+      "step": 24609
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.182457778565986e-06,
+      "loss": 0.6167,
+      "step": 24610
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.181756824240983e-06,
+      "loss": 0.5806,
+      "step": 24611
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.181055879155005e-06,
+      "loss": 0.5847,
+      "step": 24612
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.180354943311614e-06,
+      "loss": 0.6511,
+      "step": 24613
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.179654016714379e-06,
+      "loss": 0.598,
+      "step": 24614
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.17895309936685e-06,
+      "loss": 0.6586,
+      "step": 24615
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.178252191272594e-06,
+      "loss": 0.6204,
+      "step": 24616
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.177551292435171e-06,
+      "loss": 0.6118,
+      "step": 24617
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.176850402858143e-06,
+      "loss": 0.5672,
+      "step": 24618
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.176149522545074e-06,
+      "loss": 0.6124,
+      "step": 24619
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.175448651499528e-06,
+      "loss": 0.6805,
+      "step": 24620
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.174747789725055e-06,
+      "loss": 0.5523,
+      "step": 24621
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.174046937225225e-06,
+      "loss": 0.5947,
+      "step": 24622
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.173346094003597e-06,
+      "loss": 0.573,
+      "step": 24623
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.172645260063731e-06,
+      "loss": 0.5076,
+      "step": 24624
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.171944435409193e-06,
+      "loss": 0.5448,
+      "step": 24625
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.17124362004354e-06,
+      "loss": 0.6053,
+      "step": 24626
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.170542813970333e-06,
+      "loss": 0.6755,
+      "step": 24627
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.169842017193133e-06,
+      "loss": 0.6256,
+      "step": 24628
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.169141229715501e-06,
+      "loss": 0.6084,
+      "step": 24629
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.168440451541e-06,
+      "loss": 0.6327,
+      "step": 24630
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.167739682673193e-06,
+      "loss": 0.6092,
+      "step": 24631
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.167038923115632e-06,
+      "loss": 0.5516,
+      "step": 24632
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.166338172871883e-06,
+      "loss": 0.6212,
+      "step": 24633
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.165637431945508e-06,
+      "loss": 0.602,
+      "step": 24634
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.164936700340067e-06,
+      "loss": 0.6033,
+      "step": 24635
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.164235978059123e-06,
+      "loss": 0.6597,
+      "step": 24636
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.16353526510623e-06,
+      "loss": 0.593,
+      "step": 24637
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.162834561484951e-06,
+      "loss": 0.6129,
+      "step": 24638
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.16213386719885e-06,
+      "loss": 0.6008,
+      "step": 24639
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.161433182251483e-06,
+      "loss": 0.6274,
+      "step": 24640
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.160732506646417e-06,
+      "loss": 0.5679,
+      "step": 24641
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.160031840387204e-06,
+      "loss": 0.5516,
+      "step": 24642
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.159331183477407e-06,
+      "loss": 0.6649,
+      "step": 24643
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.15863053592059e-06,
+      "loss": 0.5762,
+      "step": 24644
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.15792989772031e-06,
+      "loss": 0.5841,
+      "step": 24645
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.157229268880126e-06,
+      "loss": 0.614,
+      "step": 24646
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.156528649403605e-06,
+      "loss": 0.6438,
+      "step": 24647
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.155828039294296e-06,
+      "loss": 0.602,
+      "step": 24648
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.155127438555765e-06,
+      "loss": 0.6101,
+      "step": 24649
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.154426847191573e-06,
+      "loss": 0.6223,
+      "step": 24650
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.153726265205278e-06,
+      "loss": 0.6116,
+      "step": 24651
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.153025692600447e-06,
+      "loss": 0.6264,
+      "step": 24652
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.152325129380627e-06,
+      "loss": 0.5526,
+      "step": 24653
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.151624575549383e-06,
+      "loss": 0.6554,
+      "step": 24654
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.150924031110276e-06,
+      "loss": 0.5695,
+      "step": 24655
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.150223496066868e-06,
+      "loss": 0.6115,
+      "step": 24656
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.149522970422714e-06,
+      "loss": 0.5889,
+      "step": 24657
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.148822454181381e-06,
+      "loss": 0.6243,
+      "step": 24658
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.148121947346419e-06,
+      "loss": 0.6422,
+      "step": 24659
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.147421449921393e-06,
+      "loss": 0.6415,
+      "step": 24660
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.146720961909859e-06,
+      "loss": 0.5958,
+      "step": 24661
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.146020483315381e-06,
+      "loss": 0.6213,
+      "step": 24662
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.145320014141516e-06,
+      "loss": 0.5824,
+      "step": 24663
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.14461955439182e-06,
+      "loss": 0.6084,
+      "step": 24664
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.14391910406986e-06,
+      "loss": 0.5035,
+      "step": 24665
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.143218663179187e-06,
+      "loss": 0.5985,
+      "step": 24666
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.142518231723364e-06,
+      "loss": 0.6149,
+      "step": 24667
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.141817809705956e-06,
+      "loss": 0.5917,
+      "step": 24668
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.141117397130508e-06,
+      "loss": 0.5347,
+      "step": 24669
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.14041699400059e-06,
+      "loss": 0.6264,
+      "step": 24670
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.139716600319756e-06,
+      "loss": 0.6071,
+      "step": 24671
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.139016216091568e-06,
+      "loss": 0.5888,
+      "step": 24672
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.138315841319582e-06,
+      "loss": 0.6474,
+      "step": 24673
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.137615476007364e-06,
+      "loss": 0.6275,
+      "step": 24674
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.136915120158464e-06,
+      "loss": 0.5971,
+      "step": 24675
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.13621477377644e-06,
+      "loss": 0.6169,
+      "step": 24676
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.13551443686486e-06,
+      "loss": 0.5634,
+      "step": 24677
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.134814109427272e-06,
+      "loss": 0.6085,
+      "step": 24678
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.134113791467247e-06,
+      "loss": 0.5622,
+      "step": 24679
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.133413482988332e-06,
+      "loss": 0.6422,
+      "step": 24680
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.132713183994087e-06,
+      "loss": 0.5118,
+      "step": 24681
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.132012894488075e-06,
+      "loss": 0.5707,
+      "step": 24682
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.131312614473857e-06,
+      "loss": 0.5992,
+      "step": 24683
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.130612343954981e-06,
+      "loss": 0.5695,
+      "step": 24684
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.129912082935016e-06,
+      "loss": 0.6028,
+      "step": 24685
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.129211831417514e-06,
+      "loss": 0.6132,
+      "step": 24686
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.128511589406033e-06,
+      "loss": 0.6585,
+      "step": 24687
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.127811356904133e-06,
+      "loss": 0.6473,
+      "step": 24688
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.12711113391537e-06,
+      "loss": 0.5365,
+      "step": 24689
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.12641092044331e-06,
+      "loss": 0.6104,
+      "step": 24690
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.125710716491499e-06,
+      "loss": 0.5767,
+      "step": 24691
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.125010522063502e-06,
+      "loss": 0.6153,
+      "step": 24692
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.124310337162876e-06,
+      "loss": 0.5503,
+      "step": 24693
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.123610161793178e-06,
+      "loss": 0.6519,
+      "step": 24694
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.122909995957971e-06,
+      "loss": 0.5648,
+      "step": 24695
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.122209839660802e-06,
+      "loss": 0.5689,
+      "step": 24696
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.121509692905236e-06,
+      "loss": 0.6062,
+      "step": 24697
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.12080955569483e-06,
+      "loss": 0.5284,
+      "step": 24698
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.12010942803314e-06,
+      "loss": 0.5917,
+      "step": 24699
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.119409309923726e-06,
+      "loss": 0.6174,
+      "step": 24700
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.118709201370145e-06,
+      "loss": 0.5511,
+      "step": 24701
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.118009102375952e-06,
+      "loss": 0.6217,
+      "step": 24702
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.117309012944707e-06,
+      "loss": 0.5667,
+      "step": 24703
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.116608933079966e-06,
+      "loss": 0.5281,
+      "step": 24704
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.115908862785285e-06,
+      "loss": 0.6412,
+      "step": 24705
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.115208802064226e-06,
+      "loss": 0.5768,
+      "step": 24706
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.114508750920341e-06,
+      "loss": 0.655,
+      "step": 24707
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.113808709357188e-06,
+      "loss": 0.6143,
+      "step": 24708
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.113108677378325e-06,
+      "loss": 0.5492,
+      "step": 24709
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.11240865498731e-06,
+      "loss": 0.6402,
+      "step": 24710
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.1117086421877e-06,
+      "loss": 0.6459,
+      "step": 24711
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.111008638983056e-06,
+      "loss": 0.5336,
+      "step": 24712
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.110308645376925e-06,
+      "loss": 0.5964,
+      "step": 24713
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.10960866137287e-06,
+      "loss": 0.5962,
+      "step": 24714
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.108908686974445e-06,
+      "loss": 0.5426,
+      "step": 24715
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.108208722185212e-06,
+      "loss": 0.6044,
+      "step": 24716
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.107508767008725e-06,
+      "loss": 0.6028,
+      "step": 24717
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.10680882144854e-06,
+      "loss": 0.5751,
+      "step": 24718
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.10610888550821e-06,
+      "loss": 0.6305,
+      "step": 24719
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.105408959191298e-06,
+      "loss": 0.6304,
+      "step": 24720
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.104709042501359e-06,
+      "loss": 0.604,
+      "step": 24721
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.104009135441946e-06,
+      "loss": 0.6392,
+      "step": 24722
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.103309238016621e-06,
+      "loss": 0.6356,
+      "step": 24723
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.102609350228936e-06,
+      "loss": 0.6533,
+      "step": 24724
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.101909472082448e-06,
+      "loss": 0.6331,
+      "step": 24725
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.101209603580714e-06,
+      "loss": 0.6172,
+      "step": 24726
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.100509744727287e-06,
+      "loss": 0.6065,
+      "step": 24727
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.099809895525734e-06,
+      "loss": 0.6185,
+      "step": 24728
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.099110055979597e-06,
+      "loss": 0.5787,
+      "step": 24729
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.098410226092439e-06,
+      "loss": 0.5869,
+      "step": 24730
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.097710405867817e-06,
+      "loss": 0.6244,
+      "step": 24731
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.097010595309283e-06,
+      "loss": 0.6029,
+      "step": 24732
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.096310794420402e-06,
+      "loss": 0.6075,
+      "step": 24733
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.095611003204718e-06,
+      "loss": 0.6277,
+      "step": 24734
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.094911221665793e-06,
+      "loss": 0.5386,
+      "step": 24735
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.094211449807182e-06,
+      "loss": 0.5823,
+      "step": 24736
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.093511687632439e-06,
+      "loss": 0.5752,
+      "step": 24737
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.092811935145126e-06,
+      "loss": 0.573,
+      "step": 24738
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.092112192348791e-06,
+      "loss": 0.6513,
+      "step": 24739
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.091412459246995e-06,
+      "loss": 0.5705,
+      "step": 24740
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.090712735843288e-06,
+      "loss": 0.6083,
+      "step": 24741
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.090013022141232e-06,
+      "loss": 0.564,
+      "step": 24742
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.089313318144377e-06,
+      "loss": 0.5527,
+      "step": 24743
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.088613623856282e-06,
+      "loss": 0.602,
+      "step": 24744
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.0879139392805e-06,
+      "loss": 0.5985,
+      "step": 24745
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.087214264420587e-06,
+      "loss": 0.6477,
+      "step": 24746
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.086514599280099e-06,
+      "loss": 0.6273,
+      "step": 24747
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.085814943862589e-06,
+      "loss": 0.5606,
+      "step": 24748
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.085115298171616e-06,
+      "loss": 0.6458,
+      "step": 24749
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.084415662210735e-06,
+      "loss": 0.613,
+      "step": 24750
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.083716035983495e-06,
+      "loss": 0.5541,
+      "step": 24751
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.083016419493456e-06,
+      "loss": 0.534,
+      "step": 24752
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.082316812744172e-06,
+      "loss": 0.5826,
+      "step": 24753
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.081617215739197e-06,
+      "loss": 0.6042,
+      "step": 24754
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.080917628482093e-06,
+      "loss": 0.6368,
+      "step": 24755
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.080218050976401e-06,
+      "loss": 0.5287,
+      "step": 24756
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.079518483225685e-06,
+      "loss": 0.6052,
+      "step": 24757
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.0788189252335e-06,
+      "loss": 0.6076,
+      "step": 24758
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.078119377003398e-06,
+      "loss": 0.5872,
+      "step": 24759
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.077419838538935e-06,
+      "loss": 0.6091,
+      "step": 24760
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.076720309843663e-06,
+      "loss": 0.6137,
+      "step": 24761
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.076020790921138e-06,
+      "loss": 0.5714,
+      "step": 24762
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.075321281774918e-06,
+      "loss": 0.6069,
+      "step": 24763
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.074621782408551e-06,
+      "loss": 0.6022,
+      "step": 24764
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.073922292825596e-06,
+      "loss": 0.5955,
+      "step": 24765
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.073222813029609e-06,
+      "loss": 0.589,
+      "step": 24766
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.072523343024136e-06,
+      "loss": 0.5754,
+      "step": 24767
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.071823882812738e-06,
+      "loss": 0.6018,
+      "step": 24768
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.071124432398966e-06,
+      "loss": 0.5775,
+      "step": 24769
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.070424991786377e-06,
+      "loss": 0.6221,
+      "step": 24770
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.069725560978528e-06,
+      "loss": 0.6212,
+      "step": 24771
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.069026139978965e-06,
+      "loss": 0.5486,
+      "step": 24772
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.068326728791246e-06,
+      "loss": 0.6293,
+      "step": 24773
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.067627327418922e-06,
+      "loss": 0.5732,
+      "step": 24774
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.06692793586555e-06,
+      "loss": 0.6247,
+      "step": 24775
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.066228554134686e-06,
+      "loss": 0.5911,
+      "step": 24776
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.065529182229883e-06,
+      "loss": 0.6018,
+      "step": 24777
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.064829820154688e-06,
+      "loss": 0.6126,
+      "step": 24778
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.06413046791266e-06,
+      "loss": 0.623,
+      "step": 24779
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.063431125507355e-06,
+      "loss": 0.6193,
+      "step": 24780
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.062731792942322e-06,
+      "loss": 0.6589,
+      "step": 24781
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.062032470221117e-06,
+      "loss": 0.5755,
+      "step": 24782
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.061333157347294e-06,
+      "loss": 0.6297,
+      "step": 24783
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.0606338543244e-06,
+      "loss": 0.5891,
+      "step": 24784
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.059934561155996e-06,
+      "loss": 0.6152,
+      "step": 24785
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.059235277845631e-06,
+      "loss": 0.5656,
+      "step": 24786
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.058536004396866e-06,
+      "loss": 0.5865,
+      "step": 24787
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.057836740813243e-06,
+      "loss": 0.6069,
+      "step": 24788
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.05713748709832e-06,
+      "loss": 0.636,
+      "step": 24789
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.05643824325565e-06,
+      "loss": 0.6667,
+      "step": 24790
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.055739009288788e-06,
+      "loss": 0.5764,
+      "step": 24791
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.055039785201284e-06,
+      "loss": 0.6026,
+      "step": 24792
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.054340570996697e-06,
+      "loss": 0.6105,
+      "step": 24793
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.053641366678571e-06,
+      "loss": 0.5628,
+      "step": 24794
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.052942172250464e-06,
+      "loss": 0.6203,
+      "step": 24795
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.052242987715928e-06,
+      "loss": 0.6304,
+      "step": 24796
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.051543813078515e-06,
+      "loss": 0.6093,
+      "step": 24797
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.050844648341784e-06,
+      "loss": 0.5623,
+      "step": 24798
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.050145493509277e-06,
+      "loss": 0.6334,
+      "step": 24799
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.049446348584551e-06,
+      "loss": 0.5875,
+      "step": 24800
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.048747213571163e-06,
+      "loss": 0.6028,
+      "step": 24801
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.04804808847266e-06,
+      "loss": 0.587,
+      "step": 24802
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.047348973292595e-06,
+      "loss": 0.5677,
+      "step": 24803
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.046649868034527e-06,
+      "loss": 0.5586,
+      "step": 24804
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.045950772701998e-06,
+      "loss": 0.666,
+      "step": 24805
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.045251687298566e-06,
+      "loss": 0.6447,
+      "step": 24806
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.044552611827781e-06,
+      "loss": 0.5793,
+      "step": 24807
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.043853546293199e-06,
+      "loss": 0.6419,
+      "step": 24808
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.043154490698374e-06,
+      "loss": 0.6318,
+      "step": 24809
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.042455445046848e-06,
+      "loss": 0.5963,
+      "step": 24810
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.041756409342181e-06,
+      "loss": 0.5494,
+      "step": 24811
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.041057383587922e-06,
+      "loss": 0.5937,
+      "step": 24812
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.040358367787627e-06,
+      "loss": 0.6153,
+      "step": 24813
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.039659361944842e-06,
+      "loss": 0.6239,
+      "step": 24814
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.038960366063126e-06,
+      "loss": 0.5787,
+      "step": 24815
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.038261380146024e-06,
+      "loss": 0.6027,
+      "step": 24816
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.03756240419709e-06,
+      "loss": 0.6002,
+      "step": 24817
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.036863438219876e-06,
+      "loss": 0.6507,
+      "step": 24818
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.036164482217936e-06,
+      "loss": 0.703,
+      "step": 24819
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.03546553619482e-06,
+      "loss": 0.5909,
+      "step": 24820
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.034766600154077e-06,
+      "loss": 0.5959,
+      "step": 24821
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.034067674099261e-06,
+      "loss": 0.6201,
+      "step": 24822
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.03336875803392e-06,
+      "loss": 0.6059,
+      "step": 24823
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.032669851961612e-06,
+      "loss": 0.6164,
+      "step": 24824
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.031970955885886e-06,
+      "loss": 0.5925,
+      "step": 24825
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.03127206981029e-06,
+      "loss": 0.5171,
+      "step": 24826
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.030573193738375e-06,
+      "loss": 0.6015,
+      "step": 24827
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.029874327673694e-06,
+      "loss": 0.6073,
+      "step": 24828
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.0291754716198e-06,
+      "loss": 0.7161,
+      "step": 24829
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.028476625580241e-06,
+      "loss": 0.5635,
+      "step": 24830
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.027777789558577e-06,
+      "loss": 0.6025,
+      "step": 24831
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.027078963558344e-06,
+      "loss": 0.6422,
+      "step": 24832
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.026380147583103e-06,
+      "loss": 0.5531,
+      "step": 24833
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.0256813416364e-06,
+      "loss": 0.5977,
+      "step": 24834
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.02498254572179e-06,
+      "loss": 0.6193,
+      "step": 24835
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.024283759842827e-06,
+      "loss": 0.583,
+      "step": 24836
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.023584984003052e-06,
+      "loss": 0.5764,
+      "step": 24837
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.02288621820602e-06,
+      "loss": 0.6271,
+      "step": 24838
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.022187462455285e-06,
+      "loss": 0.5486,
+      "step": 24839
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.02148871675439e-06,
+      "loss": 0.5837,
+      "step": 24840
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.020789981106893e-06,
+      "loss": 0.6452,
+      "step": 24841
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.020091255516345e-06,
+      "loss": 0.6088,
+      "step": 24842
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.01939253998629e-06,
+      "loss": 0.5957,
+      "step": 24843
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.018693834520281e-06,
+      "loss": 0.6365,
+      "step": 24844
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.01799513912187e-06,
+      "loss": 0.5851,
+      "step": 24845
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.017296453794604e-06,
+      "loss": 0.5926,
+      "step": 24846
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.016597778542042e-06,
+      "loss": 0.5717,
+      "step": 24847
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.01589911336772e-06,
+      "loss": 0.5822,
+      "step": 24848
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.015200458275198e-06,
+      "loss": 0.592,
+      "step": 24849
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.014501813268022e-06,
+      "loss": 0.6077,
+      "step": 24850
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.013803178349746e-06,
+      "loss": 0.5767,
+      "step": 24851
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.013104553523922e-06,
+      "loss": 0.6159,
+      "step": 24852
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.01240593879409e-06,
+      "loss": 0.6006,
+      "step": 24853
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.011707334163806e-06,
+      "loss": 0.5766,
+      "step": 24854
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.011008739636619e-06,
+      "loss": 0.5564,
+      "step": 24855
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.01031015521608e-06,
+      "loss": 0.599,
+      "step": 24856
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.009611580905738e-06,
+      "loss": 0.6297,
+      "step": 24857
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.008913016709144e-06,
+      "loss": 0.5564,
+      "step": 24858
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.008214462629844e-06,
+      "loss": 0.5418,
+      "step": 24859
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.00751591867139e-06,
+      "loss": 0.6006,
+      "step": 24860
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.006817384837332e-06,
+      "loss": 0.6587,
+      "step": 24861
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.006118861131218e-06,
+      "loss": 0.6431,
+      "step": 24862
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.005420347556598e-06,
+      "loss": 0.5652,
+      "step": 24863
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.004721844117023e-06,
+      "loss": 0.5654,
+      "step": 24864
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.004023350816038e-06,
+      "loss": 0.6211,
+      "step": 24865
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.003324867657196e-06,
+      "loss": 0.5306,
+      "step": 24866
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.002626394644042e-06,
+      "loss": 0.62,
+      "step": 24867
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.001927931780131e-06,
+      "loss": 0.6195,
+      "step": 24868
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.001229479069014e-06,
+      "loss": 0.6326,
+      "step": 24869
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.000531036514229e-06,
+      "loss": 0.6263,
+      "step": 24870
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.999832604119333e-06,
+      "loss": 0.5642,
+      "step": 24871
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.999134181887873e-06,
+      "loss": 0.5904,
+      "step": 24872
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.998435769823399e-06,
+      "loss": 0.6272,
+      "step": 24873
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.997737367929461e-06,
+      "loss": 0.6065,
+      "step": 24874
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.997038976209604e-06,
+      "loss": 0.6354,
+      "step": 24875
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.996340594667376e-06,
+      "loss": 0.6495,
+      "step": 24876
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.99564222330633e-06,
+      "loss": 0.5993,
+      "step": 24877
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.994943862130015e-06,
+      "loss": 0.6277,
+      "step": 24878
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.994245511141976e-06,
+      "loss": 0.6393,
+      "step": 24879
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.993547170345764e-06,
+      "loss": 0.5776,
+      "step": 24880
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.992848839744925e-06,
+      "loss": 0.6036,
+      "step": 24881
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.992150519343008e-06,
+      "loss": 0.5633,
+      "step": 24882
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.991452209143563e-06,
+      "loss": 0.6534,
+      "step": 24883
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.990753909150137e-06,
+      "loss": 0.5687,
+      "step": 24884
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.990055619366283e-06,
+      "loss": 0.5681,
+      "step": 24885
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.98935733979554e-06,
+      "loss": 0.608,
+      "step": 24886
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.988659070441464e-06,
+      "loss": 0.6448,
+      "step": 24887
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.987960811307597e-06,
+      "loss": 0.5879,
+      "step": 24888
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.987262562397491e-06,
+      "loss": 0.6681,
+      "step": 24889
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.986564323714699e-06,
+      "loss": 0.611,
+      "step": 24890
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.98586609526276e-06,
+      "loss": 0.5596,
+      "step": 24891
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.985167877045223e-06,
+      "loss": 0.5843,
+      "step": 24892
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.98446966906564e-06,
+      "loss": 0.6087,
+      "step": 24893
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.983771471327555e-06,
+      "loss": 0.6366,
+      "step": 24894
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.98307328383452e-06,
+      "loss": 0.5768,
+      "step": 24895
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.98237510659008e-06,
+      "loss": 0.6345,
+      "step": 24896
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.981676939597783e-06,
+      "loss": 0.5505,
+      "step": 24897
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.980978782861174e-06,
+      "loss": 0.5821,
+      "step": 24898
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.980280636383808e-06,
+      "loss": 0.5792,
+      "step": 24899
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.979582500169223e-06,
+      "loss": 0.5827,
+      "step": 24900
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.978884374220974e-06,
+      "loss": 0.5806,
+      "step": 24901
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.978186258542605e-06,
+      "loss": 0.6225,
+      "step": 24902
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.977488153137663e-06,
+      "loss": 0.6064,
+      "step": 24903
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.976790058009694e-06,
+      "loss": 0.649,
+      "step": 24904
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.976091973162248e-06,
+      "loss": 0.5661,
+      "step": 24905
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.975393898598872e-06,
+      "loss": 0.6483,
+      "step": 24906
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.974695834323116e-06,
+      "loss": 0.6502,
+      "step": 24907
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.97399778033852e-06,
+      "loss": 0.6059,
+      "step": 24908
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.973299736648636e-06,
+      "loss": 0.5959,
+      "step": 24909
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.972601703257007e-06,
+      "loss": 0.6059,
+      "step": 24910
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.971903680167183e-06,
+      "loss": 0.577,
+      "step": 24911
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.971205667382715e-06,
+      "loss": 0.6425,
+      "step": 24912
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.970507664907142e-06,
+      "loss": 0.6716,
+      "step": 24913
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.969809672744013e-06,
+      "loss": 0.6987,
+      "step": 24914
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.969111690896875e-06,
+      "loss": 0.6162,
+      "step": 24915
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.968413719369279e-06,
+      "loss": 0.6253,
+      "step": 24916
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.967715758164767e-06,
+      "loss": 0.5601,
+      "step": 24917
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.967017807286883e-06,
+      "loss": 0.5645,
+      "step": 24918
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.966319866739181e-06,
+      "loss": 0.5574,
+      "step": 24919
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.965621936525201e-06,
+      "loss": 0.5945,
+      "step": 24920
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.96492401664849e-06,
+      "loss": 0.5938,
+      "step": 24921
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.964226107112598e-06,
+      "loss": 0.6501,
+      "step": 24922
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.963528207921072e-06,
+      "loss": 0.6525,
+      "step": 24923
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.962830319077453e-06,
+      "loss": 0.6021,
+      "step": 24924
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.962132440585289e-06,
+      "loss": 0.5565,
+      "step": 24925
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.961434572448126e-06,
+      "loss": 0.6352,
+      "step": 24926
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.960736714669512e-06,
+      "loss": 0.5839,
+      "step": 24927
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.960038867252997e-06,
+      "loss": 0.6199,
+      "step": 24928
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.959341030202116e-06,
+      "loss": 0.5784,
+      "step": 24929
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.958643203520423e-06,
+      "loss": 0.5966,
+      "step": 24930
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.95794538721146e-06,
+      "loss": 0.6184,
+      "step": 24931
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.957247581278773e-06,
+      "loss": 0.5987,
+      "step": 24932
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.95654978572591e-06,
+      "loss": 0.5976,
+      "step": 24933
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.955852000556423e-06,
+      "loss": 0.6511,
+      "step": 24934
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.955154225773844e-06,
+      "loss": 0.5886,
+      "step": 24935
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.954456461381726e-06,
+      "loss": 0.604,
+      "step": 24936
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.953758707383616e-06,
+      "loss": 0.6084,
+      "step": 24937
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.953060963783057e-06,
+      "loss": 0.6352,
+      "step": 24938
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.952363230583592e-06,
+      "loss": 0.5755,
+      "step": 24939
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.951665507788772e-06,
+      "loss": 0.6202,
+      "step": 24940
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.950967795402137e-06,
+      "loss": 0.5986,
+      "step": 24941
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.950270093427236e-06,
+      "loss": 0.6089,
+      "step": 24942
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.949572401867612e-06,
+      "loss": 0.6192,
+      "step": 24943
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.948874720726816e-06,
+      "loss": 0.6272,
+      "step": 24944
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.948177050008383e-06,
+      "loss": 0.6459,
+      "step": 24945
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.947479389715863e-06,
+      "loss": 0.5932,
+      "step": 24946
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.946781739852803e-06,
+      "loss": 0.5947,
+      "step": 24947
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.946084100422745e-06,
+      "loss": 0.5908,
+      "step": 24948
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.945386471429235e-06,
+      "loss": 0.6037,
+      "step": 24949
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.944688852875824e-06,
+      "loss": 0.6037,
+      "step": 24950
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.943991244766044e-06,
+      "loss": 0.5575,
+      "step": 24951
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.94329364710345e-06,
+      "loss": 0.6446,
+      "step": 24952
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.94259605989158e-06,
+      "loss": 0.6389,
+      "step": 24953
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.941898483133984e-06,
+      "loss": 0.5929,
+      "step": 24954
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.941200916834208e-06,
+      "loss": 0.6047,
+      "step": 24955
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.940503360995789e-06,
+      "loss": 0.5886,
+      "step": 24956
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.939805815622274e-06,
+      "loss": 0.6221,
+      "step": 24957
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.939108280717214e-06,
+      "loss": 0.6005,
+      "step": 24958
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.938410756284144e-06,
+      "loss": 0.5823,
+      "step": 24959
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.937713242326615e-06,
+      "loss": 0.6028,
+      "step": 24960
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.937015738848171e-06,
+      "loss": 0.596,
+      "step": 24961
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.93631824585235e-06,
+      "loss": 0.5876,
+      "step": 24962
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.935620763342701e-06,
+      "loss": 0.6051,
+      "step": 24963
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.934923291322768e-06,
+      "loss": 0.6223,
+      "step": 24964
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.934225829796094e-06,
+      "loss": 0.5938,
+      "step": 24965
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.933528378766228e-06,
+      "loss": 0.6161,
+      "step": 24966
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.932830938236707e-06,
+      "loss": 0.6742,
+      "step": 24967
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.932133508211075e-06,
+      "loss": 0.5607,
+      "step": 24968
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.93143608869288e-06,
+      "loss": 0.591,
+      "step": 24969
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.930738679685662e-06,
+      "loss": 0.5558,
+      "step": 24970
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.93004128119297e-06,
+      "loss": 0.5955,
+      "step": 24971
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.929343893218343e-06,
+      "loss": 0.6071,
+      "step": 24972
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.928646515765326e-06,
+      "loss": 0.5637,
+      "step": 24973
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.927949148837461e-06,
+      "loss": 0.641,
+      "step": 24974
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.927251792438296e-06,
+      "loss": 0.5656,
+      "step": 24975
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.926554446571371e-06,
+      "loss": 0.5737,
+      "step": 24976
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.925857111240231e-06,
+      "loss": 0.5819,
+      "step": 24977
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.925159786448415e-06,
+      "loss": 0.5934,
+      "step": 24978
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.924462472199475e-06,
+      "loss": 0.5319,
+      "step": 24979
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.923765168496943e-06,
+      "loss": 0.6132,
+      "step": 24980
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.923067875344373e-06,
+      "loss": 0.6517,
+      "step": 24981
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.922370592745303e-06,
+      "loss": 0.5536,
+      "step": 24982
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.921673320703274e-06,
+      "loss": 0.563,
+      "step": 24983
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.920976059221833e-06,
+      "loss": 0.6287,
+      "step": 24984
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.92027880830452e-06,
+      "loss": 0.5997,
+      "step": 24985
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.919581567954879e-06,
+      "loss": 0.6072,
+      "step": 24986
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.918884338176455e-06,
+      "loss": 0.5827,
+      "step": 24987
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.918187118972792e-06,
+      "loss": 0.601,
+      "step": 24988
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.917489910347427e-06,
+      "loss": 0.6142,
+      "step": 24989
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.916792712303905e-06,
+      "loss": 0.5537,
+      "step": 24990
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.916095524845768e-06,
+      "loss": 0.6089,
+      "step": 24991
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.915398347976561e-06,
+      "loss": 0.6961,
+      "step": 24992
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.914701181699831e-06,
+      "loss": 0.6357,
+      "step": 24993
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.914004026019109e-06,
+      "loss": 0.6573,
+      "step": 24994
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.913306880937944e-06,
+      "loss": 0.6019,
+      "step": 24995
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.912609746459881e-06,
+      "loss": 0.5949,
+      "step": 24996
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.911912622588457e-06,
+      "loss": 0.5449,
+      "step": 24997
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.911215509327214e-06,
+      "loss": 0.6582,
+      "step": 24998
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.9105184066797e-06,
+      "loss": 0.6057,
+      "step": 24999
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.909821314649456e-06,
+      "loss": 0.6105,
+      "step": 25000
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.909124233240018e-06,
+      "loss": 0.5623,
+      "step": 25001
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.908427162454932e-06,
+      "loss": 0.583,
+      "step": 25002
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.907730102297739e-06,
+      "loss": 0.6075,
+      "step": 25003
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.907033052771988e-06,
+      "loss": 0.6207,
+      "step": 25004
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.906336013881213e-06,
+      "loss": 0.6168,
+      "step": 25005
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.905638985628954e-06,
+      "loss": 0.6365,
+      "step": 25006
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.904941968018758e-06,
+      "loss": 0.5586,
+      "step": 25007
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.904244961054166e-06,
+      "loss": 0.566,
+      "step": 25008
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.903547964738723e-06,
+      "loss": 0.5717,
+      "step": 25009
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.902850979075961e-06,
+      "loss": 0.5225,
+      "step": 25010
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.90215400406943e-06,
+      "loss": 0.5942,
+      "step": 25011
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.901457039722668e-06,
+      "loss": 0.6222,
+      "step": 25012
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.900760086039216e-06,
+      "loss": 0.6194,
+      "step": 25013
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.90006314302262e-06,
+      "loss": 0.6771,
+      "step": 25014
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89936621067642e-06,
+      "loss": 0.6528,
+      "step": 25015
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89866928900415e-06,
+      "loss": 0.6031,
+      "step": 25016
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89797237800936e-06,
+      "loss": 0.5512,
+      "step": 25017
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.897275477695587e-06,
+      "loss": 0.6436,
+      "step": 25018
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.896578588066372e-06,
+      "loss": 0.6204,
+      "step": 25019
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.895881709125263e-06,
+      "loss": 0.6591,
+      "step": 25020
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89518484087579e-06,
+      "loss": 0.6103,
+      "step": 25021
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.894487983321499e-06,
+      "loss": 0.6042,
+      "step": 25022
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89379113646593e-06,
+      "loss": 0.6062,
+      "step": 25023
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.893094300312627e-06,
+      "loss": 0.6347,
+      "step": 25024
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.892397474865128e-06,
+      "loss": 0.5647,
+      "step": 25025
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.89170066012698e-06,
+      "loss": 0.58,
+      "step": 25026
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.891003856101712e-06,
+      "loss": 0.6019,
+      "step": 25027
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.890307062792872e-06,
+      "loss": 0.5874,
+      "step": 25028
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.889610280204e-06,
+      "loss": 0.6576,
+      "step": 25029
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.888913508338636e-06,
+      "loss": 0.6298,
+      "step": 25030
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.888216747200325e-06,
+      "loss": 0.5792,
+      "step": 25031
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.887519996792601e-06,
+      "loss": 0.6021,
+      "step": 25032
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.886823257119004e-06,
+      "loss": 0.6026,
+      "step": 25033
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.886126528183078e-06,
+      "loss": 0.5968,
+      "step": 25034
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.885429809988363e-06,
+      "loss": 0.6268,
+      "step": 25035
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.8847331025384e-06,
+      "loss": 0.5897,
+      "step": 25036
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.884036405836725e-06,
+      "loss": 0.6027,
+      "step": 25037
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.883339719886883e-06,
+      "loss": 0.5882,
+      "step": 25038
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.882643044692408e-06,
+      "loss": 0.5433,
+      "step": 25039
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.881946380256845e-06,
+      "loss": 0.6275,
+      "step": 25040
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.881249726583733e-06,
+      "loss": 0.6657,
+      "step": 25041
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.880553083676616e-06,
+      "loss": 0.6234,
+      "step": 25042
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.879856451539026e-06,
+      "loss": 0.5947,
+      "step": 25043
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.879159830174505e-06,
+      "loss": 0.5229,
+      "step": 25044
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.878463219586593e-06,
+      "loss": 0.5817,
+      "step": 25045
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.877766619778832e-06,
+      "loss": 0.5905,
+      "step": 25046
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.877070030754766e-06,
+      "loss": 0.6206,
+      "step": 25047
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.876373452517922e-06,
+      "loss": 0.5887,
+      "step": 25048
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.875676885071848e-06,
+      "loss": 0.6037,
+      "step": 25049
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.874980328420082e-06,
+      "loss": 0.6533,
+      "step": 25050
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.874283782566161e-06,
+      "loss": 0.5551,
+      "step": 25051
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.873587247513632e-06,
+      "loss": 0.6072,
+      "step": 25052
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.872890723266028e-06,
+      "loss": 0.5747,
+      "step": 25053
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.872194209826886e-06,
+      "loss": 0.5766,
+      "step": 25054
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.871497707199748e-06,
+      "loss": 0.6009,
+      "step": 25055
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.870801215388156e-06,
+      "loss": 0.5167,
+      "step": 25056
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.870104734395646e-06,
+      "loss": 0.6027,
+      "step": 25057
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.869408264225758e-06,
+      "loss": 0.7,
+      "step": 25058
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.868711804882033e-06,
+      "loss": 0.6245,
+      "step": 25059
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.868015356368002e-06,
+      "loss": 0.6089,
+      "step": 25060
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.86731891868721e-06,
+      "loss": 0.5196,
+      "step": 25061
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.866622491843198e-06,
+      "loss": 0.5824,
+      "step": 25062
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.865926075839505e-06,
+      "loss": 0.6033,
+      "step": 25063
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.865229670679661e-06,
+      "loss": 0.6545,
+      "step": 25064
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.86453327636721e-06,
+      "loss": 0.5778,
+      "step": 25065
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.863836892905693e-06,
+      "loss": 0.6004,
+      "step": 25066
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.863140520298645e-06,
+      "loss": 0.5903,
+      "step": 25067
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.862444158549606e-06,
+      "loss": 0.5893,
+      "step": 25068
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.861747807662117e-06,
+      "loss": 0.5696,
+      "step": 25069
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.861051467639712e-06,
+      "loss": 0.585,
+      "step": 25070
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.86035513848593e-06,
+      "loss": 0.6336,
+      "step": 25071
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.85965882020431e-06,
+      "loss": 0.5812,
+      "step": 25072
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.858962512798391e-06,
+      "loss": 0.6068,
+      "step": 25073
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.858266216271712e-06,
+      "loss": 0.617,
+      "step": 25074
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.857569930627809e-06,
+      "loss": 0.5646,
+      "step": 25075
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.85687365587022e-06,
+      "loss": 0.5308,
+      "step": 25076
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.856177392002484e-06,
+      "loss": 0.6827,
+      "step": 25077
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.855481139028137e-06,
+      "loss": 0.5754,
+      "step": 25078
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.854784896950717e-06,
+      "loss": 0.598,
+      "step": 25079
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.854088665773769e-06,
+      "loss": 0.6095,
+      "step": 25080
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.85339244550082e-06,
+      "loss": 0.6205,
+      "step": 25081
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.852696236135414e-06,
+      "loss": 0.542,
+      "step": 25082
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.852000037681086e-06,
+      "loss": 0.6233,
+      "step": 25083
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.851303850141376e-06,
+      "loss": 0.6213,
+      "step": 25084
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.850607673519825e-06,
+      "loss": 0.6218,
+      "step": 25085
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.849911507819961e-06,
+      "loss": 0.6024,
+      "step": 25086
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.849215353045325e-06,
+      "loss": 0.6225,
+      "step": 25087
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.848519209199457e-06,
+      "loss": 0.5699,
+      "step": 25088
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.847823076285892e-06,
+      "loss": 0.5975,
+      "step": 25089
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.84712695430817e-06,
+      "loss": 0.6263,
+      "step": 25090
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.84643084326983e-06,
+      "loss": 0.601,
+      "step": 25091
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.8457347431744e-06,
+      "loss": 0.6694,
+      "step": 25092
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.845038654025424e-06,
+      "loss": 0.6065,
+      "step": 25093
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.844342575826439e-06,
+      "loss": 0.5126,
+      "step": 25094
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.84364650858098e-06,
+      "loss": 0.6214,
+      "step": 25095
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.842950452292584e-06,
+      "loss": 0.6437,
+      "step": 25096
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.842254406964791e-06,
+      "loss": 0.5442,
+      "step": 25097
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.841558372601133e-06,
+      "loss": 0.5821,
+      "step": 25098
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.840862349205149e-06,
+      "loss": 0.6337,
+      "step": 25099
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.840166336780375e-06,
+      "loss": 0.6394,
+      "step": 25100
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.839470335330352e-06,
+      "loss": 0.6707,
+      "step": 25101
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.83877434485861e-06,
+      "loss": 0.6084,
+      "step": 25102
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.838078365368687e-06,
+      "loss": 0.5484,
+      "step": 25103
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.837382396864122e-06,
+      "loss": 0.6205,
+      "step": 25104
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.83668643934845e-06,
+      "loss": 0.5759,
+      "step": 25105
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.835990492825209e-06,
+      "loss": 0.5807,
+      "step": 25106
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.835294557297938e-06,
+      "loss": 0.6229,
+      "step": 25107
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.834598632770165e-06,
+      "loss": 0.6455,
+      "step": 25108
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.83390271924543e-06,
+      "loss": 0.5618,
+      "step": 25109
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.833206816727268e-06,
+      "loss": 0.5747,
+      "step": 25110
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.83251092521922e-06,
+      "loss": 0.6169,
+      "step": 25111
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.831815044724821e-06,
+      "loss": 0.6069,
+      "step": 25112
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.8311191752476e-06,
+      "loss": 0.613,
+      "step": 25113
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.8304233167911e-06,
+      "loss": 0.6484,
+      "step": 25114
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.829727469358855e-06,
+      "loss": 0.6455,
+      "step": 25115
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.829031632954398e-06,
+      "loss": 0.5794,
+      "step": 25116
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.828335807581268e-06,
+      "loss": 0.6278,
+      "step": 25117
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.827639993243004e-06,
+      "loss": 0.593,
+      "step": 25118
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.826944189943132e-06,
+      "loss": 0.5881,
+      "step": 25119
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.826248397685194e-06,
+      "loss": 0.6293,
+      "step": 25120
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.825552616472724e-06,
+      "loss": 0.6468,
+      "step": 25121
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.824856846309259e-06,
+      "loss": 0.6119,
+      "step": 25122
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.824161087198336e-06,
+      "loss": 0.6108,
+      "step": 25123
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.823465339143484e-06,
+      "loss": 0.5406,
+      "step": 25124
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.822769602148242e-06,
+      "loss": 0.5387,
+      "step": 25125
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.822073876216146e-06,
+      "loss": 0.6296,
+      "step": 25126
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.82137816135073e-06,
+      "loss": 0.5597,
+      "step": 25127
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.820682457555535e-06,
+      "loss": 0.5278,
+      "step": 25128
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.819986764834085e-06,
+      "loss": 0.6028,
+      "step": 25129
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.81929108318992e-06,
+      "loss": 0.5895,
+      "step": 25130
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.818595412626577e-06,
+      "loss": 0.6092,
+      "step": 25131
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.817899753147593e-06,
+      "loss": 0.5855,
+      "step": 25132
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.817204104756494e-06,
+      "loss": 0.6614,
+      "step": 25133
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.816508467456825e-06,
+      "loss": 0.6539,
+      "step": 25134
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.815812841252114e-06,
+      "loss": 0.5535,
+      "step": 25135
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.815117226145897e-06,
+      "loss": 0.6247,
+      "step": 25136
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.814421622141709e-06,
+      "loss": 0.646,
+      "step": 25137
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.813726029243085e-06,
+      "loss": 0.6105,
+      "step": 25138
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.813030447453562e-06,
+      "loss": 0.6316,
+      "step": 25139
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.812334876776668e-06,
+      "loss": 0.5864,
+      "step": 25140
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.811639317215943e-06,
+      "loss": 0.6013,
+      "step": 25141
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.810943768774919e-06,
+      "loss": 0.6061,
+      "step": 25142
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.81024823145713e-06,
+      "loss": 0.6079,
+      "step": 25143
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.80955270526611e-06,
+      "loss": 0.5566,
+      "step": 25144
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.8088571902054e-06,
+      "loss": 0.5667,
+      "step": 25145
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.808161686278524e-06,
+      "loss": 0.6268,
+      "step": 25146
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.80746619348902e-06,
+      "loss": 0.5571,
+      "step": 25147
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.806770711840422e-06,
+      "loss": 0.5843,
+      "step": 25148
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.806075241336264e-06,
+      "loss": 0.6211,
+      "step": 25149
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.805379781980086e-06,
+      "loss": 0.6488,
+      "step": 25150
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.804684333775411e-06,
+      "loss": 0.5954,
+      "step": 25151
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.803988896725779e-06,
+      "loss": 0.5972,
+      "step": 25152
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.803293470834722e-06,
+      "loss": 0.5799,
+      "step": 25153
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.802598056105773e-06,
+      "loss": 0.5734,
+      "step": 25154
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.801902652542471e-06,
+      "loss": 0.6102,
+      "step": 25155
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.801207260148343e-06,
+      "loss": 0.597,
+      "step": 25156
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.800511878926924e-06,
+      "loss": 0.6084,
+      "step": 25157
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.799816508881748e-06,
+      "loss": 0.607,
+      "step": 25158
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.799121150016349e-06,
+      "loss": 0.5637,
+      "step": 25159
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.79842580233426e-06,
+      "loss": 0.5533,
+      "step": 25160
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.797730465839017e-06,
+      "loss": 0.6447,
+      "step": 25161
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.797035140534147e-06,
+      "loss": 0.6661,
+      "step": 25162
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.796339826423187e-06,
+      "loss": 0.6063,
+      "step": 25163
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.79564452350967e-06,
+      "loss": 0.6366,
+      "step": 25164
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.794949231797128e-06,
+      "loss": 0.5823,
+      "step": 25165
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.7942539512891e-06,
+      "loss": 0.6242,
+      "step": 25166
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.793558681989109e-06,
+      "loss": 0.5573,
+      "step": 25167
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.792863423900691e-06,
+      "loss": 0.5822,
+      "step": 25168
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.792168177027383e-06,
+      "loss": 0.6187,
+      "step": 25169
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.791472941372714e-06,
+      "loss": 0.5828,
+      "step": 25170
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.79077771694022e-06,
+      "loss": 0.6127,
+      "step": 25171
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.79008250373343e-06,
+      "loss": 0.5916,
+      "step": 25172
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.789387301755877e-06,
+      "loss": 0.6004,
+      "step": 25173
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.788692111011096e-06,
+      "loss": 0.6316,
+      "step": 25174
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.787996931502616e-06,
+      "loss": 0.6389,
+      "step": 25175
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.787301763233974e-06,
+      "loss": 0.5879,
+      "step": 25176
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.786606606208703e-06,
+      "loss": 0.6024,
+      "step": 25177
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.785911460430325e-06,
+      "loss": 0.6307,
+      "step": 25178
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.785216325902381e-06,
+      "loss": 0.6332,
+      "step": 25179
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.784521202628401e-06,
+      "loss": 0.6072,
+      "step": 25180
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.78382609061192e-06,
+      "loss": 0.5943,
+      "step": 25181
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.783130989856465e-06,
+      "loss": 0.6181,
+      "step": 25182
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.782435900365575e-06,
+      "loss": 0.5626,
+      "step": 25183
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.781740822142774e-06,
+      "loss": 0.6441,
+      "step": 25184
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.781045755191598e-06,
+      "loss": 0.625,
+      "step": 25185
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.780350699515577e-06,
+      "loss": 0.5384,
+      "step": 25186
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.779655655118246e-06,
+      "loss": 0.5925,
+      "step": 25187
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.778960622003137e-06,
+      "loss": 0.6207,
+      "step": 25188
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.778265600173778e-06,
+      "loss": 0.5782,
+      "step": 25189
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.7775705896337e-06,
+      "loss": 0.5911,
+      "step": 25190
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.776875590386438e-06,
+      "loss": 0.5493,
+      "step": 25191
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.776180602435523e-06,
+      "loss": 0.6288,
+      "step": 25192
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.775485625784487e-06,
+      "loss": 0.5703,
+      "step": 25193
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.774790660436857e-06,
+      "loss": 0.6262,
+      "step": 25194
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.77409570639617e-06,
+      "loss": 0.6143,
+      "step": 25195
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.773400763665953e-06,
+      "loss": 0.6335,
+      "step": 25196
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.772705832249737e-06,
+      "loss": 0.6104,
+      "step": 25197
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.772010912151055e-06,
+      "loss": 0.5737,
+      "step": 25198
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.771316003373444e-06,
+      "loss": 0.5643,
+      "step": 25199
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.770621105920424e-06,
+      "loss": 0.6193,
+      "step": 25200
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.769926219795532e-06,
+      "loss": 0.5776,
+      "step": 25201
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.769231345002298e-06,
+      "loss": 0.5455,
+      "step": 25202
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.768536481544252e-06,
+      "loss": 0.5999,
+      "step": 25203
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.767841629424927e-06,
+      "loss": 0.5866,
+      "step": 25204
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.767146788647852e-06,
+      "loss": 0.6062,
+      "step": 25205
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.766451959216558e-06,
+      "loss": 0.572,
+      "step": 25206
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.765757141134575e-06,
+      "loss": 0.6033,
+      "step": 25207
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.765062334405434e-06,
+      "loss": 0.6125,
+      "step": 25208
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.764367539032668e-06,
+      "loss": 0.574,
+      "step": 25209
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.763672755019806e-06,
+      "loss": 0.6014,
+      "step": 25210
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.762977982370374e-06,
+      "loss": 0.5551,
+      "step": 25211
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.76228322108791e-06,
+      "loss": 0.642,
+      "step": 25212
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.761588471175935e-06,
+      "loss": 0.5957,
+      "step": 25213
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.760893732637988e-06,
+      "loss": 0.5595,
+      "step": 25214
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.760199005477597e-06,
+      "loss": 0.5923,
+      "step": 25215
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.75950428969829e-06,
+      "loss": 0.6026,
+      "step": 25216
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.758809585303596e-06,
+      "loss": 0.5855,
+      "step": 25217
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.758114892297047e-06,
+      "loss": 0.6103,
+      "step": 25218
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.757420210682172e-06,
+      "loss": 0.6474,
+      "step": 25219
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.756725540462508e-06,
+      "loss": 0.6227,
+      "step": 25220
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.756030881641572e-06,
+      "loss": 0.5504,
+      "step": 25221
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.755336234222901e-06,
+      "loss": 0.5617,
+      "step": 25222
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.754641598210025e-06,
+      "loss": 0.5945,
+      "step": 25223
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.753946973606473e-06,
+      "loss": 0.5944,
+      "step": 25224
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.753252360415771e-06,
+      "loss": 0.6136,
+      "step": 25225
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.75255775864146e-06,
+      "loss": 0.5939,
+      "step": 25226
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.751863168287054e-06,
+      "loss": 0.6031,
+      "step": 25227
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.75116858935609e-06,
+      "loss": 0.7224,
+      "step": 25228
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.750474021852098e-06,
+      "loss": 0.6473,
+      "step": 25229
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.749779465778609e-06,
+      "loss": 0.6138,
+      "step": 25230
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.749084921139148e-06,
+      "loss": 0.6053,
+      "step": 25231
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.748390387937244e-06,
+      "loss": 0.6616,
+      "step": 25232
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.74769586617643e-06,
+      "loss": 0.5673,
+      "step": 25233
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.74700135586023e-06,
+      "loss": 0.6018,
+      "step": 25234
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.746306856992178e-06,
+      "loss": 0.5989,
+      "step": 25235
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.7456123695758e-06,
+      "loss": 0.5619,
+      "step": 25236
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.744917893614629e-06,
+      "loss": 0.5837,
+      "step": 25237
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.744223429112187e-06,
+      "loss": 0.6136,
+      "step": 25238
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.743528976072008e-06,
+      "loss": 0.6115,
+      "step": 25239
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.742834534497617e-06,
+      "loss": 0.5938,
+      "step": 25240
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.742140104392546e-06,
+      "loss": 0.5627,
+      "step": 25241
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.741445685760325e-06,
+      "loss": 0.6363,
+      "step": 25242
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.740751278604478e-06,
+      "loss": 0.6098,
+      "step": 25243
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.740056882928533e-06,
+      "loss": 0.638,
+      "step": 25244
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.739362498736023e-06,
+      "loss": 0.6117,
+      "step": 25245
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.738668126030473e-06,
+      "loss": 0.577,
+      "step": 25246
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.737973764815417e-06,
+      "loss": 0.605,
+      "step": 25247
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.737279415094375e-06,
+      "loss": 0.6112,
+      "step": 25248
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.736585076870878e-06,
+      "loss": 0.6163,
+      "step": 25249
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.735890750148454e-06,
+      "loss": 0.6042,
+      "step": 25250
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.735196434930636e-06,
+      "loss": 0.587,
+      "step": 25251
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.734502131220947e-06,
+      "loss": 0.5626,
+      "step": 25252
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.733807839022917e-06,
+      "loss": 0.6096,
+      "step": 25253
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.733113558340073e-06,
+      "loss": 0.5424,
+      "step": 25254
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.732419289175941e-06,
+      "loss": 0.5683,
+      "step": 25255
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.731725031534052e-06,
+      "loss": 0.6103,
+      "step": 25256
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.73103078541793e-06,
+      "loss": 0.5819,
+      "step": 25257
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.73033655083111e-06,
+      "loss": 0.6293,
+      "step": 25258
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.729642327777112e-06,
+      "loss": 0.6645,
+      "step": 25259
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.728948116259464e-06,
+      "loss": 0.6371,
+      "step": 25260
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.728253916281697e-06,
+      "loss": 0.5866,
+      "step": 25261
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.727559727847338e-06,
+      "loss": 0.6247,
+      "step": 25262
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.72686555095991e-06,
+      "loss": 0.5471,
+      "step": 25263
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.726171385622952e-06,
+      "loss": 0.6204,
+      "step": 25264
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.725477231839978e-06,
+      "loss": 0.6296,
+      "step": 25265
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.72478308961452e-06,
+      "loss": 0.5738,
+      "step": 25266
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.724088958950104e-06,
+      "loss": 0.6257,
+      "step": 25267
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.723394839850262e-06,
+      "loss": 0.6066,
+      "step": 25268
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.722700732318518e-06,
+      "loss": 0.6443,
+      "step": 25269
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.722006636358396e-06,
+      "loss": 0.5921,
+      "step": 25270
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.721312551973426e-06,
+      "loss": 0.5684,
+      "step": 25271
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.720618479167134e-06,
+      "loss": 0.585,
+      "step": 25272
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.719924417943047e-06,
+      "loss": 0.5682,
+      "step": 25273
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.719230368304692e-06,
+      "loss": 0.6061,
+      "step": 25274
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.718536330255599e-06,
+      "loss": 0.5925,
+      "step": 25275
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.717842303799287e-06,
+      "loss": 0.5679,
+      "step": 25276
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.717148288939287e-06,
+      "loss": 0.5479,
+      "step": 25277
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.716454285679123e-06,
+      "loss": 0.5968,
+      "step": 25278
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.715760294022324e-06,
+      "loss": 0.605,
+      "step": 25279
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.715066313972424e-06,
+      "loss": 0.6063,
+      "step": 25280
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.714372345532934e-06,
+      "loss": 0.5684,
+      "step": 25281
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.713678388707387e-06,
+      "loss": 0.6023,
+      "step": 25282
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.71298444349931e-06,
+      "loss": 0.5718,
+      "step": 25283
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.71229050991223e-06,
+      "loss": 0.6108,
+      "step": 25284
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.711596587949675e-06,
+      "loss": 0.5981,
+      "step": 25285
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.710902677615165e-06,
+      "loss": 0.5573,
+      "step": 25286
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.710208778912228e-06,
+      "loss": 0.5552,
+      "step": 25287
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.709514891844391e-06,
+      "loss": 0.6123,
+      "step": 25288
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.70882101641518e-06,
+      "loss": 0.5372,
+      "step": 25289
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.708127152628121e-06,
+      "loss": 0.5764,
+      "step": 25290
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.70743330048674e-06,
+      "loss": 0.5854,
+      "step": 25291
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.70673945999456e-06,
+      "loss": 0.5879,
+      "step": 25292
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.70604563115511e-06,
+      "loss": 0.5738,
+      "step": 25293
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.705351813971913e-06,
+      "loss": 0.5755,
+      "step": 25294
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.704658008448494e-06,
+      "loss": 0.5553,
+      "step": 25295
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.703964214588385e-06,
+      "loss": 0.5741,
+      "step": 25296
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.703270432395103e-06,
+      "loss": 0.689,
+      "step": 25297
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.702576661872174e-06,
+      "loss": 0.5837,
+      "step": 25298
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.701882903023128e-06,
+      "loss": 0.5897,
+      "step": 25299
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.701189155851487e-06,
+      "loss": 0.5458,
+      "step": 25300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.700495420360778e-06,
+      "loss": 0.6533,
+      "step": 25301
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.699801696554527e-06,
+      "loss": 0.5624,
+      "step": 25302
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.699107984436257e-06,
+      "loss": 0.5812,
+      "step": 25303
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.69841428400949e-06,
+      "loss": 0.5835,
+      "step": 25304
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.697720595277754e-06,
+      "loss": 0.6227,
+      "step": 25305
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.697026918244576e-06,
+      "loss": 0.6455,
+      "step": 25306
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.696333252913482e-06,
+      "loss": 0.5989,
+      "step": 25307
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.695639599287988e-06,
+      "loss": 0.6833,
+      "step": 25308
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.694945957371624e-06,
+      "loss": 0.6556,
+      "step": 25309
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.69425232716792e-06,
+      "loss": 0.5334,
+      "step": 25310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.69355870868039e-06,
+      "loss": 0.5998,
+      "step": 25311
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.692865101912566e-06,
+      "loss": 0.6265,
+      "step": 25312
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.692171506867971e-06,
+      "loss": 0.5768,
+      "step": 25313
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.691477923550127e-06,
+      "loss": 0.614,
+      "step": 25314
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.69078435196256e-06,
+      "loss": 0.5856,
+      "step": 25315
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.69009079210879e-06,
+      "loss": 0.6325,
+      "step": 25316
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.68939724399235e-06,
+      "loss": 0.6481,
+      "step": 25317
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.68870370761676e-06,
+      "loss": 0.6367,
+      "step": 25318
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.688010182985541e-06,
+      "loss": 0.5591,
+      "step": 25319
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.687316670102219e-06,
+      "loss": 0.642,
+      "step": 25320
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.68662316897032e-06,
+      "loss": 0.5847,
+      "step": 25321
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.685929679593363e-06,
+      "loss": 0.5988,
+      "step": 25322
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.685236201974882e-06,
+      "loss": 0.6029,
+      "step": 25323
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.684542736118388e-06,
+      "loss": 0.5762,
+      "step": 25324
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.68384928202741e-06,
+      "loss": 0.6019,
+      "step": 25325
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.683155839705473e-06,
+      "loss": 0.6244,
+      "step": 25326
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.6824624091561e-06,
+      "loss": 0.564,
+      "step": 25327
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.681768990382817e-06,
+      "loss": 0.5694,
+      "step": 25328
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.681075583389143e-06,
+      "loss": 0.5869,
+      "step": 25329
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.6803821881786e-06,
+      "loss": 0.5986,
+      "step": 25330
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.67968880475472e-06,
+      "loss": 0.6082,
+      "step": 25331
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.678995433121016e-06,
+      "loss": 0.6134,
+      "step": 25332
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.678302073281016e-06,
+      "loss": 0.6183,
+      "step": 25333
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.677608725238249e-06,
+      "loss": 0.6022,
+      "step": 25334
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.676915388996227e-06,
+      "loss": 0.6479,
+      "step": 25335
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.67622206455848e-06,
+      "loss": 0.6227,
+      "step": 25336
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.675528751928525e-06,
+      "loss": 0.6463,
+      "step": 25337
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.674835451109892e-06,
+      "loss": 0.6627,
+      "step": 25338
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.674142162106105e-06,
+      "loss": 0.5977,
+      "step": 25339
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.673448884920678e-06,
+      "loss": 0.606,
+      "step": 25340
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.672755619557139e-06,
+      "loss": 0.6324,
+      "step": 25341
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.672062366019012e-06,
+      "loss": 0.6546,
+      "step": 25342
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.671369124309816e-06,
+      "loss": 0.6062,
+      "step": 25343
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.670675894433075e-06,
+      "loss": 0.6014,
+      "step": 25344
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.669982676392317e-06,
+      "loss": 0.5726,
+      "step": 25345
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.669289470191055e-06,
+      "loss": 0.5916,
+      "step": 25346
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.668596275832815e-06,
+      "loss": 0.5544,
+      "step": 25347
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.667903093321125e-06,
+      "loss": 0.6081,
+      "step": 25348
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.667209922659498e-06,
+      "loss": 0.6253,
+      "step": 25349
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.666516763851461e-06,
+      "loss": 0.6274,
+      "step": 25350
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.665823616900536e-06,
+      "loss": 0.5804,
+      "step": 25351
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.665130481810248e-06,
+      "loss": 0.635,
+      "step": 25352
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.664437358584112e-06,
+      "loss": 0.6327,
+      "step": 25353
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.663744247225652e-06,
+      "loss": 0.5837,
+      "step": 25354
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.663051147738394e-06,
+      "loss": 0.6015,
+      "step": 25355
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.662358060125863e-06,
+      "loss": 0.6364,
+      "step": 25356
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.661664984391569e-06,
+      "loss": 0.586,
+      "step": 25357
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.660971920539041e-06,
+      "loss": 0.5315,
+      "step": 25358
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.660278868571799e-06,
+      "loss": 0.5797,
+      "step": 25359
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.659585828493366e-06,
+      "loss": 0.5875,
+      "step": 25360
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.658892800307266e-06,
+      "loss": 0.6162,
+      "step": 25361
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.658199784017013e-06,
+      "loss": 0.5738,
+      "step": 25362
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.657506779626134e-06,
+      "loss": 0.6257,
+      "step": 25363
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.656813787138148e-06,
+      "loss": 0.5392,
+      "step": 25364
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.656120806556579e-06,
+      "loss": 0.6203,
+      "step": 25365
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.655427837884947e-06,
+      "loss": 0.5602,
+      "step": 25366
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.654734881126773e-06,
+      "loss": 0.5963,
+      "step": 25367
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.654041936285577e-06,
+      "loss": 0.6446,
+      "step": 25368
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.653349003364882e-06,
+      "loss": 0.587,
+      "step": 25369
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.652656082368206e-06,
+      "loss": 0.5859,
+      "step": 25370
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.651963173299071e-06,
+      "loss": 0.6351,
+      "step": 25371
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.651270276161003e-06,
+      "loss": 0.6448,
+      "step": 25372
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.650577390957516e-06,
+      "loss": 0.6059,
+      "step": 25373
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.649884517692135e-06,
+      "loss": 0.6506,
+      "step": 25374
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.649191656368375e-06,
+      "loss": 0.5833,
+      "step": 25375
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.648498806989762e-06,
+      "loss": 0.6269,
+      "step": 25376
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.64780596955982e-06,
+      "loss": 0.5319,
+      "step": 25377
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.647113144082062e-06,
+      "loss": 0.5864,
+      "step": 25378
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.646420330560008e-06,
+      "loss": 0.5492,
+      "step": 25379
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.645727528997182e-06,
+      "loss": 0.624,
+      "step": 25380
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.645034739397106e-06,
+      "loss": 0.5939,
+      "step": 25381
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.644341961763297e-06,
+      "loss": 0.6104,
+      "step": 25382
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.643649196099283e-06,
+      "loss": 0.6814,
+      "step": 25383
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.64295644240857e-06,
+      "loss": 0.6182,
+      "step": 25384
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.642263700694687e-06,
+      "loss": 0.5901,
+      "step": 25385
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.641570970961151e-06,
+      "loss": 0.6067,
+      "step": 25386
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.640878253211489e-06,
+      "loss": 0.6387,
+      "step": 25387
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.640185547449212e-06,
+      "loss": 0.6595,
+      "step": 25388
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.639492853677843e-06,
+      "loss": 0.5671,
+      "step": 25389
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.638800171900904e-06,
+      "loss": 0.6001,
+      "step": 25390
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.63810750212191e-06,
+      "loss": 0.6391,
+      "step": 25391
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.637414844344386e-06,
+      "loss": 0.5846,
+      "step": 25392
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.636722198571846e-06,
+      "loss": 0.5759,
+      "step": 25393
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.636029564807819e-06,
+      "loss": 0.6038,
+      "step": 25394
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.635336943055814e-06,
+      "loss": 0.6167,
+      "step": 25395
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.634644333319352e-06,
+      "loss": 0.5874,
+      "step": 25396
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.633951735601956e-06,
+      "loss": 0.5616,
+      "step": 25397
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.633259149907145e-06,
+      "loss": 0.5992,
+      "step": 25398
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.632566576238441e-06,
+      "loss": 0.6139,
+      "step": 25399
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.631874014599356e-06,
+      "loss": 0.6503,
+      "step": 25400
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.631181464993413e-06,
+      "loss": 0.6055,
+      "step": 25401
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.630488927424129e-06,
+      "loss": 0.6059,
+      "step": 25402
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.629796401895026e-06,
+      "loss": 0.6174,
+      "step": 25403
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.629103888409626e-06,
+      "loss": 0.6243,
+      "step": 25404
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6284113869714385e-06,
+      "loss": 0.5947,
+      "step": 25405
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.627718897583987e-06,
+      "loss": 0.5829,
+      "step": 25406
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.627026420250792e-06,
+      "loss": 0.6734,
+      "step": 25407
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6263339549753714e-06,
+      "loss": 0.559,
+      "step": 25408
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.62564150176124e-06,
+      "loss": 0.6015,
+      "step": 25409
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.624949060611924e-06,
+      "loss": 0.6183,
+      "step": 25410
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.624256631530936e-06,
+      "loss": 0.624,
+      "step": 25411
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6235642145217945e-06,
+      "loss": 0.6087,
+      "step": 25412
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.62287180958802e-06,
+      "loss": 0.6089,
+      "step": 25413
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6221794167331285e-06,
+      "loss": 0.6234,
+      "step": 25414
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.621487035960644e-06,
+      "loss": 0.65,
+      "step": 25415
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.620794667274076e-06,
+      "loss": 0.5601,
+      "step": 25416
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.620102310676949e-06,
+      "loss": 0.5308,
+      "step": 25417
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.619409966172777e-06,
+      "loss": 0.6639,
+      "step": 25418
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.618717633765081e-06,
+      "loss": 0.6206,
+      "step": 25419
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.618025313457378e-06,
+      "loss": 0.6252,
+      "step": 25420
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.617333005253191e-06,
+      "loss": 0.6569,
+      "step": 25421
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.616640709156028e-06,
+      "loss": 0.6,
+      "step": 25422
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.615948425169412e-06,
+      "loss": 0.6127,
+      "step": 25423
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.615256153296859e-06,
+      "loss": 0.6119,
+      "step": 25424
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.614563893541891e-06,
+      "loss": 0.6484,
+      "step": 25425
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.613871645908022e-06,
+      "loss": 0.6077,
+      "step": 25426
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.613179410398768e-06,
+      "loss": 0.5809,
+      "step": 25427
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6124871870176485e-06,
+      "loss": 0.5915,
+      "step": 25428
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.611794975768183e-06,
+      "loss": 0.5781,
+      "step": 25429
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.611102776653885e-06,
+      "loss": 0.5913,
+      "step": 25430
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6104105896782745e-06,
+      "loss": 0.6217,
+      "step": 25431
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.609718414844867e-06,
+      "loss": 0.6209,
+      "step": 25432
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.60902625215718e-06,
+      "loss": 0.6774,
+      "step": 25433
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.608334101618729e-06,
+      "loss": 0.6756,
+      "step": 25434
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6076419632330345e-06,
+      "loss": 0.6121,
+      "step": 25435
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.606949837003612e-06,
+      "loss": 0.6297,
+      "step": 25436
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.606257722933981e-06,
+      "loss": 0.5716,
+      "step": 25437
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.605565621027652e-06,
+      "loss": 0.5537,
+      "step": 25438
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.604873531288146e-06,
+      "loss": 0.6272,
+      "step": 25439
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.604181453718977e-06,
+      "loss": 0.5932,
+      "step": 25440
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.603489388323666e-06,
+      "loss": 0.6206,
+      "step": 25441
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.6027973351057305e-06,
+      "loss": 0.5743,
+      "step": 25442
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.60210529406868e-06,
+      "loss": 0.5838,
+      "step": 25443
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.601413265216033e-06,
+      "loss": 0.617,
+      "step": 25444
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.600721248551311e-06,
+      "loss": 0.6207,
+      "step": 25445
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.600029244078027e-06,
+      "loss": 0.6041,
+      "step": 25446
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.599337251799695e-06,
+      "loss": 0.6018,
+      "step": 25447
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.598645271719836e-06,
+      "loss": 0.5945,
+      "step": 25448
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.597953303841963e-06,
+      "loss": 0.5645,
+      "step": 25449
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.597261348169591e-06,
+      "loss": 0.5843,
+      "step": 25450
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.596569404706238e-06,
+      "loss": 0.5849,
+      "step": 25451
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5958774734554194e-06,
+      "loss": 0.6069,
+      "step": 25452
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5951855544206565e-06,
+      "loss": 0.6229,
+      "step": 25453
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.594493647605454e-06,
+      "loss": 0.5225,
+      "step": 25454
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.593801753013336e-06,
+      "loss": 0.5676,
+      "step": 25455
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.593109870647815e-06,
+      "loss": 0.6277,
+      "step": 25456
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.592418000512407e-06,
+      "loss": 0.6266,
+      "step": 25457
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.591726142610629e-06,
+      "loss": 0.6256,
+      "step": 25458
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.591034296946001e-06,
+      "loss": 0.5799,
+      "step": 25459
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.590342463522027e-06,
+      "loss": 0.6126,
+      "step": 25460
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.58965064234223e-06,
+      "loss": 0.5563,
+      "step": 25461
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.588958833410123e-06,
+      "loss": 0.5883,
+      "step": 25462
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5882670367292235e-06,
+      "loss": 0.6002,
+      "step": 25463
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5875752523030485e-06,
+      "loss": 0.5903,
+      "step": 25464
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.586883480135106e-06,
+      "loss": 0.588,
+      "step": 25465
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.586191720228916e-06,
+      "loss": 0.6346,
+      "step": 25466
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.585499972587995e-06,
+      "loss": 0.5953,
+      "step": 25467
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.584808237215853e-06,
+      "loss": 0.646,
+      "step": 25468
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.58411651411601e-06,
+      "loss": 0.5784,
+      "step": 25469
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.583424803291979e-06,
+      "loss": 0.5462,
+      "step": 25470
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.582733104747271e-06,
+      "loss": 0.6082,
+      "step": 25471
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5820414184854055e-06,
+      "loss": 0.6372,
+      "step": 25472
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.581349744509894e-06,
+      "loss": 0.6385,
+      "step": 25473
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.580658082824254e-06,
+      "loss": 0.6376,
+      "step": 25474
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.579966433432002e-06,
+      "loss": 0.599,
+      "step": 25475
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5792747963366465e-06,
+      "loss": 0.5966,
+      "step": 25476
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5785831715417045e-06,
+      "loss": 0.5398,
+      "step": 25477
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.577891559050689e-06,
+      "loss": 0.6085,
+      "step": 25478
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.577199958867117e-06,
+      "loss": 0.6096,
+      "step": 25479
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.576508370994506e-06,
+      "loss": 0.5626,
+      "step": 25480
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.575816795436361e-06,
+      "loss": 0.5898,
+      "step": 25481
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.575125232196201e-06,
+      "loss": 0.608,
+      "step": 25482
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.57443368127754e-06,
+      "loss": 0.6533,
+      "step": 25483
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.573742142683892e-06,
+      "loss": 0.565,
+      "step": 25484
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.573050616418772e-06,
+      "loss": 0.6164,
+      "step": 25485
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5723591024856936e-06,
+      "loss": 0.5797,
+      "step": 25486
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.5716676008881675e-06,
+      "loss": 0.5925,
+      "step": 25487
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.570976111629711e-06,
+      "loss": 0.6247,
+      "step": 25488
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.570284634713834e-06,
+      "loss": 0.6003,
+      "step": 25489
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.569593170144053e-06,
+      "loss": 0.599,
+      "step": 25490
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.568901717923884e-06,
+      "loss": 0.6115,
+      "step": 25491
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.568210278056833e-06,
+      "loss": 0.5949,
+      "step": 25492
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.56751885054642e-06,
+      "loss": 0.5702,
+      "step": 25493
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.566827435396154e-06,
+      "loss": 0.541,
+      "step": 25494
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.566136032609551e-06,
+      "loss": 0.6229,
+      "step": 25495
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.565444642190129e-06,
+      "loss": 0.604,
+      "step": 25496
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.564753264141389e-06,
+      "loss": 0.6893,
+      "step": 25497
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.564061898466852e-06,
+      "loss": 0.5697,
+      "step": 25498
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.563370545170031e-06,
+      "loss": 0.6092,
+      "step": 25499
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.562679204254438e-06,
+      "loss": 0.6162,
+      "step": 25500
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.561987875723585e-06,
+      "loss": 0.6133,
+      "step": 25501
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.561296559580988e-06,
+      "loss": 0.6106,
+      "step": 25502
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.560605255830156e-06,
+      "loss": 0.5513,
+      "step": 25503
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5599139644746015e-06,
+      "loss": 0.6131,
+      "step": 25504
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.559222685517841e-06,
+      "loss": 0.6134,
+      "step": 25505
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.558531418963383e-06,
+      "loss": 0.5849,
+      "step": 25506
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5578401648147435e-06,
+      "loss": 0.6766,
+      "step": 25507
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.557148923075432e-06,
+      "loss": 0.6057,
+      "step": 25508
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5564576937489645e-06,
+      "loss": 0.6407,
+      "step": 25509
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.555766476838849e-06,
+      "loss": 0.5643,
+      "step": 25510
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.555075272348599e-06,
+      "loss": 0.5465,
+      "step": 25511
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.554384080281727e-06,
+      "loss": 0.6543,
+      "step": 25512
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.553692900641751e-06,
+      "loss": 0.6037,
+      "step": 25513
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.553001733432174e-06,
+      "loss": 0.6228,
+      "step": 25514
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.552310578656512e-06,
+      "loss": 0.569,
+      "step": 25515
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.551619436318276e-06,
+      "loss": 0.6416,
+      "step": 25516
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.550928306420978e-06,
+      "loss": 0.6104,
+      "step": 25517
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.550237188968136e-06,
+      "loss": 0.5654,
+      "step": 25518
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.549546083963252e-06,
+      "loss": 0.6144,
+      "step": 25519
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.548854991409841e-06,
+      "loss": 0.62,
+      "step": 25520
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.548163911311417e-06,
+      "loss": 0.6475,
+      "step": 25521
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.547472843671489e-06,
+      "loss": 0.6431,
+      "step": 25522
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5467817884935745e-06,
+      "loss": 0.6077,
+      "step": 25523
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.546090745781176e-06,
+      "loss": 0.591,
+      "step": 25524
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.545399715537809e-06,
+      "loss": 0.5844,
+      "step": 25525
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.544708697766987e-06,
+      "loss": 0.5939,
+      "step": 25526
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.544017692472216e-06,
+      "loss": 0.614,
+      "step": 25527
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.543326699657012e-06,
+      "loss": 0.6495,
+      "step": 25528
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.542635719324886e-06,
+      "loss": 0.6047,
+      "step": 25529
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.541944751479348e-06,
+      "loss": 0.6229,
+      "step": 25530
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.541253796123906e-06,
+      "loss": 0.5768,
+      "step": 25531
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.540562853262074e-06,
+      "loss": 0.6452,
+      "step": 25532
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.539871922897362e-06,
+      "loss": 0.5892,
+      "step": 25533
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.539181005033284e-06,
+      "loss": 0.5776,
+      "step": 25534
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.538490099673347e-06,
+      "loss": 0.6825,
+      "step": 25535
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.53779920682106e-06,
+      "loss": 0.5346,
+      "step": 25536
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.537108326479938e-06,
+      "loss": 0.575,
+      "step": 25537
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5364174586534884e-06,
+      "loss": 0.5436,
+      "step": 25538
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.535726603345225e-06,
+      "loss": 0.5856,
+      "step": 25539
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.53503576055866e-06,
+      "loss": 0.6289,
+      "step": 25540
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.534344930297295e-06,
+      "loss": 0.5886,
+      "step": 25541
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.533654112564646e-06,
+      "loss": 0.6089,
+      "step": 25542
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.532963307364223e-06,
+      "loss": 0.6397,
+      "step": 25543
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.532272514699538e-06,
+      "loss": 0.571,
+      "step": 25544
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5315817345740995e-06,
+      "loss": 0.5929,
+      "step": 25545
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.530890966991415e-06,
+      "loss": 0.5828,
+      "step": 25546
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.530200211954998e-06,
+      "loss": 0.5856,
+      "step": 25547
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.529509469468354e-06,
+      "loss": 0.6009,
+      "step": 25548
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.528818739534998e-06,
+      "loss": 0.6462,
+      "step": 25549
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.528128022158436e-06,
+      "loss": 0.5801,
+      "step": 25550
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.527437317342184e-06,
+      "loss": 0.5696,
+      "step": 25551
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5267466250897435e-06,
+      "loss": 0.6668,
+      "step": 25552
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.526055945404628e-06,
+      "loss": 0.5815,
+      "step": 25553
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.525365278290346e-06,
+      "loss": 0.59,
+      "step": 25554
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.524674623750409e-06,
+      "loss": 0.6537,
+      "step": 25555
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.523983981788327e-06,
+      "loss": 0.5488,
+      "step": 25556
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.523293352407606e-06,
+      "loss": 0.5909,
+      "step": 25557
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.522602735611755e-06,
+      "loss": 0.6162,
+      "step": 25558
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5219121314042854e-06,
+      "loss": 0.6279,
+      "step": 25559
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.521221539788707e-06,
+      "loss": 0.6169,
+      "step": 25560
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.520530960768531e-06,
+      "loss": 0.5799,
+      "step": 25561
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.51984039434726e-06,
+      "loss": 0.6333,
+      "step": 25562
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5191498405284055e-06,
+      "loss": 0.5524,
+      "step": 25563
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.518459299315477e-06,
+      "loss": 0.5444,
+      "step": 25564
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5177687707119865e-06,
+      "loss": 0.6225,
+      "step": 25565
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.517078254721437e-06,
+      "loss": 0.63,
+      "step": 25566
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.516387751347343e-06,
+      "loss": 0.5687,
+      "step": 25567
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.515697260593209e-06,
+      "loss": 0.5914,
+      "step": 25568
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.515006782462544e-06,
+      "loss": 0.5544,
+      "step": 25569
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.514316316958856e-06,
+      "loss": 0.6383,
+      "step": 25570
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.513625864085656e-06,
+      "loss": 0.6282,
+      "step": 25571
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.512935423846454e-06,
+      "loss": 0.6287,
+      "step": 25572
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.512244996244753e-06,
+      "loss": 0.5627,
+      "step": 25573
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5115545812840615e-06,
+      "loss": 0.6202,
+      "step": 25574
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.510864178967892e-06,
+      "loss": 0.679,
+      "step": 25575
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.510173789299751e-06,
+      "loss": 0.5757,
+      "step": 25576
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.509483412283144e-06,
+      "loss": 0.65,
+      "step": 25577
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.508793047921586e-06,
+      "loss": 0.5959,
+      "step": 25578
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.508102696218577e-06,
+      "loss": 0.6547,
+      "step": 25579
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.507412357177627e-06,
+      "loss": 0.5153,
+      "step": 25580
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5067220308022445e-06,
+      "loss": 0.5537,
+      "step": 25581
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5060317170959404e-06,
+      "loss": 0.5884,
+      "step": 25582
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.50534141606222e-06,
+      "loss": 0.5923,
+      "step": 25583
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.504651127704587e-06,
+      "loss": 0.5817,
+      "step": 25584
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.503960852026555e-06,
+      "loss": 0.6213,
+      "step": 25585
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.503270589031628e-06,
+      "loss": 0.6051,
+      "step": 25586
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.502580338723313e-06,
+      "loss": 0.5979,
+      "step": 25587
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.5018901011051216e-06,
+      "loss": 0.5642,
+      "step": 25588
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.501199876180557e-06,
+      "loss": 0.5842,
+      "step": 25589
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.500509663953126e-06,
+      "loss": 0.5818,
+      "step": 25590
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.499819464426339e-06,
+      "loss": 0.6174,
+      "step": 25591
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.499129277603699e-06,
+      "loss": 0.6935,
+      "step": 25592
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.498439103488717e-06,
+      "loss": 0.638,
+      "step": 25593
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.497748942084903e-06,
+      "loss": 0.591,
+      "step": 25594
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.497058793395755e-06,
+      "loss": 0.5695,
+      "step": 25595
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.496368657424784e-06,
+      "loss": 0.6016,
+      "step": 25596
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.495678534175497e-06,
+      "loss": 0.6466,
+      "step": 25597
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.494988423651402e-06,
+      "loss": 0.6191,
+      "step": 25598
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4942983258560066e-06,
+      "loss": 0.6136,
+      "step": 25599
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.493608240792811e-06,
+      "loss": 0.5694,
+      "step": 25600
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.492918168465328e-06,
+      "loss": 0.6344,
+      "step": 25601
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.492228108877061e-06,
+      "loss": 0.573,
+      "step": 25602
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.491538062031519e-06,
+      "loss": 0.566,
+      "step": 25603
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.490848027932205e-06,
+      "loss": 0.5867,
+      "step": 25604
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.490158006582629e-06,
+      "loss": 0.6506,
+      "step": 25605
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.489467997986296e-06,
+      "loss": 0.6626,
+      "step": 25606
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4887780021467085e-06,
+      "loss": 0.5711,
+      "step": 25607
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.488088019067377e-06,
+      "loss": 0.5886,
+      "step": 25608
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.487398048751804e-06,
+      "loss": 0.6173,
+      "step": 25609
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.486708091203502e-06,
+      "loss": 0.5729,
+      "step": 25610
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.486018146425968e-06,
+      "loss": 0.5597,
+      "step": 25611
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.485328214422711e-06,
+      "loss": 0.5731,
+      "step": 25612
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.484638295197239e-06,
+      "loss": 0.5881,
+      "step": 25613
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.483948388753057e-06,
+      "loss": 0.6282,
+      "step": 25614
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.483258495093675e-06,
+      "loss": 0.5154,
+      "step": 25615
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4825686142225876e-06,
+      "loss": 0.6152,
+      "step": 25616
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.481878746143307e-06,
+      "loss": 0.602,
+      "step": 25617
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4811888908593376e-06,
+      "loss": 0.5415,
+      "step": 25618
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.480499048374186e-06,
+      "loss": 0.5806,
+      "step": 25619
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.479809218691356e-06,
+      "loss": 0.56,
+      "step": 25620
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.479119401814357e-06,
+      "loss": 0.607,
+      "step": 25621
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.478429597746688e-06,
+      "loss": 0.5892,
+      "step": 25622
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.477739806491855e-06,
+      "loss": 0.5901,
+      "step": 25623
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.477050028053368e-06,
+      "loss": 0.6308,
+      "step": 25624
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4763602624347254e-06,
+      "loss": 0.5475,
+      "step": 25625
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.475670509639439e-06,
+      "loss": 0.6032,
+      "step": 25626
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.47498076967101e-06,
+      "loss": 0.5956,
+      "step": 25627
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.47429104253294e-06,
+      "loss": 0.5423,
+      "step": 25628
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.473601328228737e-06,
+      "loss": 0.5755,
+      "step": 25629
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.472911626761905e-06,
+      "loss": 0.606,
+      "step": 25630
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.4722219381359504e-06,
+      "loss": 0.6128,
+      "step": 25631
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.47153226235438e-06,
+      "loss": 0.5668,
+      "step": 25632
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.470842599420691e-06,
+      "loss": 0.6822,
+      "step": 25633
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.470152949338391e-06,
+      "loss": 0.6364,
+      "step": 25634
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.469463312110985e-06,
+      "loss": 0.6436,
+      "step": 25635
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.468773687741977e-06,
+      "loss": 0.5792,
+      "step": 25636
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.468084076234874e-06,
+      "loss": 0.5975,
+      "step": 25637
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.467394477593175e-06,
+      "loss": 0.6016,
+      "step": 25638
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.466704891820385e-06,
+      "loss": 0.6086,
+      "step": 25639
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.46601531892001e-06,
+      "loss": 0.5829,
+      "step": 25640
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.465325758895556e-06,
+      "loss": 0.5597,
+      "step": 25641
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.464636211750522e-06,
+      "loss": 0.5923,
+      "step": 25642
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.463946677488415e-06,
+      "loss": 0.577,
+      "step": 25643
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.463257156112737e-06,
+      "loss": 0.6402,
+      "step": 25644
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.462567647626993e-06,
+      "loss": 0.6094,
+      "step": 25645
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.461878152034684e-06,
+      "loss": 0.5891,
+      "step": 25646
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4611886693393164e-06,
+      "loss": 0.6598,
+      "step": 25647
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.460499199544395e-06,
+      "loss": 0.6194,
+      "step": 25648
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.459809742653419e-06,
+      "loss": 0.5599,
+      "step": 25649
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.459120298669892e-06,
+      "loss": 0.5918,
+      "step": 25650
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.458430867597319e-06,
+      "loss": 0.5961,
+      "step": 25651
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.457741449439204e-06,
+      "loss": 0.6041,
+      "step": 25652
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.457052044199054e-06,
+      "loss": 0.5569,
+      "step": 25653
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.456362651880361e-06,
+      "loss": 0.5841,
+      "step": 25654
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.455673272486636e-06,
+      "loss": 0.5511,
+      "step": 25655
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.454983906021379e-06,
+      "loss": 0.5831,
+      "step": 25656
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.454294552488095e-06,
+      "loss": 0.5343,
+      "step": 25657
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.453605211890285e-06,
+      "loss": 0.5849,
+      "step": 25658
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4529158842314575e-06,
+      "loss": 0.5701,
+      "step": 25659
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.452226569515107e-06,
+      "loss": 0.6854,
+      "step": 25660
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.451537267744737e-06,
+      "loss": 0.5943,
+      "step": 25661
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.450847978923857e-06,
+      "loss": 0.5778,
+      "step": 25662
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.450158703055963e-06,
+      "loss": 0.5778,
+      "step": 25663
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.44946944014456e-06,
+      "loss": 0.5983,
+      "step": 25664
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.448780190193148e-06,
+      "loss": 0.6101,
+      "step": 25665
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.448090953205234e-06,
+      "loss": 0.6102,
+      "step": 25666
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.447401729184313e-06,
+      "loss": 0.6084,
+      "step": 25667
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.446712518133894e-06,
+      "loss": 0.592,
+      "step": 25668
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.446023320057475e-06,
+      "loss": 0.6373,
+      "step": 25669
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.445334134958564e-06,
+      "loss": 0.6286,
+      "step": 25670
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.444644962840655e-06,
+      "loss": 0.6624,
+      "step": 25671
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4439558037072514e-06,
+      "loss": 0.5824,
+      "step": 25672
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.443266657561858e-06,
+      "loss": 0.6223,
+      "step": 25673
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.442577524407976e-06,
+      "loss": 0.5434,
+      "step": 25674
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.441888404249109e-06,
+      "loss": 0.5954,
+      "step": 25675
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.441199297088755e-06,
+      "loss": 0.6415,
+      "step": 25676
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4405102029304146e-06,
+      "loss": 0.6639,
+      "step": 25677
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.439821121777591e-06,
+      "loss": 0.6173,
+      "step": 25678
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.439132053633787e-06,
+      "loss": 0.5877,
+      "step": 25679
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4384429985025075e-06,
+      "loss": 0.6292,
+      "step": 25680
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.437753956387244e-06,
+      "loss": 0.6309,
+      "step": 25681
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.437064927291504e-06,
+      "loss": 0.6306,
+      "step": 25682
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4363759112187895e-06,
+      "loss": 0.6371,
+      "step": 25683
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.435686908172599e-06,
+      "loss": 0.6165,
+      "step": 25684
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.434997918156433e-06,
+      "loss": 0.5743,
+      "step": 25685
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.434308941173798e-06,
+      "loss": 0.6345,
+      "step": 25686
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.433619977228186e-06,
+      "loss": 0.5856,
+      "step": 25687
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4329310263231025e-06,
+      "loss": 0.5819,
+      "step": 25688
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.43224208846205e-06,
+      "loss": 0.547,
+      "step": 25689
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.431553163648526e-06,
+      "loss": 0.5634,
+      "step": 25690
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.430864251886037e-06,
+      "loss": 0.6357,
+      "step": 25691
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4301753531780754e-06,
+      "loss": 0.6095,
+      "step": 25692
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4294864675281444e-06,
+      "loss": 0.6166,
+      "step": 25693
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.428797594939745e-06,
+      "loss": 0.5922,
+      "step": 25694
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4281087354163795e-06,
+      "loss": 0.5588,
+      "step": 25695
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.427419888961547e-06,
+      "loss": 0.6161,
+      "step": 25696
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.426731055578751e-06,
+      "loss": 0.6085,
+      "step": 25697
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.426042235271483e-06,
+      "loss": 0.6037,
+      "step": 25698
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.42535342804325e-06,
+      "loss": 0.5739,
+      "step": 25699
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.424664633897549e-06,
+      "loss": 0.5897,
+      "step": 25700
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.423975852837884e-06,
+      "loss": 0.5867,
+      "step": 25701
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.423287084867752e-06,
+      "loss": 0.6367,
+      "step": 25702
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.42259832999065e-06,
+      "loss": 0.6603,
+      "step": 25703
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.421909588210083e-06,
+      "loss": 0.6722,
+      "step": 25704
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.421220859529548e-06,
+      "loss": 0.6086,
+      "step": 25705
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.420532143952543e-06,
+      "loss": 0.5803,
+      "step": 25706
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.419843441482575e-06,
+      "loss": 0.647,
+      "step": 25707
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.419154752123133e-06,
+      "loss": 0.5886,
+      "step": 25708
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.418466075877723e-06,
+      "loss": 0.6561,
+      "step": 25709
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.417777412749841e-06,
+      "loss": 0.5647,
+      "step": 25710
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4170887627429896e-06,
+      "loss": 0.5732,
+      "step": 25711
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.416400125860665e-06,
+      "loss": 0.5879,
+      "step": 25712
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.415711502106375e-06,
+      "loss": 0.544,
+      "step": 25713
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.415022891483605e-06,
+      "loss": 0.5623,
+      "step": 25714
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.414334293995862e-06,
+      "loss": 0.6059,
+      "step": 25715
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.413645709646644e-06,
+      "loss": 0.5791,
+      "step": 25716
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.412957138439449e-06,
+      "loss": 0.6417,
+      "step": 25717
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.41226858037778e-06,
+      "loss": 0.6744,
+      "step": 25718
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.411580035465129e-06,
+      "loss": 0.6055,
+      "step": 25719
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.410891503704998e-06,
+      "loss": 0.5909,
+      "step": 25720
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.410202985100885e-06,
+      "loss": 0.5968,
+      "step": 25721
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.40951447965629e-06,
+      "loss": 0.5502,
+      "step": 25722
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.40882598737471e-06,
+      "loss": 0.5609,
+      "step": 25723
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4081375082596465e-06,
+      "loss": 0.587,
+      "step": 25724
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.407449042314594e-06,
+      "loss": 0.5821,
+      "step": 25725
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4067605895430515e-06,
+      "loss": 0.5886,
+      "step": 25726
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.406072149948517e-06,
+      "loss": 0.6764,
+      "step": 25727
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4053837235344894e-06,
+      "loss": 0.5759,
+      "step": 25728
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.404695310304472e-06,
+      "loss": 0.587,
+      "step": 25729
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.404006910261953e-06,
+      "loss": 0.5789,
+      "step": 25730
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.403318523410434e-06,
+      "loss": 0.6186,
+      "step": 25731
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4026301497534155e-06,
+      "loss": 0.5726,
+      "step": 25732
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.401941789294395e-06,
+      "loss": 0.5239,
+      "step": 25733
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.4012534420368706e-06,
+      "loss": 0.5703,
+      "step": 25734
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.400565107984337e-06,
+      "loss": 0.5978,
+      "step": 25735
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.399876787140292e-06,
+      "loss": 0.5776,
+      "step": 25736
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.399188479508235e-06,
+      "loss": 0.5444,
+      "step": 25737
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.398500185091661e-06,
+      "loss": 0.6435,
+      "step": 25738
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.397811903894074e-06,
+      "loss": 0.6534,
+      "step": 25739
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3971236359189665e-06,
+      "loss": 0.6243,
+      "step": 25740
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.396435381169833e-06,
+      "loss": 0.6025,
+      "step": 25741
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.395747139650178e-06,
+      "loss": 0.5915,
+      "step": 25742
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.395058911363491e-06,
+      "loss": 0.6393,
+      "step": 25743
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.394370696313271e-06,
+      "loss": 0.6989,
+      "step": 25744
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.393682494503021e-06,
+      "loss": 0.6469,
+      "step": 25745
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.392994305936232e-06,
+      "loss": 0.6162,
+      "step": 25746
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3923061306164015e-06,
+      "loss": 0.6025,
+      "step": 25747
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.391617968547027e-06,
+      "loss": 0.5522,
+      "step": 25748
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.390929819731605e-06,
+      "loss": 0.6259,
+      "step": 25749
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3902416841736324e-06,
+      "loss": 0.5776,
+      "step": 25750
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.389553561876611e-06,
+      "loss": 0.5929,
+      "step": 25751
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.388865452844027e-06,
+      "loss": 0.5796,
+      "step": 25752
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3881773570793845e-06,
+      "loss": 0.5962,
+      "step": 25753
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.387489274586177e-06,
+      "loss": 0.618,
+      "step": 25754
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3868012053679005e-06,
+      "loss": 0.5601,
+      "step": 25755
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.386113149428058e-06,
+      "loss": 0.6047,
+      "step": 25756
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.385425106770134e-06,
+      "loss": 0.5983,
+      "step": 25757
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.384737077397632e-06,
+      "loss": 0.6024,
+      "step": 25758
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.384049061314047e-06,
+      "loss": 0.6028,
+      "step": 25759
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.383361058522878e-06,
+      "loss": 0.6162,
+      "step": 25760
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.382673069027614e-06,
+      "loss": 0.5796,
+      "step": 25761
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.381985092831757e-06,
+      "loss": 0.6311,
+      "step": 25762
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3812971299388005e-06,
+      "loss": 0.6135,
+      "step": 25763
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.380609180352239e-06,
+      "loss": 0.5811,
+      "step": 25764
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.379921244075567e-06,
+      "loss": 0.5808,
+      "step": 25765
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.379233321112285e-06,
+      "loss": 0.6425,
+      "step": 25766
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3785454114658896e-06,
+      "loss": 0.5978,
+      "step": 25767
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.377857515139868e-06,
+      "loss": 0.6003,
+      "step": 25768
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.377169632137722e-06,
+      "loss": 0.5915,
+      "step": 25769
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.376481762462943e-06,
+      "loss": 0.6365,
+      "step": 25770
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.375793906119029e-06,
+      "loss": 0.5993,
+      "step": 25771
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3751060631094786e-06,
+      "loss": 0.5802,
+      "step": 25772
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.37441823343778e-06,
+      "loss": 0.6184,
+      "step": 25773
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.373730417107431e-06,
+      "loss": 0.5294,
+      "step": 25774
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3730426141219255e-06,
+      "loss": 0.6682,
+      "step": 25775
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.372354824484762e-06,
+      "loss": 0.5913,
+      "step": 25776
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.371667048199433e-06,
+      "loss": 0.6372,
+      "step": 25777
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.370979285269435e-06,
+      "loss": 0.605,
+      "step": 25778
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.37029153569826e-06,
+      "loss": 0.5826,
+      "step": 25779
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.369603799489402e-06,
+      "loss": 0.6169,
+      "step": 25780
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.36891607664636e-06,
+      "loss": 0.7033,
+      "step": 25781
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.368228367172624e-06,
+      "loss": 0.6763,
+      "step": 25782
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.367540671071692e-06,
+      "loss": 0.6467,
+      "step": 25783
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.366852988347057e-06,
+      "loss": 0.621,
+      "step": 25784
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3661653190022116e-06,
+      "loss": 0.5523,
+      "step": 25785
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.3654776630406525e-06,
+      "loss": 0.5811,
+      "step": 25786
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.364790020465871e-06,
+      "loss": 0.6229,
+      "step": 25787
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.364102391281365e-06,
+      "loss": 0.5959,
+      "step": 25788
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.363414775490629e-06,
+      "loss": 0.5573,
+      "step": 25789
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3627271730971505e-06,
+      "loss": 0.573,
+      "step": 25790
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.362039584104429e-06,
+      "loss": 0.5714,
+      "step": 25791
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3613520085159565e-06,
+      "loss": 0.5833,
+      "step": 25792
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3606644463352265e-06,
+      "loss": 0.5956,
+      "step": 25793
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3599768975657394e-06,
+      "loss": 0.5759,
+      "step": 25794
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3592893622109774e-06,
+      "loss": 0.6498,
+      "step": 25795
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.358601840274439e-06,
+      "loss": 0.5509,
+      "step": 25796
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.357914331759619e-06,
+      "loss": 0.6255,
+      "step": 25797
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3572268366700115e-06,
+      "loss": 0.5735,
+      "step": 25798
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.356539355009109e-06,
+      "loss": 0.6002,
+      "step": 25799
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3558518867804014e-06,
+      "loss": 0.5635,
+      "step": 25800
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.355164431987385e-06,
+      "loss": 0.5632,
+      "step": 25801
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3544769906335555e-06,
+      "loss": 0.5704,
+      "step": 25802
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.353789562722401e-06,
+      "loss": 0.5949,
+      "step": 25803
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.353102148257416e-06,
+      "loss": 0.5834,
+      "step": 25804
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3524147472420985e-06,
+      "loss": 0.6188,
+      "step": 25805
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.351727359679933e-06,
+      "loss": 0.6274,
+      "step": 25806
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.351039985574417e-06,
+      "loss": 0.5687,
+      "step": 25807
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.350352624929043e-06,
+      "loss": 0.567,
+      "step": 25808
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.349665277747302e-06,
+      "loss": 0.5624,
+      "step": 25809
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.348977944032691e-06,
+      "loss": 0.5935,
+      "step": 25810
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.348290623788698e-06,
+      "loss": 0.6093,
+      "step": 25811
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.347603317018814e-06,
+      "loss": 0.6674,
+      "step": 25812
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.346916023726537e-06,
+      "loss": 0.6274,
+      "step": 25813
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.346228743915355e-06,
+      "loss": 0.6022,
+      "step": 25814
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.345541477588763e-06,
+      "loss": 0.5931,
+      "step": 25815
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.344854224750257e-06,
+      "loss": 0.6005,
+      "step": 25816
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.344166985403318e-06,
+      "loss": 0.5539,
+      "step": 25817
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3434797595514464e-06,
+      "loss": 0.5664,
+      "step": 25818
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3427925471981324e-06,
+      "loss": 0.6363,
+      "step": 25819
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.342105348346867e-06,
+      "loss": 0.6415,
+      "step": 25820
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.341418163001144e-06,
+      "loss": 0.6425,
+      "step": 25821
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.340730991164454e-06,
+      "loss": 0.6146,
+      "step": 25822
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.340043832840287e-06,
+      "loss": 0.6172,
+      "step": 25823
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3393566880321356e-06,
+      "loss": 0.5806,
+      "step": 25824
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3386695567434915e-06,
+      "loss": 0.5928,
+      "step": 25825
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.337982438977853e-06,
+      "loss": 0.5849,
+      "step": 25826
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3372953347387e-06,
+      "loss": 0.6324,
+      "step": 25827
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.336608244029529e-06,
+      "loss": 0.5569,
+      "step": 25828
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.335921166853832e-06,
+      "loss": 0.6245,
+      "step": 25829
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3352341032151e-06,
+      "loss": 0.619,
+      "step": 25830
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.334547053116823e-06,
+      "loss": 0.6143,
+      "step": 25831
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.333860016562498e-06,
+      "loss": 0.6417,
+      "step": 25832
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.333172993555606e-06,
+      "loss": 0.5957,
+      "step": 25833
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.332485984099645e-06,
+      "loss": 0.5737,
+      "step": 25834
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.331798988198102e-06,
+      "loss": 0.6061,
+      "step": 25835
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.33111200585447e-06,
+      "loss": 0.6166,
+      "step": 25836
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.330425037072244e-06,
+      "loss": 0.5852,
+      "step": 25837
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.329738081854906e-06,
+      "loss": 0.5898,
+      "step": 25838
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.329051140205952e-06,
+      "loss": 0.572,
+      "step": 25839
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.328364212128871e-06,
+      "loss": 0.5827,
+      "step": 25840
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.327677297627154e-06,
+      "loss": 0.5693,
+      "step": 25841
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.32699039670429e-06,
+      "loss": 0.6262,
+      "step": 25842
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.326303509363775e-06,
+      "loss": 0.5903,
+      "step": 25843
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3256166356090905e-06,
+      "loss": 0.5578,
+      "step": 25844
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.324929775443732e-06,
+      "loss": 0.5899,
+      "step": 25845
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.324242928871188e-06,
+      "loss": 0.5868,
+      "step": 25846
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.323556095894951e-06,
+      "loss": 0.6135,
+      "step": 25847
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.322869276518512e-06,
+      "loss": 0.6019,
+      "step": 25848
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.3221824707453535e-06,
+      "loss": 0.6426,
+      "step": 25849
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.321495678578971e-06,
+      "loss": 0.5791,
+      "step": 25850
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.320808900022852e-06,
+      "loss": 0.5773,
+      "step": 25851
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.320122135080489e-06,
+      "loss": 0.6225,
+      "step": 25852
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.31943538375537e-06,
+      "loss": 0.5827,
+      "step": 25853
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.318748646050989e-06,
+      "loss": 0.5523,
+      "step": 25854
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.318061921970827e-06,
+      "loss": 0.6241,
+      "step": 25855
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.317375211518379e-06,
+      "loss": 0.6213,
+      "step": 25856
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.316688514697132e-06,
+      "loss": 0.5859,
+      "step": 25857
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.316001831510578e-06,
+      "loss": 0.6435,
+      "step": 25858
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.315315161962207e-06,
+      "loss": 0.6349,
+      "step": 25859
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.314628506055501e-06,
+      "loss": 0.5912,
+      "step": 25860
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.313941863793958e-06,
+      "loss": 0.6666,
+      "step": 25861
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.31325523518106e-06,
+      "loss": 0.6074,
+      "step": 25862
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.312568620220299e-06,
+      "loss": 0.6084,
+      "step": 25863
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.31188201891517e-06,
+      "loss": 0.5687,
+      "step": 25864
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.311195431269149e-06,
+      "loss": 0.5932,
+      "step": 25865
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.310508857285733e-06,
+      "loss": 0.6115,
+      "step": 25866
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.309822296968409e-06,
+      "loss": 0.5894,
+      "step": 25867
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.309135750320665e-06,
+      "loss": 0.5797,
+      "step": 25868
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.308449217345991e-06,
+      "loss": 0.6248,
+      "step": 25869
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.307762698047877e-06,
+      "loss": 0.5837,
+      "step": 25870
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.307076192429808e-06,
+      "loss": 0.6251,
+      "step": 25871
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.306389700495273e-06,
+      "loss": 0.5745,
+      "step": 25872
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.30570322224776e-06,
+      "loss": 0.5556,
+      "step": 25873
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.305016757690757e-06,
+      "loss": 0.5979,
+      "step": 25874
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.304330306827758e-06,
+      "loss": 0.6116,
+      "step": 25875
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.303643869662242e-06,
+      "loss": 0.6149,
+      "step": 25876
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.302957446197703e-06,
+      "loss": 0.6269,
+      "step": 25877
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.302271036437627e-06,
+      "loss": 0.6189,
+      "step": 25878
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.301584640385501e-06,
+      "loss": 0.5404,
+      "step": 25879
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.300898258044813e-06,
+      "loss": 0.551,
+      "step": 25880
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.300211889419053e-06,
+      "loss": 0.5908,
+      "step": 25881
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2995255345117085e-06,
+      "loss": 0.5715,
+      "step": 25882
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.298839193326262e-06,
+      "loss": 0.6207,
+      "step": 25883
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.298152865866206e-06,
+      "loss": 0.6278,
+      "step": 25884
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.297466552135026e-06,
+      "loss": 0.6534,
+      "step": 25885
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.296780252136214e-06,
+      "loss": 0.6496,
+      "step": 25886
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.29609396587325e-06,
+      "loss": 0.5908,
+      "step": 25887
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.295407693349623e-06,
+      "loss": 0.5852,
+      "step": 25888
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.294721434568822e-06,
+      "loss": 0.6035,
+      "step": 25889
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.294035189534334e-06,
+      "loss": 0.6306,
+      "step": 25890
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2933489582496495e-06,
+      "loss": 0.5931,
+      "step": 25891
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.292662740718248e-06,
+      "loss": 0.6191,
+      "step": 25892
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.291976536943619e-06,
+      "loss": 0.6331,
+      "step": 25893
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.291290346929251e-06,
+      "loss": 0.6242,
+      "step": 25894
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.29060417067863e-06,
+      "loss": 0.6129,
+      "step": 25895
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.289918008195245e-06,
+      "loss": 0.577,
+      "step": 25896
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.289231859482579e-06,
+      "loss": 0.5857,
+      "step": 25897
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2885457245441204e-06,
+      "loss": 0.6678,
+      "step": 25898
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.287859603383355e-06,
+      "loss": 0.5921,
+      "step": 25899
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2871734960037675e-06,
+      "loss": 0.6094,
+      "step": 25900
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.286487402408846e-06,
+      "loss": 0.5428,
+      "step": 25901
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.28580132260208e-06,
+      "loss": 0.6416,
+      "step": 25902
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.28511525658695e-06,
+      "loss": 0.6426,
+      "step": 25903
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.284429204366944e-06,
+      "loss": 0.5998,
+      "step": 25904
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.283743165945547e-06,
+      "loss": 0.6132,
+      "step": 25905
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.283057141326248e-06,
+      "loss": 0.5786,
+      "step": 25906
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.282371130512531e-06,
+      "loss": 0.5565,
+      "step": 25907
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.281685133507887e-06,
+      "loss": 0.622,
+      "step": 25908
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.280999150315792e-06,
+      "loss": 0.57,
+      "step": 25909
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.280313180939737e-06,
+      "loss": 0.6069,
+      "step": 25910
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.279627225383206e-06,
+      "loss": 0.5677,
+      "step": 25911
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.278941283649688e-06,
+      "loss": 0.6086,
+      "step": 25912
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2782553557426685e-06,
+      "loss": 0.6073,
+      "step": 25913
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2775694416656264e-06,
+      "loss": 0.5632,
+      "step": 25914
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.276883541422053e-06,
+      "loss": 0.5689,
+      "step": 25915
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.27619765501543e-06,
+      "loss": 0.5998,
+      "step": 25916
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2755117824492474e-06,
+      "loss": 0.5986,
+      "step": 25917
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.274825923726987e-06,
+      "loss": 0.6121,
+      "step": 25918
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.274140078852132e-06,
+      "loss": 0.5673,
+      "step": 25919
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2734542478281725e-06,
+      "loss": 0.6648,
+      "step": 25920
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.272768430658588e-06,
+      "loss": 0.5616,
+      "step": 25921
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.272082627346866e-06,
+      "loss": 0.5086,
+      "step": 25922
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.271396837896489e-06,
+      "loss": 0.5957,
+      "step": 25923
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.270711062310951e-06,
+      "loss": 0.5909,
+      "step": 25924
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.270025300593724e-06,
+      "loss": 0.6326,
+      "step": 25925
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.269339552748298e-06,
+      "loss": 0.5826,
+      "step": 25926
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.268653818778156e-06,
+      "loss": 0.5667,
+      "step": 25927
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.267968098686786e-06,
+      "loss": 0.5575,
+      "step": 25928
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.267282392477673e-06,
+      "loss": 0.5664,
+      "step": 25929
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.266596700154296e-06,
+      "loss": 0.5854,
+      "step": 25930
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.26591102172014e-06,
+      "loss": 0.6411,
+      "step": 25931
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.265225357178692e-06,
+      "loss": 0.5994,
+      "step": 25932
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.264539706533434e-06,
+      "loss": 0.5966,
+      "step": 25933
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.263854069787853e-06,
+      "loss": 0.6316,
+      "step": 25934
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.26316844694543e-06,
+      "loss": 0.564,
+      "step": 25935
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.262482838009648e-06,
+      "loss": 0.6481,
+      "step": 25936
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2617972429839935e-06,
+      "loss": 0.6052,
+      "step": 25937
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2611116618719515e-06,
+      "loss": 0.637,
+      "step": 25938
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.260426094677e-06,
+      "loss": 0.59,
+      "step": 25939
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2597405414026275e-06,
+      "loss": 0.6829,
+      "step": 25940
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2590550020523175e-06,
+      "loss": 0.5894,
+      "step": 25941
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2583694766295476e-06,
+      "loss": 0.5394,
+      "step": 25942
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.257683965137807e-06,
+      "loss": 0.6002,
+      "step": 25943
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.256998467580576e-06,
+      "loss": 0.5748,
+      "step": 25944
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.25631298396134e-06,
+      "loss": 0.5947,
+      "step": 25945
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.255627514283585e-06,
+      "loss": 0.6926,
+      "step": 25946
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.254942058550787e-06,
+      "loss": 0.6674,
+      "step": 25947
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.254256616766432e-06,
+      "loss": 0.6065,
+      "step": 25948
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.253571188934003e-06,
+      "loss": 0.6297,
+      "step": 25949
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2528857750569835e-06,
+      "loss": 0.5321,
+      "step": 25950
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.252200375138859e-06,
+      "loss": 0.6514,
+      "step": 25951
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.251514989183106e-06,
+      "loss": 0.5977,
+      "step": 25952
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.25082961719321e-06,
+      "loss": 0.5867,
+      "step": 25953
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.250144259172653e-06,
+      "loss": 0.6027,
+      "step": 25954
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.249458915124922e-06,
+      "loss": 0.6577,
+      "step": 25955
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2487735850534945e-06,
+      "loss": 0.6177,
+      "step": 25956
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.248088268961853e-06,
+      "loss": 0.5882,
+      "step": 25957
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.247402966853481e-06,
+      "loss": 0.5643,
+      "step": 25958
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2467176787318614e-06,
+      "loss": 0.5968,
+      "step": 25959
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.246032404600475e-06,
+      "loss": 0.5749,
+      "step": 25960
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.245347144462804e-06,
+      "loss": 0.6503,
+      "step": 25961
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.244661898322335e-06,
+      "loss": 0.6003,
+      "step": 25962
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.243976666182542e-06,
+      "loss": 0.599,
+      "step": 25963
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.24329144804691e-06,
+      "loss": 0.5899,
+      "step": 25964
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.242606243918921e-06,
+      "loss": 0.5851,
+      "step": 25965
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.241921053802059e-06,
+      "loss": 0.528,
+      "step": 25966
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.241235877699806e-06,
+      "loss": 0.5817,
+      "step": 25967
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.240550715615639e-06,
+      "loss": 0.6102,
+      "step": 25968
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.239865567553041e-06,
+      "loss": 0.6236,
+      "step": 25969
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.239180433515495e-06,
+      "loss": 0.595,
+      "step": 25970
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.238495313506481e-06,
+      "loss": 0.5851,
+      "step": 25971
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.237810207529481e-06,
+      "loss": 0.5742,
+      "step": 25972
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.237125115587982e-06,
+      "loss": 0.6006,
+      "step": 25973
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.236440037685454e-06,
+      "loss": 0.604,
+      "step": 25974
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.235754973825384e-06,
+      "loss": 0.5771,
+      "step": 25975
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.235069924011253e-06,
+      "loss": 0.6398,
+      "step": 25976
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.234384888246542e-06,
+      "loss": 0.606,
+      "step": 25977
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.233699866534731e-06,
+      "loss": 0.6052,
+      "step": 25978
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2330148588793036e-06,
+      "loss": 0.5253,
+      "step": 25979
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.232329865283734e-06,
+      "loss": 0.5869,
+      "step": 25980
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2316448857515076e-06,
+      "loss": 0.5204,
+      "step": 25981
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.230959920286105e-06,
+      "loss": 0.5856,
+      "step": 25982
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.230274968891009e-06,
+      "loss": 0.6196,
+      "step": 25983
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.229590031569693e-06,
+      "loss": 0.5645,
+      "step": 25984
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.228905108325641e-06,
+      "loss": 0.5923,
+      "step": 25985
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2282201991623345e-06,
+      "loss": 0.6013,
+      "step": 25986
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.227535304083253e-06,
+      "loss": 0.6454,
+      "step": 25987
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.226850423091876e-06,
+      "loss": 0.5887,
+      "step": 25988
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.226165556191689e-06,
+      "loss": 0.5338,
+      "step": 25989
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.225480703386163e-06,
+      "loss": 0.6265,
+      "step": 25990
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.224795864678782e-06,
+      "loss": 0.6455,
+      "step": 25991
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2241110400730255e-06,
+      "loss": 0.5861,
+      "step": 25992
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2234262295723735e-06,
+      "loss": 0.6625,
+      "step": 25993
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2227414331803115e-06,
+      "loss": 0.546,
+      "step": 25994
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.222056650900308e-06,
+      "loss": 0.6641,
+      "step": 25995
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.221371882735848e-06,
+      "loss": 0.5979,
+      "step": 25996
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.220687128690415e-06,
+      "loss": 0.6274,
+      "step": 25997
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.220002388767482e-06,
+      "loss": 0.6095,
+      "step": 25998
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.219317662970531e-06,
+      "loss": 0.5707,
+      "step": 25999
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.218632951303044e-06,
+      "loss": 0.5523,
+      "step": 26000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.217948253768494e-06,
+      "loss": 0.6661,
+      "step": 26001
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.217263570370365e-06,
+      "loss": 0.5525,
+      "step": 26002
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2165789011121345e-06,
+      "loss": 0.5951,
+      "step": 26003
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.215894245997281e-06,
+      "loss": 0.6021,
+      "step": 26004
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.215209605029289e-06,
+      "loss": 0.5885,
+      "step": 26005
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.214524978211628e-06,
+      "loss": 0.5848,
+      "step": 26006
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.213840365547781e-06,
+      "loss": 0.5997,
+      "step": 26007
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.213155767041227e-06,
+      "loss": 0.6283,
+      "step": 26008
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.212471182695446e-06,
+      "loss": 0.5477,
+      "step": 26009
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.211786612513918e-06,
+      "loss": 0.5951,
+      "step": 26010
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2111020565001155e-06,
+      "loss": 0.5923,
+      "step": 26011
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.21041751465752e-06,
+      "loss": 0.5997,
+      "step": 26012
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.20973298698961e-06,
+      "loss": 0.6475,
+      "step": 26013
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.209048473499867e-06,
+      "loss": 0.6164,
+      "step": 26014
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.208363974191762e-06,
+      "loss": 0.6213,
+      "step": 26015
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.207679489068779e-06,
+      "loss": 0.6076,
+      "step": 26016
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2069950181343955e-06,
+      "loss": 0.6531,
+      "step": 26017
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.206310561392087e-06,
+      "loss": 0.5769,
+      "step": 26018
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.205626118845332e-06,
+      "loss": 0.6781,
+      "step": 26019
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.204941690497609e-06,
+      "loss": 0.6912,
+      "step": 26020
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2042572763524e-06,
+      "loss": 0.6574,
+      "step": 26021
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.203572876413175e-06,
+      "loss": 0.6293,
+      "step": 26022
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2028884906834155e-06,
+      "loss": 0.602,
+      "step": 26023
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.202204119166597e-06,
+      "loss": 0.633,
+      "step": 26024
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2015197618662e-06,
+      "loss": 0.5745,
+      "step": 26025
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.2008354187857e-06,
+      "loss": 0.6609,
+      "step": 26026
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.200151089928579e-06,
+      "loss": 0.646,
+      "step": 26027
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.199466775298307e-06,
+      "loss": 0.6654,
+      "step": 26028
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.198782474898363e-06,
+      "loss": 0.5897,
+      "step": 26029
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.198098188732227e-06,
+      "loss": 0.5547,
+      "step": 26030
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.197413916803375e-06,
+      "loss": 0.5992,
+      "step": 26031
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1967296591152864e-06,
+      "loss": 0.5371,
+      "step": 26032
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.196045415671433e-06,
+      "loss": 0.5753,
+      "step": 26033
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.195361186475292e-06,
+      "loss": 0.6376,
+      "step": 26034
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1946769715303455e-06,
+      "loss": 0.6769,
+      "step": 26035
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.193992770840065e-06,
+      "loss": 0.6547,
+      "step": 26036
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1933085844079286e-06,
+      "loss": 0.5893,
+      "step": 26037
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.192624412237415e-06,
+      "loss": 0.6319,
+      "step": 26038
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.191940254331999e-06,
+      "loss": 0.6562,
+      "step": 26039
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.191256110695155e-06,
+      "loss": 0.6099,
+      "step": 26040
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1905719813303625e-06,
+      "loss": 0.5786,
+      "step": 26041
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1898878662410954e-06,
+      "loss": 0.6436,
+      "step": 26042
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1892037654308345e-06,
+      "loss": 0.6303,
+      "step": 26043
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.18851967890305e-06,
+      "loss": 0.5898,
+      "step": 26044
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.187835606661219e-06,
+      "loss": 0.5754,
+      "step": 26045
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.18715154870882e-06,
+      "loss": 0.5786,
+      "step": 26046
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.186467505049326e-06,
+      "loss": 0.6249,
+      "step": 26047
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.18578347568622e-06,
+      "loss": 0.5986,
+      "step": 26048
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.185099460622968e-06,
+      "loss": 0.584,
+      "step": 26049
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.18441545986305e-06,
+      "loss": 0.6198,
+      "step": 26050
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1837314734099416e-06,
+      "loss": 0.6179,
+      "step": 26051
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1830475012671185e-06,
+      "loss": 0.5923,
+      "step": 26052
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.182363543438058e-06,
+      "loss": 0.5136,
+      "step": 26053
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1816795999262335e-06,
+      "loss": 0.6374,
+      "step": 26054
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.180995670735118e-06,
+      "loss": 0.62,
+      "step": 26055
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.180311755868192e-06,
+      "loss": 0.582,
+      "step": 26056
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1796278553289255e-06,
+      "loss": 0.5565,
+      "step": 26057
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.178943969120795e-06,
+      "loss": 0.5906,
+      "step": 26058
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.17826009724728e-06,
+      "loss": 0.6422,
+      "step": 26059
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1775762397118494e-06,
+      "loss": 0.6537,
+      "step": 26060
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.176892396517981e-06,
+      "loss": 0.6221,
+      "step": 26061
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.176208567669147e-06,
+      "loss": 0.5908,
+      "step": 26062
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.175524753168825e-06,
+      "loss": 0.6412,
+      "step": 26063
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.17484095302049e-06,
+      "loss": 0.5594,
+      "step": 26064
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.174157167227618e-06,
+      "loss": 0.6541,
+      "step": 26065
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.173473395793678e-06,
+      "loss": 0.6095,
+      "step": 26066
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1727896387221465e-06,
+      "loss": 0.6064,
+      "step": 26067
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.172105896016499e-06,
+      "loss": 0.6427,
+      "step": 26068
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.171422167680211e-06,
+      "loss": 0.5707,
+      "step": 26069
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.170738453716758e-06,
+      "loss": 0.6791,
+      "step": 26070
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.170054754129607e-06,
+      "loss": 0.5848,
+      "step": 26071
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.1693710689222385e-06,
+      "loss": 0.6264,
+      "step": 26072
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.168687398098123e-06,
+      "loss": 0.6689,
+      "step": 26073
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.168003741660739e-06,
+      "loss": 0.6073,
+      "step": 26074
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.167320099613557e-06,
+      "loss": 0.5946,
+      "step": 26075
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.166636471960049e-06,
+      "loss": 0.6138,
+      "step": 26076
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.165952858703693e-06,
+      "loss": 0.6288,
+      "step": 26077
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.165269259847959e-06,
+      "loss": 0.6263,
+      "step": 26078
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.164585675396321e-06,
+      "loss": 0.5439,
+      "step": 26079
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.163902105352254e-06,
+      "loss": 0.5868,
+      "step": 26080
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.163218549719235e-06,
+      "loss": 0.5639,
+      "step": 26081
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.162535008500731e-06,
+      "loss": 0.616,
+      "step": 26082
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.161851481700215e-06,
+      "loss": 0.6004,
+      "step": 26083
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.161167969321165e-06,
+      "loss": 0.5628,
+      "step": 26084
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.160484471367051e-06,
+      "loss": 0.5188,
+      "step": 26085
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.15980098784135e-06,
+      "loss": 0.5982,
+      "step": 26086
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1591175187475295e-06,
+      "loss": 0.6056,
+      "step": 26087
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.158434064089064e-06,
+      "loss": 0.5756,
+      "step": 26088
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.157750623869427e-06,
+      "loss": 0.6089,
+      "step": 26089
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.157067198092091e-06,
+      "loss": 0.5938,
+      "step": 26090
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.156383786760532e-06,
+      "loss": 0.5718,
+      "step": 26091
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.155700389878221e-06,
+      "loss": 0.5785,
+      "step": 26092
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1550170074486256e-06,
+      "loss": 0.6198,
+      "step": 26093
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.154333639475223e-06,
+      "loss": 0.5999,
+      "step": 26094
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.153650285961486e-06,
+      "loss": 0.594,
+      "step": 26095
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.152966946910884e-06,
+      "loss": 0.5689,
+      "step": 26096
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.152283622326892e-06,
+      "loss": 0.5791,
+      "step": 26097
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.151600312212981e-06,
+      "loss": 0.6498,
+      "step": 26098
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.150917016572622e-06,
+      "loss": 0.6303,
+      "step": 26099
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.150233735409286e-06,
+      "loss": 0.6218,
+      "step": 26100
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.14955046872645e-06,
+      "loss": 0.5608,
+      "step": 26101
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.148867216527584e-06,
+      "loss": 0.5659,
+      "step": 26102
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.148183978816157e-06,
+      "loss": 0.5858,
+      "step": 26103
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1475007555956396e-06,
+      "loss": 0.6454,
+      "step": 26104
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.146817546869509e-06,
+      "loss": 0.5773,
+      "step": 26105
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.146134352641232e-06,
+      "loss": 0.6182,
+      "step": 26106
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.145451172914283e-06,
+      "loss": 0.5129,
+      "step": 26107
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.144768007692136e-06,
+      "loss": 0.5679,
+      "step": 26108
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.144084856978256e-06,
+      "loss": 0.6812,
+      "step": 26109
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.143401720776117e-06,
+      "loss": 0.5934,
+      "step": 26110
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.14271859908919e-06,
+      "loss": 0.6494,
+      "step": 26111
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.142035491920949e-06,
+      "loss": 0.6266,
+      "step": 26112
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.141352399274862e-06,
+      "loss": 0.5676,
+      "step": 26113
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1406693211544e-06,
+      "loss": 0.6306,
+      "step": 26114
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.139986257563035e-06,
+      "loss": 0.6573,
+      "step": 26115
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1393032085042355e-06,
+      "loss": 0.6481,
+      "step": 26116
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1386201739814765e-06,
+      "loss": 0.5794,
+      "step": 26117
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.137937153998224e-06,
+      "loss": 0.5737,
+      "step": 26118
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.137254148557955e-06,
+      "loss": 0.5668,
+      "step": 26119
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.136571157664134e-06,
+      "loss": 0.6386,
+      "step": 26120
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.135888181320232e-06,
+      "loss": 0.5451,
+      "step": 26121
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.135205219529722e-06,
+      "loss": 0.5861,
+      "step": 26122
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.134522272296071e-06,
+      "loss": 0.6525,
+      "step": 26123
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.133839339622758e-06,
+      "loss": 0.6028,
+      "step": 26124
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.133156421513241e-06,
+      "loss": 0.5843,
+      "step": 26125
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.132473517970997e-06,
+      "loss": 0.6592,
+      "step": 26126
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.131790628999494e-06,
+      "loss": 0.6467,
+      "step": 26127
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.131107754602203e-06,
+      "loss": 0.5936,
+      "step": 26128
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.130424894782595e-06,
+      "loss": 0.6133,
+      "step": 26129
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.129742049544142e-06,
+      "loss": 0.5926,
+      "step": 26130
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.129059218890304e-06,
+      "loss": 0.6276,
+      "step": 26131
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1283764028245594e-06,
+      "loss": 0.5836,
+      "step": 26132
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.127693601350376e-06,
+      "loss": 0.6475,
+      "step": 26133
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1270108144712215e-06,
+      "loss": 0.6305,
+      "step": 26134
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.126328042190568e-06,
+      "loss": 0.6316,
+      "step": 26135
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1256452845118836e-06,
+      "loss": 0.6457,
+      "step": 26136
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.124962541438635e-06,
+      "loss": 0.5889,
+      "step": 26137
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.124279812974293e-06,
+      "loss": 0.5506,
+      "step": 26138
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.123597099122329e-06,
+      "loss": 0.6431,
+      "step": 26139
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1229143998862136e-06,
+      "loss": 0.5552,
+      "step": 26140
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.12223171526941e-06,
+      "loss": 0.5727,
+      "step": 26141
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.121549045275388e-06,
+      "loss": 0.6297,
+      "step": 26142
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.120866389907619e-06,
+      "loss": 0.541,
+      "step": 26143
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.120183749169571e-06,
+      "loss": 0.6662,
+      "step": 26144
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.119501123064713e-06,
+      "loss": 0.5165,
+      "step": 26145
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.118818511596517e-06,
+      "loss": 0.6337,
+      "step": 26146
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1181359147684445e-06,
+      "loss": 0.6081,
+      "step": 26147
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.117453332583967e-06,
+      "loss": 0.6512,
+      "step": 26148
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.116770765046552e-06,
+      "loss": 0.615,
+      "step": 26149
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.116088212159671e-06,
+      "loss": 0.6281,
+      "step": 26150
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.115405673926794e-06,
+      "loss": 0.6165,
+      "step": 26151
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.114723150351382e-06,
+      "loss": 0.5812,
+      "step": 26152
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.114040641436906e-06,
+      "loss": 0.5898,
+      "step": 26153
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.113358147186837e-06,
+      "loss": 0.5435,
+      "step": 26154
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.112675667604638e-06,
+      "loss": 0.561,
+      "step": 26155
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.111993202693781e-06,
+      "loss": 0.6302,
+      "step": 26156
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.111310752457735e-06,
+      "loss": 0.579,
+      "step": 26157
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1106283168999615e-06,
+      "loss": 0.5865,
+      "step": 26158
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.109945896023931e-06,
+      "loss": 0.562,
+      "step": 26159
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1092634898331135e-06,
+      "loss": 0.5897,
+      "step": 26160
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.108581098330974e-06,
+      "loss": 0.6052,
+      "step": 26161
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.107898721520984e-06,
+      "loss": 0.5669,
+      "step": 26162
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.107216359406604e-06,
+      "loss": 0.6116,
+      "step": 26163
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.106534011991304e-06,
+      "loss": 0.5877,
+      "step": 26164
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.105851679278553e-06,
+      "loss": 0.5895,
+      "step": 26165
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.105169361271817e-06,
+      "loss": 0.5886,
+      "step": 26166
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.1044870579745684e-06,
+      "loss": 0.6436,
+      "step": 26167
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.103804769390263e-06,
+      "loss": 0.6087,
+      "step": 26168
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.103122495522377e-06,
+      "loss": 0.57,
+      "step": 26169
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.102440236374371e-06,
+      "loss": 0.6052,
+      "step": 26170
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.101757991949718e-06,
+      "loss": 0.5989,
+      "step": 26171
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.101075762251879e-06,
+      "loss": 0.5781,
+      "step": 26172
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.100393547284325e-06,
+      "loss": 0.6228,
+      "step": 26173
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.099711347050518e-06,
+      "loss": 0.6101,
+      "step": 26174
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.099029161553929e-06,
+      "loss": 0.5973,
+      "step": 26175
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.098346990798022e-06,
+      "loss": 0.6192,
+      "step": 26176
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.097664834786262e-06,
+      "loss": 0.6305,
+      "step": 26177
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.096982693522123e-06,
+      "loss": 0.568,
+      "step": 26178
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.09630056700906e-06,
+      "loss": 0.6268,
+      "step": 26179
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.095618455250544e-06,
+      "loss": 0.5712,
+      "step": 26180
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.094936358250041e-06,
+      "loss": 0.608,
+      "step": 26181
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.094254276011017e-06,
+      "loss": 0.623,
+      "step": 26182
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0935722085369386e-06,
+      "loss": 0.5868,
+      "step": 26183
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.092890155831276e-06,
+      "loss": 0.6419,
+      "step": 26184
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.092208117897483e-06,
+      "loss": 0.5537,
+      "step": 26185
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.091526094739035e-06,
+      "loss": 0.6516,
+      "step": 26186
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.090844086359394e-06,
+      "loss": 0.6317,
+      "step": 26187
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.090162092762025e-06,
+      "loss": 0.6709,
+      "step": 26188
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0894801139503996e-06,
+      "loss": 0.5917,
+      "step": 26189
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0887981499279735e-06,
+      "loss": 0.6541,
+      "step": 26190
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.088116200698219e-06,
+      "loss": 0.5751,
+      "step": 26191
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0874342662645975e-06,
+      "loss": 0.594,
+      "step": 26192
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.086752346630575e-06,
+      "loss": 0.5973,
+      "step": 26193
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.08607044179962e-06,
+      "loss": 0.5297,
+      "step": 26194
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.08538855177519e-06,
+      "loss": 0.6106,
+      "step": 26195
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.084706676560757e-06,
+      "loss": 0.6174,
+      "step": 26196
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.084024816159782e-06,
+      "loss": 0.6049,
+      "step": 26197
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.083342970575732e-06,
+      "loss": 0.6561,
+      "step": 26198
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.082661139812068e-06,
+      "loss": 0.6498,
+      "step": 26199
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.081979323872263e-06,
+      "loss": 0.6125,
+      "step": 26200
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.081297522759771e-06,
+      "loss": 0.6176,
+      "step": 26201
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.080615736478061e-06,
+      "loss": 0.6518,
+      "step": 26202
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.079933965030599e-06,
+      "loss": 0.557,
+      "step": 26203
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.079252208420846e-06,
+      "loss": 0.6273,
+      "step": 26204
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.078570466652272e-06,
+      "loss": 0.5556,
+      "step": 26205
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0778887397283335e-06,
+      "loss": 0.5673,
+      "step": 26206
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.077207027652497e-06,
+      "loss": 0.5741,
+      "step": 26207
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.076525330428228e-06,
+      "loss": 0.6004,
+      "step": 26208
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.075843648058991e-06,
+      "loss": 0.5805,
+      "step": 26209
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.075161980548251e-06,
+      "loss": 0.6125,
+      "step": 26210
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.074480327899468e-06,
+      "loss": 0.6228,
+      "step": 26211
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0737986901161046e-06,
+      "loss": 0.6699,
+      "step": 26212
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.073117067201631e-06,
+      "loss": 0.6577,
+      "step": 26213
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.072435459159504e-06,
+      "loss": 0.5889,
+      "step": 26214
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.071753865993191e-06,
+      "loss": 0.6111,
+      "step": 26215
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.071072287706158e-06,
+      "loss": 0.63,
+      "step": 26216
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.070390724301859e-06,
+      "loss": 0.6149,
+      "step": 26217
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.069709175783763e-06,
+      "loss": 0.5844,
+      "step": 26218
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.069027642155333e-06,
+      "loss": 0.6102,
+      "step": 26219
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.068346123420031e-06,
+      "loss": 0.62,
+      "step": 26220
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.067664619581323e-06,
+      "loss": 0.5774,
+      "step": 26221
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0669831306426725e-06,
+      "loss": 0.5951,
+      "step": 26222
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0663016566075374e-06,
+      "loss": 0.6587,
+      "step": 26223
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.06562019747938e-06,
+      "loss": 0.6481,
+      "step": 26224
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.064938753261667e-06,
+      "loss": 0.6345,
+      "step": 26225
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.06425732395786e-06,
+      "loss": 0.6527,
+      "step": 26226
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.0635759095714256e-06,
+      "loss": 0.5755,
+      "step": 26227
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.062894510105818e-06,
+      "loss": 0.5894,
+      "step": 26228
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.062213125564503e-06,
+      "loss": 0.5926,
+      "step": 26229
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.061531755950944e-06,
+      "loss": 0.6219,
+      "step": 26230
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.060850401268604e-06,
+      "loss": 0.6266,
+      "step": 26231
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.060169061520945e-06,
+      "loss": 0.6274,
+      "step": 26232
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.059487736711424e-06,
+      "loss": 0.6445,
+      "step": 26233
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0588064268435104e-06,
+      "loss": 0.6223,
+      "step": 26234
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.058125131920661e-06,
+      "loss": 0.6318,
+      "step": 26235
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.057443851946339e-06,
+      "loss": 0.5831,
+      "step": 26236
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.056762586924007e-06,
+      "loss": 0.5565,
+      "step": 26237
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.056081336857129e-06,
+      "loss": 0.5957,
+      "step": 26238
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.055400101749161e-06,
+      "loss": 0.5715,
+      "step": 26239
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.054718881603566e-06,
+      "loss": 0.6327,
+      "step": 26240
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.054037676423808e-06,
+      "loss": 0.6471,
+      "step": 26241
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.053356486213347e-06,
+      "loss": 0.6087,
+      "step": 26242
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.05267531097565e-06,
+      "loss": 0.5907,
+      "step": 26243
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.051994150714168e-06,
+      "loss": 0.608,
+      "step": 26244
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0513130054323665e-06,
+      "loss": 0.5928,
+      "step": 26245
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.050631875133708e-06,
+      "loss": 0.5871,
+      "step": 26246
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0499507598216525e-06,
+      "loss": 0.5947,
+      "step": 26247
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.049269659499663e-06,
+      "loss": 0.6412,
+      "step": 26248
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.048588574171198e-06,
+      "loss": 0.641,
+      "step": 26249
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.047907503839717e-06,
+      "loss": 0.6071,
+      "step": 26250
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.047226448508686e-06,
+      "loss": 0.5971,
+      "step": 26251
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.046545408181559e-06,
+      "loss": 0.5603,
+      "step": 26252
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.045864382861799e-06,
+      "loss": 0.6049,
+      "step": 26253
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.04518337255287e-06,
+      "loss": 0.5799,
+      "step": 26254
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0445023772582295e-06,
+      "loss": 0.6471,
+      "step": 26255
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.043821396981336e-06,
+      "loss": 0.6161,
+      "step": 26256
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0431404317256525e-06,
+      "loss": 0.606,
+      "step": 26257
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.042459481494638e-06,
+      "loss": 0.5818,
+      "step": 26258
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.041778546291756e-06,
+      "loss": 0.6478,
+      "step": 26259
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0410976261204615e-06,
+      "loss": 0.6059,
+      "step": 26260
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.040416720984216e-06,
+      "loss": 0.6076,
+      "step": 26261
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.039735830886478e-06,
+      "loss": 0.6037,
+      "step": 26262
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.039054955830711e-06,
+      "loss": 0.6453,
+      "step": 26263
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0383740958203726e-06,
+      "loss": 0.5929,
+      "step": 26264
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0376932508589256e-06,
+      "loss": 0.6474,
+      "step": 26265
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.037012420949825e-06,
+      "loss": 0.6588,
+      "step": 26266
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.036331606096529e-06,
+      "loss": 0.6226,
+      "step": 26267
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.035650806302503e-06,
+      "loss": 0.661,
+      "step": 26268
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.034970021571203e-06,
+      "loss": 0.6099,
+      "step": 26269
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.034289251906091e-06,
+      "loss": 0.579,
+      "step": 26270
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.03360849731062e-06,
+      "loss": 0.5728,
+      "step": 26271
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.032927757788257e-06,
+      "loss": 0.6296,
+      "step": 26272
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.032247033342455e-06,
+      "loss": 0.6158,
+      "step": 26273
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.031566323976672e-06,
+      "loss": 0.5925,
+      "step": 26274
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0308856296943715e-06,
+      "loss": 0.6008,
+      "step": 26275
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0302049504990155e-06,
+      "loss": 0.6065,
+      "step": 26276
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.029524286394053e-06,
+      "loss": 0.6614,
+      "step": 26277
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.028843637382947e-06,
+      "loss": 0.5777,
+      "step": 26278
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.028163003469158e-06,
+      "loss": 0.6538,
+      "step": 26279
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.027482384656141e-06,
+      "loss": 0.6649,
+      "step": 26280
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0268017809473635e-06,
+      "loss": 0.5328,
+      "step": 26281
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0261211923462715e-06,
+      "loss": 0.6141,
+      "step": 26282
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.025440618856328e-06,
+      "loss": 0.6264,
+      "step": 26283
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.024760060480993e-06,
+      "loss": 0.5446,
+      "step": 26284
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.024079517223722e-06,
+      "loss": 0.5888,
+      "step": 26285
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.02339898908798e-06,
+      "loss": 0.543,
+      "step": 26286
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.022718476077215e-06,
+      "loss": 0.6244,
+      "step": 26287
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.022037978194889e-06,
+      "loss": 0.577,
+      "step": 26288
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.021357495444459e-06,
+      "loss": 0.6011,
+      "step": 26289
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.020677027829388e-06,
+      "loss": 0.579,
+      "step": 26290
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.019996575353128e-06,
+      "loss": 0.6642,
+      "step": 26291
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.019316138019138e-06,
+      "loss": 0.5767,
+      "step": 26292
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.018635715830877e-06,
+      "loss": 0.5973,
+      "step": 26293
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0179553087917976e-06,
+      "loss": 0.5753,
+      "step": 26294
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.017274916905362e-06,
+      "loss": 0.5919,
+      "step": 26295
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.016594540175026e-06,
+      "loss": 0.6159,
+      "step": 26296
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.015914178604251e-06,
+      "loss": 0.5868,
+      "step": 26297
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0152338321964855e-06,
+      "loss": 0.5994,
+      "step": 26298
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.01455350095519e-06,
+      "loss": 0.579,
+      "step": 26299
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.013873184883824e-06,
+      "loss": 0.5818,
+      "step": 26300
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.013192883985842e-06,
+      "loss": 0.6053,
+      "step": 26301
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.012512598264702e-06,
+      "loss": 0.5951,
+      "step": 26302
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.011832327723866e-06,
+      "loss": 0.6098,
+      "step": 26303
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.011152072366779e-06,
+      "loss": 0.6471,
+      "step": 26304
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.010471832196904e-06,
+      "loss": 0.593,
+      "step": 26305
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0097916072176976e-06,
+      "loss": 0.6344,
+      "step": 26306
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.009111397432617e-06,
+      "loss": 0.6061,
+      "step": 26307
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.008431202845119e-06,
+      "loss": 0.5933,
+      "step": 26308
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.007751023458656e-06,
+      "loss": 0.6018,
+      "step": 26309
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.007070859276685e-06,
+      "loss": 0.5592,
+      "step": 26310
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.006390710302666e-06,
+      "loss": 0.565,
+      "step": 26311
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.00571057654005e-06,
+      "loss": 0.5513,
+      "step": 26312
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.005030457992297e-06,
+      "loss": 0.5997,
+      "step": 26313
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.004350354662863e-06,
+      "loss": 0.6343,
+      "step": 26314
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0036702665552e-06,
+      "loss": 0.6911,
+      "step": 26315
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.002990193672765e-06,
+      "loss": 0.5975,
+      "step": 26316
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.002310136019015e-06,
+      "loss": 0.5692,
+      "step": 26317
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.0016300935974045e-06,
+      "loss": 0.5951,
+      "step": 26318
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.000950066411395e-06,
+      "loss": 0.6036,
+      "step": 26319
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.000270054464431e-06,
+      "loss": 0.5574,
+      "step": 26320
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9995900577599754e-06,
+      "loss": 0.5498,
+      "step": 26321
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.998910076301479e-06,
+      "loss": 0.5648,
+      "step": 26322
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.998230110092401e-06,
+      "loss": 0.5324,
+      "step": 26323
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9975501591361985e-06,
+      "loss": 0.6407,
+      "step": 26324
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9968702234363185e-06,
+      "loss": 0.6532,
+      "step": 26325
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.996190302996221e-06,
+      "loss": 0.5793,
+      "step": 26326
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.995510397819359e-06,
+      "loss": 0.5698,
+      "step": 26327
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.994830507909192e-06,
+      "loss": 0.6041,
+      "step": 26328
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.99415063326917e-06,
+      "loss": 0.5893,
+      "step": 26329
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.993470773902749e-06,
+      "loss": 0.5526,
+      "step": 26330
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.992790929813383e-06,
+      "loss": 0.6162,
+      "step": 26331
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.992111101004528e-06,
+      "loss": 0.5998,
+      "step": 26332
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.991431287479635e-06,
+      "loss": 0.5711,
+      "step": 26333
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.990751489242162e-06,
+      "loss": 0.5965,
+      "step": 26334
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.990071706295564e-06,
+      "loss": 0.5577,
+      "step": 26335
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.989391938643292e-06,
+      "loss": 0.5873,
+      "step": 26336
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9887121862888e-06,
+      "loss": 0.6109,
+      "step": 26337
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.988032449235542e-06,
+      "loss": 0.5546,
+      "step": 26338
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.987352727486976e-06,
+      "loss": 0.6529,
+      "step": 26339
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.98667302104655e-06,
+      "loss": 0.5899,
+      "step": 26340
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.985993329917727e-06,
+      "loss": 0.5996,
+      "step": 26341
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.985313654103949e-06,
+      "loss": 0.5698,
+      "step": 26342
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.984633993608677e-06,
+      "loss": 0.6371,
+      "step": 26343
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.983954348435363e-06,
+      "loss": 0.5628,
+      "step": 26344
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9832747185874585e-06,
+      "loss": 0.542,
+      "step": 26345
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.982595104068422e-06,
+      "loss": 0.5842,
+      "step": 26346
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9819155048817015e-06,
+      "loss": 0.5714,
+      "step": 26347
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.981235921030751e-06,
+      "loss": 0.5548,
+      "step": 26348
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.980556352519028e-06,
+      "loss": 0.5239,
+      "step": 26349
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.97987679934998e-06,
+      "loss": 0.5643,
+      "step": 26350
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.979197261527064e-06,
+      "loss": 0.6028,
+      "step": 26351
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.978517739053732e-06,
+      "loss": 0.5746,
+      "step": 26352
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9778382319334335e-06,
+      "loss": 0.558,
+      "step": 26353
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.977158740169624e-06,
+      "loss": 0.5906,
+      "step": 26354
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.976479263765757e-06,
+      "loss": 0.5783,
+      "step": 26355
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.975799802725283e-06,
+      "loss": 0.5573,
+      "step": 26356
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.975120357051661e-06,
+      "loss": 0.6065,
+      "step": 26357
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.974440926748335e-06,
+      "loss": 0.5895,
+      "step": 26358
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.97376151181876e-06,
+      "loss": 0.6474,
+      "step": 26359
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.973082112266388e-06,
+      "loss": 0.5957,
+      "step": 26360
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.972402728094674e-06,
+      "loss": 0.6387,
+      "step": 26361
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.971723359307072e-06,
+      "loss": 0.5905,
+      "step": 26362
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.971044005907026e-06,
+      "loss": 0.5978,
+      "step": 26363
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.970364667897992e-06,
+      "loss": 0.62,
+      "step": 26364
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.969685345283423e-06,
+      "loss": 0.5554,
+      "step": 26365
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9690060380667705e-06,
+      "loss": 0.5889,
+      "step": 26366
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.968326746251487e-06,
+      "loss": 0.599,
+      "step": 26367
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.967647469841023e-06,
+      "loss": 0.5668,
+      "step": 26368
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9669682088388305e-06,
+      "loss": 0.6672,
+      "step": 26369
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.966288963248361e-06,
+      "loss": 0.5562,
+      "step": 26370
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9656097330730645e-06,
+      "loss": 0.6248,
+      "step": 26371
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.964930518316393e-06,
+      "loss": 0.605,
+      "step": 26372
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.964251318981803e-06,
+      "loss": 0.5635,
+      "step": 26373
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9635721350727355e-06,
+      "loss": 0.6175,
+      "step": 26374
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.9628929665926494e-06,
+      "loss": 0.6513,
+      "step": 26375
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 6.962213813544994e-06,
+      "loss": 0.5868,
+      "step": 26376
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.961534675933217e-06,
+      "loss": 0.5791,
+      "step": 26377
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.960855553760779e-06,
+      "loss": 0.606,
+      "step": 26378
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.960176447031117e-06,
+      "loss": 0.5695,
+      "step": 26379
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.959497355747691e-06,
+      "loss": 0.6236,
+      "step": 26380
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9588182799139495e-06,
+      "loss": 0.588,
+      "step": 26381
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.958139219533342e-06,
+      "loss": 0.6098,
+      "step": 26382
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.95746017460932e-06,
+      "loss": 0.5604,
+      "step": 26383
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9567811451453396e-06,
+      "loss": 0.6465,
+      "step": 26384
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.95610213114484e-06,
+      "loss": 0.5945,
+      "step": 26385
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9554231326112766e-06,
+      "loss": 0.6483,
+      "step": 26386
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.954744149548099e-06,
+      "loss": 0.6394,
+      "step": 26387
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.954065181958762e-06,
+      "loss": 0.5983,
+      "step": 26388
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.953386229846711e-06,
+      "loss": 0.6128,
+      "step": 26389
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.952707293215395e-06,
+      "loss": 0.6059,
+      "step": 26390
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.952028372068269e-06,
+      "loss": 0.631,
+      "step": 26391
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.951349466408777e-06,
+      "loss": 0.6046,
+      "step": 26392
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9506705762403695e-06,
+      "loss": 0.616,
+      "step": 26393
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.949991701566499e-06,
+      "loss": 0.6391,
+      "step": 26394
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.949312842390617e-06,
+      "loss": 0.6137,
+      "step": 26395
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.948633998716168e-06,
+      "loss": 0.6352,
+      "step": 26396
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9479551705466005e-06,
+      "loss": 0.5937,
+      "step": 26397
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.947276357885369e-06,
+      "loss": 0.6114,
+      "step": 26398
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.946597560735919e-06,
+      "loss": 0.6029,
+      "step": 26399
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.945918779101706e-06,
+      "loss": 0.6816,
+      "step": 26400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.94524001298617e-06,
+      "loss": 0.5747,
+      "step": 26401
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.944561262392763e-06,
+      "loss": 0.599,
+      "step": 26402
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.943882527324936e-06,
+      "loss": 0.542,
+      "step": 26403
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.943203807786136e-06,
+      "loss": 0.6104,
+      "step": 26404
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.942525103779816e-06,
+      "loss": 0.5874,
+      "step": 26405
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.94184641530942e-06,
+      "loss": 0.6053,
+      "step": 26406
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9411677423783965e-06,
+      "loss": 0.6636,
+      "step": 26407
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.940489084990198e-06,
+      "loss": 0.5908,
+      "step": 26408
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.939810443148268e-06,
+      "loss": 0.6151,
+      "step": 26409
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.939131816856058e-06,
+      "loss": 0.5736,
+      "step": 26410
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.938453206117017e-06,
+      "loss": 0.5805,
+      "step": 26411
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9377746109345915e-06,
+      "loss": 0.651,
+      "step": 26412
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.93709603131223e-06,
+      "loss": 0.5805,
+      "step": 26413
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.936417467253378e-06,
+      "loss": 0.6227,
+      "step": 26414
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.935738918761488e-06,
+      "loss": 0.5935,
+      "step": 26415
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.935060385840009e-06,
+      "loss": 0.5846,
+      "step": 26416
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.934381868492381e-06,
+      "loss": 0.6044,
+      "step": 26417
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.933703366722057e-06,
+      "loss": 0.6033,
+      "step": 26418
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.933024880532485e-06,
+      "loss": 0.5784,
+      "step": 26419
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.932346409927111e-06,
+      "loss": 0.5721,
+      "step": 26420
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.931667954909384e-06,
+      "loss": 0.5942,
+      "step": 26421
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.930989515482756e-06,
+      "loss": 0.6269,
+      "step": 26422
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.930311091650664e-06,
+      "loss": 0.6177,
+      "step": 26423
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.929632683416559e-06,
+      "loss": 0.6097,
+      "step": 26424
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.928954290783889e-06,
+      "loss": 0.5733,
+      "step": 26425
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.928275913756106e-06,
+      "loss": 0.6252,
+      "step": 26426
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9275975523366515e-06,
+      "loss": 0.6019,
+      "step": 26427
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.926919206528972e-06,
+      "loss": 0.5998,
+      "step": 26428
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.926240876336519e-06,
+      "loss": 0.5652,
+      "step": 26429
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.925562561762733e-06,
+      "loss": 0.5774,
+      "step": 26430
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.924884262811063e-06,
+      "loss": 0.5886,
+      "step": 26431
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.924205979484958e-06,
+      "loss": 0.5906,
+      "step": 26432
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.923527711787868e-06,
+      "loss": 0.6152,
+      "step": 26433
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.922849459723229e-06,
+      "loss": 0.6512,
+      "step": 26434
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.922171223294494e-06,
+      "loss": 0.6411,
+      "step": 26435
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.921493002505109e-06,
+      "loss": 0.5159,
+      "step": 26436
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.920814797358519e-06,
+      "loss": 0.6522,
+      "step": 26437
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.920136607858174e-06,
+      "loss": 0.659,
+      "step": 26438
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.919458434007513e-06,
+      "loss": 0.6084,
+      "step": 26439
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9187802758099865e-06,
+      "loss": 0.5781,
+      "step": 26440
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9181021332690404e-06,
+      "loss": 0.5222,
+      "step": 26441
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.917424006388118e-06,
+      "loss": 0.6399,
+      "step": 26442
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.916745895170673e-06,
+      "loss": 0.5809,
+      "step": 26443
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.916067799620141e-06,
+      "loss": 0.6059,
+      "step": 26444
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.915389719739971e-06,
+      "loss": 0.5898,
+      "step": 26445
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.914711655533608e-06,
+      "loss": 0.7301,
+      "step": 26446
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.914033607004502e-06,
+      "loss": 0.5512,
+      "step": 26447
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9133555741560935e-06,
+      "loss": 0.5641,
+      "step": 26448
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.91267755699183e-06,
+      "loss": 0.5555,
+      "step": 26449
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.911999555515157e-06,
+      "loss": 0.5637,
+      "step": 26450
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9113215697295164e-06,
+      "loss": 0.5668,
+      "step": 26451
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.910643599638356e-06,
+      "loss": 0.5886,
+      "step": 26452
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.909965645245119e-06,
+      "loss": 0.576,
+      "step": 26453
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.909287706553256e-06,
+      "loss": 0.6426,
+      "step": 26454
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.908609783566203e-06,
+      "loss": 0.6179,
+      "step": 26455
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9079318762874105e-06,
+      "loss": 0.5682,
+      "step": 26456
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.90725398472032e-06,
+      "loss": 0.5792,
+      "step": 26457
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.906576108868379e-06,
+      "loss": 0.6009,
+      "step": 26458
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.9058982487350304e-06,
+      "loss": 0.6327,
+      "step": 26459
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.905220404323723e-06,
+      "loss": 0.5835,
+      "step": 26460
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.904542575637892e-06,
+      "loss": 0.596,
+      "step": 26461
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.903864762680987e-06,
+      "loss": 0.638,
+      "step": 26462
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.903186965456452e-06,
+      "loss": 0.6022,
+      "step": 26463
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.902509183967734e-06,
+      "loss": 0.5589,
+      "step": 26464
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.901831418218274e-06,
+      "loss": 0.6238,
+      "step": 26465
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.901153668211514e-06,
+      "loss": 0.5953,
+      "step": 26466
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.900475933950898e-06,
+      "loss": 0.5854,
+      "step": 26467
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.899798215439875e-06,
+      "loss": 0.5549,
+      "step": 26468
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8991205126818814e-06,
+      "loss": 0.6694,
+      "step": 26469
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.898442825680368e-06,
+      "loss": 0.5766,
+      "step": 26470
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.897765154438776e-06,
+      "loss": 0.6307,
+      "step": 26471
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8970874989605444e-06,
+      "loss": 0.6313,
+      "step": 26472
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.89640985924912e-06,
+      "loss": 0.5882,
+      "step": 26473
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.895732235307946e-06,
+      "loss": 0.6075,
+      "step": 26474
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.895054627140465e-06,
+      "loss": 0.5811,
+      "step": 26475
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.894377034750126e-06,
+      "loss": 0.5549,
+      "step": 26476
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.893699458140363e-06,
+      "loss": 0.5735,
+      "step": 26477
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.893021897314622e-06,
+      "loss": 0.5979,
+      "step": 26478
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.892344352276347e-06,
+      "loss": 0.5851,
+      "step": 26479
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.89166682302898e-06,
+      "loss": 0.6108,
+      "step": 26480
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.890989309575969e-06,
+      "loss": 0.5603,
+      "step": 26481
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8903118119207465e-06,
+      "loss": 0.5915,
+      "step": 26482
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.889634330066761e-06,
+      "loss": 0.6007,
+      "step": 26483
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8889568640174555e-06,
+      "loss": 0.6028,
+      "step": 26484
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.888279413776272e-06,
+      "loss": 0.6341,
+      "step": 26485
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8876019793466495e-06,
+      "loss": 0.6762,
+      "step": 26486
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.886924560732037e-06,
+      "loss": 0.5611,
+      "step": 26487
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8862471579358716e-06,
+      "loss": 0.6156,
+      "step": 26488
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.885569770961593e-06,
+      "loss": 0.632,
+      "step": 26489
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8848923998126485e-06,
+      "loss": 0.5729,
+      "step": 26490
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.884215044492477e-06,
+      "loss": 0.5737,
+      "step": 26491
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.883537705004526e-06,
+      "loss": 0.6682,
+      "step": 26492
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.882860381352228e-06,
+      "loss": 0.5773,
+      "step": 26493
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.882183073539028e-06,
+      "loss": 0.5553,
+      "step": 26494
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.881505781568372e-06,
+      "loss": 0.5936,
+      "step": 26495
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.880828505443695e-06,
+      "loss": 0.6273,
+      "step": 26496
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.880151245168445e-06,
+      "loss": 0.5925,
+      "step": 26497
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.879474000746063e-06,
+      "loss": 0.5992,
+      "step": 26498
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.878796772179983e-06,
+      "loss": 0.5574,
+      "step": 26499
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.87811955947365e-06,
+      "loss": 0.6124,
+      "step": 26500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8774423626305065e-06,
+      "loss": 0.5883,
+      "step": 26501
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.876765181653994e-06,
+      "loss": 0.634,
+      "step": 26502
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.876088016547555e-06,
+      "loss": 0.6088,
+      "step": 26503
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8754108673146244e-06,
+      "loss": 0.626,
+      "step": 26504
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.874733733958645e-06,
+      "loss": 0.599,
+      "step": 26505
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.874056616483062e-06,
+      "loss": 0.5857,
+      "step": 26506
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.873379514891311e-06,
+      "loss": 0.6411,
+      "step": 26507
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.872702429186835e-06,
+      "loss": 0.6173,
+      "step": 26508
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8720253593730754e-06,
+      "loss": 0.5461,
+      "step": 26509
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.871348305453468e-06,
+      "loss": 0.5939,
+      "step": 26510
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.870671267431457e-06,
+      "loss": 0.6014,
+      "step": 26511
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.86999424531048e-06,
+      "loss": 0.6359,
+      "step": 26512
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.86931723909398e-06,
+      "loss": 0.638,
+      "step": 26513
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.868640248785401e-06,
+      "loss": 0.6618,
+      "step": 26514
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.867963274388173e-06,
+      "loss": 0.5743,
+      "step": 26515
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.8672863159057415e-06,
+      "loss": 0.567,
+      "step": 26516
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.866609373341544e-06,
+      "loss": 0.5594,
+      "step": 26517
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.865932446699024e-06,
+      "loss": 0.5806,
+      "step": 26518
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.865255535981621e-06,
+      "loss": 0.5939,
+      "step": 26519
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.86457864119277e-06,
+      "loss": 0.6379,
+      "step": 26520
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.863901762335914e-06,
+      "loss": 0.5497,
+      "step": 26521
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.863224899414489e-06,
+      "loss": 0.5849,
+      "step": 26522
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 6.862548052431939e-06,
+      "loss": 0.6525,
+      "step": 26523
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.861871221391703e-06,
+      "loss": 0.6083,
+      "step": 26524
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.861194406297218e-06,
+      "loss": 0.5797,
+      "step": 26525
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.860517607151922e-06,
+      "loss": 0.6619,
+      "step": 26526
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.859840823959256e-06,
+      "loss": 0.5121,
+      "step": 26527
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.859164056722658e-06,
+      "loss": 0.5972,
+      "step": 26528
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.858487305445567e-06,
+      "loss": 0.6132,
+      "step": 26529
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.857810570131425e-06,
+      "loss": 0.5904,
+      "step": 26530
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.857133850783664e-06,
+      "loss": 0.6154,
+      "step": 26531
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.856457147405726e-06,
+      "loss": 0.6059,
+      "step": 26532
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8557804600010515e-06,
+      "loss": 0.5817,
+      "step": 26533
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.855103788573077e-06,
+      "loss": 0.5867,
+      "step": 26534
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.854427133125244e-06,
+      "loss": 0.6101,
+      "step": 26535
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.853750493660984e-06,
+      "loss": 0.6012,
+      "step": 26536
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.853073870183739e-06,
+      "loss": 0.6711,
+      "step": 26537
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.852397262696949e-06,
+      "loss": 0.6153,
+      "step": 26538
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8517206712040495e-06,
+      "loss": 0.5525,
+      "step": 26539
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.851044095708479e-06,
+      "loss": 0.6348,
+      "step": 26540
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.85036753621368e-06,
+      "loss": 0.5967,
+      "step": 26541
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.849690992723082e-06,
+      "loss": 0.6314,
+      "step": 26542
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.849014465240126e-06,
+      "loss": 0.6345,
+      "step": 26543
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.848337953768253e-06,
+      "loss": 0.5899,
+      "step": 26544
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.847661458310894e-06,
+      "loss": 0.6378,
+      "step": 26545
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.846984978871496e-06,
+      "loss": 0.5686,
+      "step": 26546
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.846308515453486e-06,
+      "loss": 0.5446,
+      "step": 26547
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.845632068060308e-06,
+      "loss": 0.6187,
+      "step": 26548
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.844955636695396e-06,
+      "loss": 0.6032,
+      "step": 26549
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8442792213621885e-06,
+      "loss": 0.6547,
+      "step": 26550
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.843602822064122e-06,
+      "loss": 0.6314,
+      "step": 26551
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.842926438804639e-06,
+      "loss": 0.5884,
+      "step": 26552
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.842250071587166e-06,
+      "loss": 0.5929,
+      "step": 26553
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.841573720415146e-06,
+      "loss": 0.5955,
+      "step": 26554
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.840897385292016e-06,
+      "loss": 0.551,
+      "step": 26555
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.840221066221209e-06,
+      "loss": 0.6008,
+      "step": 26556
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.83954476320617e-06,
+      "loss": 0.5825,
+      "step": 26557
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.838868476250324e-06,
+      "loss": 0.6088,
+      "step": 26558
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.838192205357115e-06,
+      "loss": 0.6183,
+      "step": 26559
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.837515950529976e-06,
+      "loss": 0.678,
+      "step": 26560
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.836839711772345e-06,
+      "loss": 0.6238,
+      "step": 26561
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.836163489087663e-06,
+      "loss": 0.6283,
+      "step": 26562
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.835487282479356e-06,
+      "loss": 0.5476,
+      "step": 26563
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.834811091950864e-06,
+      "loss": 0.6561,
+      "step": 26564
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.834134917505626e-06,
+      "loss": 0.5901,
+      "step": 26565
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8334587591470745e-06,
+      "loss": 0.5699,
+      "step": 26566
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8327826168786456e-06,
+      "loss": 0.5623,
+      "step": 26567
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.832106490703779e-06,
+      "loss": 0.5926,
+      "step": 26568
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8314303806259065e-06,
+      "loss": 0.5945,
+      "step": 26569
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8307542866484625e-06,
+      "loss": 0.6774,
+      "step": 26570
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.830078208774884e-06,
+      "loss": 0.57,
+      "step": 26571
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8294021470086055e-06,
+      "loss": 0.5741,
+      "step": 26572
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.828726101353068e-06,
+      "loss": 0.5962,
+      "step": 26573
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.828050071811698e-06,
+      "loss": 0.5891,
+      "step": 26574
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.827374058387934e-06,
+      "loss": 0.6332,
+      "step": 26575
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.826698061085212e-06,
+      "loss": 0.601,
+      "step": 26576
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.826022079906966e-06,
+      "loss": 0.5437,
+      "step": 26577
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.825346114856632e-06,
+      "loss": 0.6046,
+      "step": 26578
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.824670165937648e-06,
+      "loss": 0.6038,
+      "step": 26579
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.823994233153442e-06,
+      "loss": 0.6136,
+      "step": 26580
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.823318316507449e-06,
+      "loss": 0.6699,
+      "step": 26581
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.822642416003107e-06,
+      "loss": 0.6224,
+      "step": 26582
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.821966531643851e-06,
+      "loss": 0.5998,
+      "step": 26583
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.821290663433116e-06,
+      "loss": 0.5438,
+      "step": 26584
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8206148113743284e-06,
+      "loss": 0.5898,
+      "step": 26585
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.819938975470932e-06,
+      "loss": 0.6317,
+      "step": 26586
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8192631557263565e-06,
+      "loss": 0.6184,
+      "step": 26587
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.818587352144035e-06,
+      "loss": 0.6043,
+      "step": 26588
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.817911564727402e-06,
+      "loss": 0.597,
+      "step": 26589
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.817235793479897e-06,
+      "loss": 0.5606,
+      "step": 26590
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8165600384049455e-06,
+      "loss": 0.5963,
+      "step": 26591
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.815884299505984e-06,
+      "loss": 0.5983,
+      "step": 26592
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.815208576786447e-06,
+      "loss": 0.561,
+      "step": 26593
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.814532870249769e-06,
+      "loss": 0.5665,
+      "step": 26594
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.813857179899385e-06,
+      "loss": 0.6611,
+      "step": 26595
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.813181505738723e-06,
+      "loss": 0.5566,
+      "step": 26596
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.812505847771218e-06,
+      "loss": 0.5713,
+      "step": 26597
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.811830206000305e-06,
+      "loss": 0.578,
+      "step": 26598
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.811154580429416e-06,
+      "loss": 0.6413,
+      "step": 26599
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.810478971061989e-06,
+      "loss": 0.5761,
+      "step": 26600
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.809803377901448e-06,
+      "loss": 0.561,
+      "step": 26601
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.8091278009512305e-06,
+      "loss": 0.5988,
+      "step": 26602
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.808452240214769e-06,
+      "loss": 0.6131,
+      "step": 26603
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.807776695695498e-06,
+      "loss": 0.6526,
+      "step": 26604
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.807101167396848e-06,
+      "loss": 0.6467,
+      "step": 26605
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.806425655322253e-06,
+      "loss": 0.5789,
+      "step": 26606
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.805750159475144e-06,
+      "loss": 0.6064,
+      "step": 26607
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.805074679858953e-06,
+      "loss": 0.5663,
+      "step": 26608
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.804399216477113e-06,
+      "loss": 0.6443,
+      "step": 26609
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.803723769333055e-06,
+      "loss": 0.6701,
+      "step": 26610
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.803048338430218e-06,
+      "loss": 0.5615,
+      "step": 26611
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.802372923772025e-06,
+      "loss": 0.5723,
+      "step": 26612
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.80169752536191e-06,
+      "loss": 0.5927,
+      "step": 26613
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.801022143203308e-06,
+      "loss": 0.6106,
+      "step": 26614
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.800346777299649e-06,
+      "loss": 0.5981,
+      "step": 26615
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.799671427654364e-06,
+      "loss": 0.5468,
+      "step": 26616
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.798996094270891e-06,
+      "loss": 0.5772,
+      "step": 26617
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.798320777152652e-06,
+      "loss": 0.5676,
+      "step": 26618
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.797645476303083e-06,
+      "loss": 0.5557,
+      "step": 26619
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.796970191725616e-06,
+      "loss": 0.6556,
+      "step": 26620
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.796294923423682e-06,
+      "loss": 0.6729,
+      "step": 26621
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.795619671400713e-06,
+      "loss": 0.6139,
+      "step": 26622
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7949444356601354e-06,
+      "loss": 0.6209,
+      "step": 26623
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.794269216205385e-06,
+      "loss": 0.624,
+      "step": 26624
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7935940130398925e-06,
+      "loss": 0.5654,
+      "step": 26625
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.792918826167087e-06,
+      "loss": 0.6172,
+      "step": 26626
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.792243655590401e-06,
+      "loss": 0.5836,
+      "step": 26627
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.791568501313266e-06,
+      "loss": 0.5771,
+      "step": 26628
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.790893363339107e-06,
+      "loss": 0.5979,
+      "step": 26629
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.79021824167136e-06,
+      "loss": 0.5889,
+      "step": 26630
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7895431363134544e-06,
+      "loss": 0.6233,
+      "step": 26631
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.78886804726882e-06,
+      "loss": 0.6611,
+      "step": 26632
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7881929745408905e-06,
+      "loss": 0.5638,
+      "step": 26633
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.787517918133091e-06,
+      "loss": 0.5937,
+      "step": 26634
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.786842878048852e-06,
+      "loss": 0.5596,
+      "step": 26635
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7861678542916055e-06,
+      "loss": 0.6224,
+      "step": 26636
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.785492846864783e-06,
+      "loss": 0.538,
+      "step": 26637
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.784817855771815e-06,
+      "loss": 0.5826,
+      "step": 26638
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.784142881016126e-06,
+      "loss": 0.6268,
+      "step": 26639
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.783467922601148e-06,
+      "loss": 0.5875,
+      "step": 26640
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.782792980530313e-06,
+      "loss": 0.622,
+      "step": 26641
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.78211805480705e-06,
+      "loss": 0.6258,
+      "step": 26642
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7814431454347855e-06,
+      "loss": 0.6282,
+      "step": 26643
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.780768252416954e-06,
+      "loss": 0.5797,
+      "step": 26644
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.78009337575698e-06,
+      "loss": 0.6392,
+      "step": 26645
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.779418515458293e-06,
+      "loss": 0.6653,
+      "step": 26646
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.778743671524325e-06,
+      "loss": 0.5991,
+      "step": 26647
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.778068843958503e-06,
+      "loss": 0.5503,
+      "step": 26648
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.777394032764262e-06,
+      "loss": 0.5992,
+      "step": 26649
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7767192379450195e-06,
+      "loss": 0.6147,
+      "step": 26650
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.776044459504212e-06,
+      "loss": 0.5834,
+      "step": 26651
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.775369697445267e-06,
+      "loss": 0.6223,
+      "step": 26652
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.774694951771612e-06,
+      "loss": 0.5751,
+      "step": 26653
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.77402022248668e-06,
+      "loss": 0.5614,
+      "step": 26654
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.773345509593894e-06,
+      "loss": 0.6139,
+      "step": 26655
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.772670813096682e-06,
+      "loss": 0.5772,
+      "step": 26656
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.771996132998476e-06,
+      "loss": 0.639,
+      "step": 26657
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.771321469302703e-06,
+      "loss": 0.692,
+      "step": 26658
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.770646822012792e-06,
+      "loss": 0.6124,
+      "step": 26659
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.769972191132173e-06,
+      "loss": 0.5792,
+      "step": 26660
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.769297576664269e-06,
+      "loss": 0.5945,
+      "step": 26661
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.768622978612508e-06,
+      "loss": 0.646,
+      "step": 26662
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.767948396980323e-06,
+      "loss": 0.5818,
+      "step": 26663
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.767273831771137e-06,
+      "loss": 0.5939,
+      "step": 26664
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.766599282988381e-06,
+      "loss": 0.5801,
+      "step": 26665
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.765924750635481e-06,
+      "loss": 0.5942,
+      "step": 26666
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.765250234715862e-06,
+      "loss": 0.6473,
+      "step": 26667
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.7645757352329546e-06,
+      "loss": 0.5768,
+      "step": 26668
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.763901252190186e-06,
+      "loss": 0.5828,
+      "step": 26669
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.763226785590982e-06,
+      "loss": 0.5554,
+      "step": 26670
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.762552335438775e-06,
+      "loss": 0.5858,
+      "step": 26671
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.761877901736983e-06,
+      "loss": 0.591,
+      "step": 26672
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.761203484489039e-06,
+      "loss": 0.636,
+      "step": 26673
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7605290836983664e-06,
+      "loss": 0.598,
+      "step": 26674
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.759854699368395e-06,
+      "loss": 0.6316,
+      "step": 26675
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.759180331502557e-06,
+      "loss": 0.6368,
+      "step": 26676
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.758505980104266e-06,
+      "loss": 0.5637,
+      "step": 26677
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.757831645176957e-06,
+      "loss": 0.4968,
+      "step": 26678
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7571573267240555e-06,
+      "loss": 0.5804,
+      "step": 26679
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.756483024748988e-06,
+      "loss": 0.6376,
+      "step": 26680
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.755808739255178e-06,
+      "loss": 0.575,
+      "step": 26681
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.755134470246058e-06,
+      "loss": 0.5842,
+      "step": 26682
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.754460217725046e-06,
+      "loss": 0.6352,
+      "step": 26683
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.753785981695575e-06,
+      "loss": 0.665,
+      "step": 26684
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7531117621610666e-06,
+      "loss": 0.5921,
+      "step": 26685
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.752437559124947e-06,
+      "loss": 0.6092,
+      "step": 26686
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.751763372590647e-06,
+      "loss": 0.5692,
+      "step": 26687
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.751089202561586e-06,
+      "loss": 0.5903,
+      "step": 26688
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7504150490411925e-06,
+      "loss": 0.606,
+      "step": 26689
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7497409120328896e-06,
+      "loss": 0.6354,
+      "step": 26690
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.749066791540106e-06,
+      "loss": 0.5419,
+      "step": 26691
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.748392687566272e-06,
+      "loss": 0.5952,
+      "step": 26692
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.747718600114801e-06,
+      "loss": 0.5539,
+      "step": 26693
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.747044529189124e-06,
+      "loss": 0.5936,
+      "step": 26694
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7463704747926676e-06,
+      "loss": 0.5404,
+      "step": 26695
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7456964369288545e-06,
+      "loss": 0.5596,
+      "step": 26696
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7450224156011125e-06,
+      "loss": 0.6134,
+      "step": 26697
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.744348410812869e-06,
+      "loss": 0.5808,
+      "step": 26698
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.743674422567538e-06,
+      "loss": 0.5316,
+      "step": 26699
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.743000450868553e-06,
+      "loss": 0.6037,
+      "step": 26700
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7423264957193375e-06,
+      "loss": 0.6215,
+      "step": 26701
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.741652557123315e-06,
+      "loss": 0.6048,
+      "step": 26702
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.740978635083911e-06,
+      "loss": 0.5502,
+      "step": 26703
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.740304729604548e-06,
+      "loss": 0.6236,
+      "step": 26704
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.739630840688651e-06,
+      "loss": 0.6472,
+      "step": 26705
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.738956968339644e-06,
+      "loss": 0.5987,
+      "step": 26706
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.738283112560952e-06,
+      "loss": 0.66,
+      "step": 26707
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.737609273355998e-06,
+      "loss": 0.5832,
+      "step": 26708
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.736935450728211e-06,
+      "loss": 0.68,
+      "step": 26709
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7362616446810056e-06,
+      "loss": 0.5885,
+      "step": 26710
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.735587855217812e-06,
+      "loss": 0.6103,
+      "step": 26711
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.734914082342052e-06,
+      "loss": 0.6131,
+      "step": 26712
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.73424032605715e-06,
+      "loss": 0.6499,
+      "step": 26713
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.733566586366533e-06,
+      "loss": 0.6324,
+      "step": 26714
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.732892863273617e-06,
+      "loss": 0.571,
+      "step": 26715
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.732219156781829e-06,
+      "loss": 0.6033,
+      "step": 26716
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.731545466894593e-06,
+      "loss": 0.5505,
+      "step": 26717
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.730871793615331e-06,
+      "loss": 0.6139,
+      "step": 26718
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.730198136947472e-06,
+      "loss": 0.6018,
+      "step": 26719
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.729524496894428e-06,
+      "loss": 0.5855,
+      "step": 26720
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.728850873459629e-06,
+      "loss": 0.5549,
+      "step": 26721
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7281772666464985e-06,
+      "loss": 0.5769,
+      "step": 26722
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7275036764584555e-06,
+      "loss": 0.5922,
+      "step": 26723
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7268301028989246e-06,
+      "loss": 0.5658,
+      "step": 26724
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.726156545971334e-06,
+      "loss": 0.5856,
+      "step": 26725
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.725483005679095e-06,
+      "loss": 0.6014,
+      "step": 26726
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.724809482025636e-06,
+      "loss": 0.579,
+      "step": 26727
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7241359750143795e-06,
+      "loss": 0.607,
+      "step": 26728
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.723462484648748e-06,
+      "loss": 0.6564,
+      "step": 26729
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.722789010932167e-06,
+      "loss": 0.6535,
+      "step": 26730
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.722115553868051e-06,
+      "loss": 0.6181,
+      "step": 26731
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.721442113459824e-06,
+      "loss": 0.5855,
+      "step": 26732
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7207686897109125e-06,
+      "loss": 0.5614,
+      "step": 26733
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.720095282624733e-06,
+      "loss": 0.6338,
+      "step": 26734
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7194218922047116e-06,
+      "loss": 0.5862,
+      "step": 26735
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.718748518454272e-06,
+      "loss": 0.6041,
+      "step": 26736
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.718075161376829e-06,
+      "loss": 0.6049,
+      "step": 26737
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.717401820975806e-06,
+      "loss": 0.5583,
+      "step": 26738
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.716728497254626e-06,
+      "loss": 0.6816,
+      "step": 26739
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.716055190216713e-06,
+      "loss": 0.5908,
+      "step": 26740
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.715381899865483e-06,
+      "loss": 0.6209,
+      "step": 26741
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.71470862620436e-06,
+      "loss": 0.6123,
+      "step": 26742
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7140353692367655e-06,
+      "loss": 0.5911,
+      "step": 26743
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.713362128966118e-06,
+      "loss": 0.689,
+      "step": 26744
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.712688905395839e-06,
+      "loss": 0.56,
+      "step": 26745
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.712015698529355e-06,
+      "loss": 0.6263,
+      "step": 26746
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.711342508370079e-06,
+      "loss": 0.6138,
+      "step": 26747
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.710669334921434e-06,
+      "loss": 0.6252,
+      "step": 26748
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7099961781868415e-06,
+      "loss": 0.6199,
+      "step": 26749
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.709323038169722e-06,
+      "loss": 0.6024,
+      "step": 26750
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.708649914873496e-06,
+      "loss": 0.6073,
+      "step": 26751
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.707976808301588e-06,
+      "loss": 0.5929,
+      "step": 26752
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.70730371845741e-06,
+      "loss": 0.5791,
+      "step": 26753
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.706630645344385e-06,
+      "loss": 0.6623,
+      "step": 26754
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.705957588965935e-06,
+      "loss": 0.5815,
+      "step": 26755
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.705284549325479e-06,
+      "loss": 0.6402,
+      "step": 26756
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.704611526426441e-06,
+      "loss": 0.5437,
+      "step": 26757
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.703938520272233e-06,
+      "loss": 0.5791,
+      "step": 26758
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.703265530866278e-06,
+      "loss": 0.5493,
+      "step": 26759
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.702592558211997e-06,
+      "loss": 0.548,
+      "step": 26760
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.701919602312812e-06,
+      "loss": 0.6136,
+      "step": 26761
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7012466631721364e-06,
+      "loss": 0.5806,
+      "step": 26762
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.7005737407933945e-06,
+      "loss": 0.5829,
+      "step": 26763
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.699900835180004e-06,
+      "loss": 0.5761,
+      "step": 26764
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.699227946335382e-06,
+      "loss": 0.5671,
+      "step": 26765
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.698555074262949e-06,
+      "loss": 0.6383,
+      "step": 26766
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.697882218966125e-06,
+      "loss": 0.5971,
+      "step": 26767
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.697209380448333e-06,
+      "loss": 0.6144,
+      "step": 26768
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.696536558712982e-06,
+      "loss": 0.6132,
+      "step": 26769
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.695863753763497e-06,
+      "loss": 0.6092,
+      "step": 26770
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.695190965603295e-06,
+      "loss": 0.6132,
+      "step": 26771
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.694518194235798e-06,
+      "loss": 0.5664,
+      "step": 26772
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.693845439664419e-06,
+      "loss": 0.6173,
+      "step": 26773
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.693172701892585e-06,
+      "loss": 0.6365,
+      "step": 26774
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.692499980923706e-06,
+      "loss": 0.6102,
+      "step": 26775
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.691827276761202e-06,
+      "loss": 0.564,
+      "step": 26776
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6911545894084926e-06,
+      "loss": 0.5979,
+      "step": 26777
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.690481918868997e-06,
+      "loss": 0.5973,
+      "step": 26778
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6898092651461334e-06,
+      "loss": 0.5734,
+      "step": 26779
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.689136628243316e-06,
+      "loss": 0.6039,
+      "step": 26780
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.688464008163966e-06,
+      "loss": 0.5706,
+      "step": 26781
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.687791404911499e-06,
+      "loss": 0.6532,
+      "step": 26782
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6871188184893345e-06,
+      "loss": 0.5772,
+      "step": 26783
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.686446248900891e-06,
+      "loss": 0.6191,
+      "step": 26784
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.685773696149584e-06,
+      "loss": 0.6027,
+      "step": 26785
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.685101160238829e-06,
+      "loss": 0.5447,
+      "step": 26786
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6844286411720476e-06,
+      "loss": 0.5739,
+      "step": 26787
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.683756138952655e-06,
+      "loss": 0.6061,
+      "step": 26788
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6830836535840685e-06,
+      "loss": 0.6052,
+      "step": 26789
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.682411185069708e-06,
+      "loss": 0.6177,
+      "step": 26790
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.681738733412986e-06,
+      "loss": 0.5802,
+      "step": 26791
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.681066298617319e-06,
+      "loss": 0.6155,
+      "step": 26792
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.680393880686127e-06,
+      "loss": 0.6008,
+      "step": 26793
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.679721479622827e-06,
+      "loss": 0.5969,
+      "step": 26794
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.679049095430837e-06,
+      "loss": 0.6366,
+      "step": 26795
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.678376728113568e-06,
+      "loss": 0.5583,
+      "step": 26796
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.677704377674439e-06,
+      "loss": 0.6261,
+      "step": 26797
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.677032044116868e-06,
+      "loss": 0.5959,
+      "step": 26798
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.676359727444272e-06,
+      "loss": 0.5858,
+      "step": 26799
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.675687427660063e-06,
+      "loss": 0.5626,
+      "step": 26800
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.675015144767663e-06,
+      "loss": 0.6292,
+      "step": 26801
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.674342878770484e-06,
+      "loss": 0.5831,
+      "step": 26802
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.67367062967194e-06,
+      "loss": 0.6031,
+      "step": 26803
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.67299839747545e-06,
+      "loss": 0.5619,
+      "step": 26804
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6723261821844295e-06,
+      "loss": 0.5943,
+      "step": 26805
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6716539838023e-06,
+      "loss": 0.6394,
+      "step": 26806
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6709818023324655e-06,
+      "loss": 0.6623,
+      "step": 26807
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.670309637778348e-06,
+      "loss": 0.602,
+      "step": 26808
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.669637490143363e-06,
+      "loss": 0.6568,
+      "step": 26809
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.668965359430923e-06,
+      "loss": 0.5863,
+      "step": 26810
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6682932456444524e-06,
+      "loss": 0.62,
+      "step": 26811
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.667621148787354e-06,
+      "loss": 0.5827,
+      "step": 26812
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.666949068863049e-06,
+      "loss": 0.6592,
+      "step": 26813
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6662770058749515e-06,
+      "loss": 0.5909,
+      "step": 26814
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.665604959826477e-06,
+      "loss": 0.601,
+      "step": 26815
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.6649329307210395e-06,
+      "loss": 0.5906,
+      "step": 26816
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 6.66426091856206e-06,
+      "loss": 0.6295,
+      "step": 26817
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.663588923352944e-06,
+      "loss": 0.5671,
+      "step": 26818
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.662916945097107e-06,
+      "loss": 0.5592,
+      "step": 26819
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.662244983797971e-06,
+      "loss": 0.6292,
+      "step": 26820
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.661573039458943e-06,
+      "loss": 0.6558,
+      "step": 26821
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.660901112083442e-06,
+      "loss": 0.6477,
+      "step": 26822
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.660229201674879e-06,
+      "loss": 0.623,
+      "step": 26823
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.659557308236671e-06,
+      "loss": 0.5641,
+      "step": 26824
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6588854317722265e-06,
+      "loss": 0.5929,
+      "step": 26825
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.658213572284966e-06,
+      "loss": 0.5749,
+      "step": 26826
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.657541729778301e-06,
+      "loss": 0.6152,
+      "step": 26827
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.656869904255649e-06,
+      "loss": 0.6198,
+      "step": 26828
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.656198095720415e-06,
+      "loss": 0.5529,
+      "step": 26829
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.655526304176019e-06,
+      "loss": 0.6579,
+      "step": 26830
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.654854529625874e-06,
+      "loss": 0.6288,
+      "step": 26831
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.654182772073391e-06,
+      "loss": 0.5842,
+      "step": 26832
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.653511031521991e-06,
+      "loss": 0.5515,
+      "step": 26833
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.652839307975076e-06,
+      "loss": 0.5726,
+      "step": 26834
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6521676014360645e-06,
+      "loss": 0.5754,
+      "step": 26835
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.651495911908371e-06,
+      "loss": 0.5916,
+      "step": 26836
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.650824239395409e-06,
+      "loss": 0.6145,
+      "step": 26837
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.650152583900591e-06,
+      "loss": 0.6322,
+      "step": 26838
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.649480945427325e-06,
+      "loss": 0.6192,
+      "step": 26839
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6488093239790284e-06,
+      "loss": 0.52,
+      "step": 26840
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6481377195591155e-06,
+      "loss": 0.5506,
+      "step": 26841
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.647466132170993e-06,
+      "loss": 0.6386,
+      "step": 26842
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6467945618180785e-06,
+      "loss": 0.6061,
+      "step": 26843
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.646123008503787e-06,
+      "loss": 0.612,
+      "step": 26844
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6454514722315215e-06,
+      "loss": 0.6202,
+      "step": 26845
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6447799530047e-06,
+      "loss": 0.5637,
+      "step": 26846
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.644108450826734e-06,
+      "loss": 0.5965,
+      "step": 26847
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.643436965701036e-06,
+      "loss": 0.6083,
+      "step": 26848
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.642765497631021e-06,
+      "loss": 0.5865,
+      "step": 26849
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.642094046620096e-06,
+      "loss": 0.6059,
+      "step": 26850
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.641422612671672e-06,
+      "loss": 0.6178,
+      "step": 26851
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.640751195789164e-06,
+      "loss": 0.5637,
+      "step": 26852
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.640079795975983e-06,
+      "loss": 0.5411,
+      "step": 26853
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6394084132355415e-06,
+      "loss": 0.6112,
+      "step": 26854
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.638737047571253e-06,
+      "loss": 0.6565,
+      "step": 26855
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.638065698986522e-06,
+      "loss": 0.5462,
+      "step": 26856
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.637394367484764e-06,
+      "loss": 0.6176,
+      "step": 26857
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.636723053069392e-06,
+      "loss": 0.6349,
+      "step": 26858
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.636051755743814e-06,
+      "loss": 0.6131,
+      "step": 26859
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.635380475511444e-06,
+      "loss": 0.6136,
+      "step": 26860
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.634709212375688e-06,
+      "loss": 0.6162,
+      "step": 26861
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.634037966339964e-06,
+      "loss": 0.6166,
+      "step": 26862
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.633366737407673e-06,
+      "loss": 0.6098,
+      "step": 26863
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.632695525582236e-06,
+      "loss": 0.6323,
+      "step": 26864
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.632024330867057e-06,
+      "loss": 0.5885,
+      "step": 26865
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.631353153265554e-06,
+      "loss": 0.6331,
+      "step": 26866
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.630681992781128e-06,
+      "loss": 0.5639,
+      "step": 26867
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.630010849417191e-06,
+      "loss": 0.6338,
+      "step": 26868
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.629339723177157e-06,
+      "loss": 0.5867,
+      "step": 26869
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.628668614064436e-06,
+      "loss": 0.5952,
+      "step": 26870
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.627997522082441e-06,
+      "loss": 0.5768,
+      "step": 26871
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.627326447234575e-06,
+      "loss": 0.6393,
+      "step": 26872
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.626655389524249e-06,
+      "loss": 0.6559,
+      "step": 26873
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.625984348954876e-06,
+      "loss": 0.5814,
+      "step": 26874
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.625313325529864e-06,
+      "loss": 0.6116,
+      "step": 26875
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6246423192526275e-06,
+      "loss": 0.591,
+      "step": 26876
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.623971330126568e-06,
+      "loss": 0.5819,
+      "step": 26877
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.623300358155101e-06,
+      "loss": 0.5126,
+      "step": 26878
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.622629403341633e-06,
+      "loss": 0.5734,
+      "step": 26879
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.621958465689573e-06,
+      "loss": 0.538,
+      "step": 26880
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.62128754520233e-06,
+      "loss": 0.6364,
+      "step": 26881
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.620616641883322e-06,
+      "loss": 0.6271,
+      "step": 26882
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.619945755735944e-06,
+      "loss": 0.6469,
+      "step": 26883
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.619274886763611e-06,
+      "loss": 0.5456,
+      "step": 26884
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.618604034969734e-06,
+      "loss": 0.6213,
+      "step": 26885
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.617933200357719e-06,
+      "loss": 0.6117,
+      "step": 26886
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.61726238293098e-06,
+      "loss": 0.5967,
+      "step": 26887
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6165915826929175e-06,
+      "loss": 0.598,
+      "step": 26888
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6159207996469445e-06,
+      "loss": 0.5716,
+      "step": 26889
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.61525003379647e-06,
+      "loss": 0.5796,
+      "step": 26890
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6145792851449e-06,
+      "loss": 0.5936,
+      "step": 26891
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.613908553695644e-06,
+      "loss": 0.69,
+      "step": 26892
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.613237839452114e-06,
+      "loss": 0.5803,
+      "step": 26893
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.612567142417712e-06,
+      "loss": 0.5861,
+      "step": 26894
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.611896462595847e-06,
+      "loss": 0.619,
+      "step": 26895
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.611225799989929e-06,
+      "loss": 0.5511,
+      "step": 26896
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.610555154603367e-06,
+      "loss": 0.6333,
+      "step": 26897
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.609884526439567e-06,
+      "loss": 0.6552,
+      "step": 26898
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.609213915501934e-06,
+      "loss": 0.6578,
+      "step": 26899
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.608543321793881e-06,
+      "loss": 0.659,
+      "step": 26900
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.607872745318809e-06,
+      "loss": 0.6148,
+      "step": 26901
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.60720218608013e-06,
+      "loss": 0.5606,
+      "step": 26902
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.606531644081255e-06,
+      "loss": 0.5874,
+      "step": 26903
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.605861119325583e-06,
+      "loss": 0.5817,
+      "step": 26904
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.6051906118165235e-06,
+      "loss": 0.557,
+      "step": 26905
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.604520121557485e-06,
+      "loss": 0.5886,
+      "step": 26906
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.603849648551873e-06,
+      "loss": 0.5406,
+      "step": 26907
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.603179192803098e-06,
+      "loss": 0.5289,
+      "step": 26908
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.602508754314566e-06,
+      "loss": 0.521,
+      "step": 26909
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.601838333089679e-06,
+      "loss": 0.5573,
+      "step": 26910
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.601167929131846e-06,
+      "loss": 0.6747,
+      "step": 26911
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.600497542444475e-06,
+      "loss": 0.6396,
+      "step": 26912
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.599827173030972e-06,
+      "loss": 0.5939,
+      "step": 26913
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.599156820894745e-06,
+      "loss": 0.633,
+      "step": 26914
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.598486486039196e-06,
+      "loss": 0.5892,
+      "step": 26915
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.597816168467733e-06,
+      "loss": 0.5772,
+      "step": 26916
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.597145868183765e-06,
+      "loss": 0.6476,
+      "step": 26917
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.596475585190692e-06,
+      "loss": 0.5927,
+      "step": 26918
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.595805319491924e-06,
+      "loss": 0.5921,
+      "step": 26919
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.595135071090868e-06,
+      "loss": 0.5722,
+      "step": 26920
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.594464839990928e-06,
+      "loss": 0.5738,
+      "step": 26921
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.593794626195506e-06,
+      "loss": 0.5813,
+      "step": 26922
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.593124429708013e-06,
+      "loss": 0.5691,
+      "step": 26923
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.592454250531853e-06,
+      "loss": 0.5577,
+      "step": 26924
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.591784088670434e-06,
+      "loss": 0.5527,
+      "step": 26925
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.591113944127154e-06,
+      "loss": 0.6366,
+      "step": 26926
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.590443816905423e-06,
+      "loss": 0.6438,
+      "step": 26927
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.589773707008644e-06,
+      "loss": 0.5776,
+      "step": 26928
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.589103614440224e-06,
+      "loss": 0.6319,
+      "step": 26929
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.5884335392035735e-06,
+      "loss": 0.5478,
+      "step": 26930
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.587763481302085e-06,
+      "loss": 0.644,
+      "step": 26931
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.587093440739171e-06,
+      "loss": 0.5905,
+      "step": 26932
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.586423417518233e-06,
+      "loss": 0.5684,
+      "step": 26933
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.585753411642679e-06,
+      "loss": 0.5722,
+      "step": 26934
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.5850834231159135e-06,
+      "loss": 0.6272,
+      "step": 26935
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.584413451941338e-06,
+      "loss": 0.6565,
+      "step": 26936
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.583743498122358e-06,
+      "loss": 0.5735,
+      "step": 26937
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.583073561662379e-06,
+      "loss": 0.5793,
+      "step": 26938
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.582403642564801e-06,
+      "loss": 0.5641,
+      "step": 26939
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.5817337408330315e-06,
+      "loss": 0.6191,
+      "step": 26940
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.581063856470477e-06,
+      "loss": 0.6054,
+      "step": 26941
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.580393989480538e-06,
+      "loss": 0.6048,
+      "step": 26942
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.579724139866616e-06,
+      "loss": 0.5576,
+      "step": 26943
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.579054307632117e-06,
+      "loss": 0.5967,
+      "step": 26944
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.578384492780446e-06,
+      "loss": 0.6318,
+      "step": 26945
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.577714695315004e-06,
+      "loss": 0.5797,
+      "step": 26946
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.5770449152392015e-06,
+      "loss": 0.5393,
+      "step": 26947
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.576375152556431e-06,
+      "loss": 0.6081,
+      "step": 26948
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.575705407270101e-06,
+      "loss": 0.6311,
+      "step": 26949
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.575035679383615e-06,
+      "loss": 0.5807,
+      "step": 26950
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.574365968900376e-06,
+      "loss": 0.6497,
+      "step": 26951
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.573696275823791e-06,
+      "loss": 0.512,
+      "step": 26952
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.573026600157255e-06,
+      "loss": 0.6214,
+      "step": 26953
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.572356941904173e-06,
+      "loss": 0.5792,
+      "step": 26954
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.57168730106795e-06,
+      "loss": 0.5546,
+      "step": 26955
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.57101767765199e-06,
+      "loss": 0.6051,
+      "step": 26956
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.570348071659691e-06,
+      "loss": 0.5433,
+      "step": 26957
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.56967848309446e-06,
+      "loss": 0.5697,
+      "step": 26958
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.569008911959698e-06,
+      "loss": 0.5996,
+      "step": 26959
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.568339358258804e-06,
+      "loss": 0.6463,
+      "step": 26960
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.567669821995182e-06,
+      "loss": 0.5984,
+      "step": 26961
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.5670003031722364e-06,
+      "loss": 0.5911,
+      "step": 26962
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 6.56633080179337e-06,
+      "loss": 0.6355,
+      "step": 26963
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5656613178619795e-06,
+      "loss": 0.6136,
+      "step": 26964
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.564991851381469e-06,
+      "loss": 0.5821,
+      "step": 26965
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.564322402355241e-06,
+      "loss": 0.5795,
+      "step": 26966
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.563652970786698e-06,
+      "loss": 0.5998,
+      "step": 26967
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.562983556679243e-06,
+      "loss": 0.5901,
+      "step": 26968
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.562314160036273e-06,
+      "loss": 0.5918,
+      "step": 26969
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.56164478086119e-06,
+      "loss": 0.5913,
+      "step": 26970
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5609754191573985e-06,
+      "loss": 0.6499,
+      "step": 26971
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.560306074928295e-06,
+      "loss": 0.6242,
+      "step": 26972
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.559636748177288e-06,
+      "loss": 0.6497,
+      "step": 26973
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5589674389077754e-06,
+      "loss": 0.5804,
+      "step": 26974
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.558298147123152e-06,
+      "loss": 0.6156,
+      "step": 26975
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5576288728268255e-06,
+      "loss": 0.6242,
+      "step": 26976
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.556959616022196e-06,
+      "loss": 0.6006,
+      "step": 26977
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.556290376712661e-06,
+      "loss": 0.6482,
+      "step": 26978
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.555621154901627e-06,
+      "loss": 0.6061,
+      "step": 26979
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.55495195059249e-06,
+      "loss": 0.5797,
+      "step": 26980
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.554282763788647e-06,
+      "loss": 0.5552,
+      "step": 26981
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5536135944935045e-06,
+      "loss": 0.5903,
+      "step": 26982
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5529444427104595e-06,
+      "loss": 0.5947,
+      "step": 26983
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.552275308442913e-06,
+      "loss": 0.6424,
+      "step": 26984
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.551606191694271e-06,
+      "loss": 0.6152,
+      "step": 26985
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.550937092467923e-06,
+      "loss": 0.599,
+      "step": 26986
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.550268010767273e-06,
+      "loss": 0.5957,
+      "step": 26987
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.549598946595722e-06,
+      "loss": 0.659,
+      "step": 26988
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.54892989995667e-06,
+      "loss": 0.6439,
+      "step": 26989
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.548260870853519e-06,
+      "loss": 0.6307,
+      "step": 26990
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5475918592896616e-06,
+      "loss": 0.5707,
+      "step": 26991
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.546922865268501e-06,
+      "loss": 0.5528,
+      "step": 26992
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.546253888793437e-06,
+      "loss": 0.6065,
+      "step": 26993
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5455849298678695e-06,
+      "loss": 0.6077,
+      "step": 26994
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.544915988495198e-06,
+      "loss": 0.6385,
+      "step": 26995
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.544247064678817e-06,
+      "loss": 0.5728,
+      "step": 26996
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.54357815842213e-06,
+      "loss": 0.5626,
+      "step": 26997
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5429092697285366e-06,
+      "loss": 0.5641,
+      "step": 26998
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.542240398601431e-06,
+      "loss": 0.5623,
+      "step": 26999
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.541571545044214e-06,
+      "loss": 0.5953,
+      "step": 27000
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.540902709060289e-06,
+      "loss": 0.6436,
+      "step": 27001
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5402338906530476e-06,
+      "loss": 0.579,
+      "step": 27002
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.53956508982589e-06,
+      "loss": 0.6405,
+      "step": 27003
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.538896306582217e-06,
+      "loss": 0.6277,
+      "step": 27004
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.538227540925424e-06,
+      "loss": 0.6448,
+      "step": 27005
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.537558792858915e-06,
+      "loss": 0.6113,
+      "step": 27006
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.536890062386079e-06,
+      "loss": 0.5823,
+      "step": 27007
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.53622134951032e-06,
+      "loss": 0.5471,
+      "step": 27008
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.535552654235033e-06,
+      "loss": 0.584,
+      "step": 27009
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.534883976563619e-06,
+      "loss": 0.6307,
+      "step": 27010
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.534215316499475e-06,
+      "loss": 0.6291,
+      "step": 27011
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.533546674046002e-06,
+      "loss": 0.5747,
+      "step": 27012
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.532878049206588e-06,
+      "loss": 0.6099,
+      "step": 27013
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.532209441984638e-06,
+      "loss": 0.562,
+      "step": 27014
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.531540852383548e-06,
+      "loss": 0.5979,
+      "step": 27015
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.530872280406713e-06,
+      "loss": 0.6022,
+      "step": 27016
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.530203726057534e-06,
+      "loss": 0.6014,
+      "step": 27017
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.529535189339406e-06,
+      "loss": 0.6004,
+      "step": 27018
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.528866670255723e-06,
+      "loss": 0.617,
+      "step": 27019
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.528198168809887e-06,
+      "loss": 0.6295,
+      "step": 27020
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.527529685005292e-06,
+      "loss": 0.6334,
+      "step": 27021
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.52686121884534e-06,
+      "loss": 0.6347,
+      "step": 27022
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5261927703334185e-06,
+      "loss": 0.5767,
+      "step": 27023
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.525524339472928e-06,
+      "loss": 0.6168,
+      "step": 27024
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.524855926267268e-06,
+      "loss": 0.5761,
+      "step": 27025
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.524187530719832e-06,
+      "loss": 0.5721,
+      "step": 27026
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.523519152834017e-06,
+      "loss": 0.6013,
+      "step": 27027
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.522850792613222e-06,
+      "loss": 0.6216,
+      "step": 27028
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.522182450060839e-06,
+      "loss": 0.6096,
+      "step": 27029
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5215141251802635e-06,
+      "loss": 0.5771,
+      "step": 27030
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.520845817974894e-06,
+      "loss": 0.6726,
+      "step": 27031
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.520177528448126e-06,
+      "loss": 0.5735,
+      "step": 27032
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.51950925660336e-06,
+      "loss": 0.6467,
+      "step": 27033
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.518841002443982e-06,
+      "loss": 0.5799,
+      "step": 27034
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5181727659733915e-06,
+      "loss": 0.6046,
+      "step": 27035
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.517504547194988e-06,
+      "loss": 0.5358,
+      "step": 27036
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.516836346112161e-06,
+      "loss": 0.5913,
+      "step": 27037
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.516168162728309e-06,
+      "loss": 0.5793,
+      "step": 27038
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.515499997046831e-06,
+      "loss": 0.5991,
+      "step": 27039
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.514831849071115e-06,
+      "loss": 0.5805,
+      "step": 27040
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.514163718804558e-06,
+      "loss": 0.6205,
+      "step": 27041
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.513495606250555e-06,
+      "loss": 0.5442,
+      "step": 27042
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.512827511412503e-06,
+      "loss": 0.5885,
+      "step": 27043
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5121594342938e-06,
+      "loss": 0.5819,
+      "step": 27044
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.511491374897832e-06,
+      "loss": 0.6274,
+      "step": 27045
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.510823333227998e-06,
+      "loss": 0.5907,
+      "step": 27046
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.510155309287691e-06,
+      "loss": 0.5854,
+      "step": 27047
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.509487303080308e-06,
+      "loss": 0.6091,
+      "step": 27048
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5088193146092435e-06,
+      "loss": 0.6075,
+      "step": 27049
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.508151343877895e-06,
+      "loss": 0.6451,
+      "step": 27050
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.507483390889646e-06,
+      "loss": 0.5497,
+      "step": 27051
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.506815455647897e-06,
+      "loss": 0.573,
+      "step": 27052
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.506147538156043e-06,
+      "loss": 0.5894,
+      "step": 27053
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.505479638417478e-06,
+      "loss": 0.6236,
+      "step": 27054
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.504811756435594e-06,
+      "loss": 0.624,
+      "step": 27055
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.504143892213782e-06,
+      "loss": 0.5898,
+      "step": 27056
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.503476045755442e-06,
+      "loss": 0.6306,
+      "step": 27057
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.502808217063963e-06,
+      "loss": 0.5458,
+      "step": 27058
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.502140406142739e-06,
+      "loss": 0.5437,
+      "step": 27059
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.501472612995168e-06,
+      "loss": 0.5626,
+      "step": 27060
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5008048376246344e-06,
+      "loss": 0.5761,
+      "step": 27061
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.5001370800345366e-06,
+      "loss": 0.647,
+      "step": 27062
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.499469340228267e-06,
+      "loss": 0.5693,
+      "step": 27063
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.49880161820922e-06,
+      "loss": 0.5882,
+      "step": 27064
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.498133913980786e-06,
+      "loss": 0.6067,
+      "step": 27065
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.497466227546364e-06,
+      "loss": 0.6713,
+      "step": 27066
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.496798558909338e-06,
+      "loss": 0.5727,
+      "step": 27067
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.496130908073104e-06,
+      "loss": 0.6246,
+      "step": 27068
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4954632750410555e-06,
+      "loss": 0.564,
+      "step": 27069
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.494795659816584e-06,
+      "loss": 0.6239,
+      "step": 27070
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.494128062403087e-06,
+      "loss": 0.6047,
+      "step": 27071
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.493460482803946e-06,
+      "loss": 0.5693,
+      "step": 27072
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4927929210225615e-06,
+      "loss": 0.5988,
+      "step": 27073
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.492125377062324e-06,
+      "loss": 0.6321,
+      "step": 27074
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.491457850926623e-06,
+      "loss": 0.6282,
+      "step": 27075
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.490790342618854e-06,
+      "loss": 0.5971,
+      "step": 27076
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.490122852142407e-06,
+      "loss": 0.6121,
+      "step": 27077
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.489455379500673e-06,
+      "loss": 0.5647,
+      "step": 27078
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.488787924697044e-06,
+      "loss": 0.6146,
+      "step": 27079
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.488120487734912e-06,
+      "loss": 0.5886,
+      "step": 27080
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.487453068617667e-06,
+      "loss": 0.6256,
+      "step": 27081
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4867856673487075e-06,
+      "loss": 0.6055,
+      "step": 27082
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.486118283931413e-06,
+      "loss": 0.5778,
+      "step": 27083
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.485450918369182e-06,
+      "loss": 0.6198,
+      "step": 27084
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4847835706654025e-06,
+      "loss": 0.5265,
+      "step": 27085
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.484116240823468e-06,
+      "loss": 0.614,
+      "step": 27086
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.483448928846774e-06,
+      "loss": 0.666,
+      "step": 27087
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.482781634738699e-06,
+      "loss": 0.6174,
+      "step": 27088
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.482114358502643e-06,
+      "loss": 0.592,
+      "step": 27089
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.481447100141993e-06,
+      "loss": 0.6706,
+      "step": 27090
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.480779859660142e-06,
+      "loss": 0.655,
+      "step": 27091
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.480112637060481e-06,
+      "loss": 0.612,
+      "step": 27092
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4794454323463965e-06,
+      "loss": 0.5816,
+      "step": 27093
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.47877824552128e-06,
+      "loss": 0.6515,
+      "step": 27094
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.478111076588525e-06,
+      "loss": 0.64,
+      "step": 27095
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.477443925551517e-06,
+      "loss": 0.5388,
+      "step": 27096
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.476776792413647e-06,
+      "loss": 0.6314,
+      "step": 27097
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.476109677178308e-06,
+      "loss": 0.6255,
+      "step": 27098
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.47544257984889e-06,
+      "loss": 0.6384,
+      "step": 27099
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.474775500428776e-06,
+      "loss": 0.6199,
+      "step": 27100
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.474108438921361e-06,
+      "loss": 0.6527,
+      "step": 27101
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.473441395330034e-06,
+      "loss": 0.5772,
+      "step": 27102
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.472774369658183e-06,
+      "loss": 0.6754,
+      "step": 27103
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.472107361909204e-06,
+      "loss": 0.6163,
+      "step": 27104
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.471440372086475e-06,
+      "loss": 0.6036,
+      "step": 27105
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.470773400193392e-06,
+      "loss": 0.6368,
+      "step": 27106
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.4701064462333415e-06,
+      "loss": 0.6247,
+      "step": 27107
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.469439510209717e-06,
+      "loss": 0.5885,
+      "step": 27108
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.468772592125905e-06,
+      "loss": 0.6019,
+      "step": 27109
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.468105691985291e-06,
+      "loss": 0.6355,
+      "step": 27110
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.467438809791267e-06,
+      "loss": 0.5916,
+      "step": 27111
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.466771945547219e-06,
+      "loss": 0.5624,
+      "step": 27112
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.466105099256541e-06,
+      "loss": 0.6288,
+      "step": 27113
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.465438270922617e-06,
+      "loss": 0.6424,
+      "step": 27114
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4647714605488356e-06,
+      "loss": 0.5348,
+      "step": 27115
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4641046681385865e-06,
+      "loss": 0.5911,
+      "step": 27116
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4634378936952564e-06,
+      "loss": 0.6222,
+      "step": 27117
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4627711372222335e-06,
+      "loss": 0.5515,
+      "step": 27118
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.462104398722906e-06,
+      "loss": 0.5626,
+      "step": 27119
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.461437678200667e-06,
+      "loss": 0.5953,
+      "step": 27120
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4607709756588945e-06,
+      "loss": 0.6206,
+      "step": 27121
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.460104291100981e-06,
+      "loss": 0.6015,
+      "step": 27122
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.459437624530316e-06,
+      "loss": 0.6124,
+      "step": 27123
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.458770975950284e-06,
+      "loss": 0.6441,
+      "step": 27124
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4581043453642776e-06,
+      "loss": 0.5807,
+      "step": 27125
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4574377327756775e-06,
+      "loss": 0.5878,
+      "step": 27126
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.456771138187873e-06,
+      "loss": 0.5449,
+      "step": 27127
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.456104561604252e-06,
+      "loss": 0.5846,
+      "step": 27128
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4554380030282e-06,
+      "loss": 0.631,
+      "step": 27129
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.45477146246311e-06,
+      "loss": 0.5743,
+      "step": 27130
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.454104939912365e-06,
+      "loss": 0.6131,
+      "step": 27131
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.453438435379347e-06,
+      "loss": 0.6044,
+      "step": 27132
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.452771948867448e-06,
+      "loss": 0.5454,
+      "step": 27133
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.452105480380055e-06,
+      "loss": 0.5772,
+      "step": 27134
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.451439029920552e-06,
+      "loss": 0.5659,
+      "step": 27135
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.450772597492328e-06,
+      "loss": 0.6077,
+      "step": 27136
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.450106183098768e-06,
+      "loss": 0.6889,
+      "step": 27137
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.449439786743257e-06,
+      "loss": 0.5865,
+      "step": 27138
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.448773408429181e-06,
+      "loss": 0.6307,
+      "step": 27139
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.448107048159927e-06,
+      "loss": 0.6148,
+      "step": 27140
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4474407059388836e-06,
+      "loss": 0.6108,
+      "step": 27141
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.446774381769436e-06,
+      "loss": 0.6164,
+      "step": 27142
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.446108075654965e-06,
+      "loss": 0.6524,
+      "step": 27143
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.445441787598861e-06,
+      "loss": 0.6655,
+      "step": 27144
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.444775517604507e-06,
+      "loss": 0.547,
+      "step": 27145
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.444109265675289e-06,
+      "loss": 0.5708,
+      "step": 27146
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.443443031814598e-06,
+      "loss": 0.5428,
+      "step": 27147
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.442776816025811e-06,
+      "loss": 0.6568,
+      "step": 27148
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.442110618312316e-06,
+      "loss": 0.5866,
+      "step": 27149
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4414444386775e-06,
+      "loss": 0.5999,
+      "step": 27150
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4407782771247486e-06,
+      "loss": 0.6087,
+      "step": 27151
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.440112133657444e-06,
+      "loss": 0.6063,
+      "step": 27152
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.439446008278971e-06,
+      "loss": 0.5468,
+      "step": 27153
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.438779900992718e-06,
+      "loss": 0.5887,
+      "step": 27154
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4381138118020644e-06,
+      "loss": 0.5926,
+      "step": 27155
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.437447740710398e-06,
+      "loss": 0.6806,
+      "step": 27156
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.436781687721104e-06,
+      "loss": 0.6432,
+      "step": 27157
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.436115652837569e-06,
+      "loss": 0.6438,
+      "step": 27158
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.435449636063171e-06,
+      "loss": 0.5868,
+      "step": 27159
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.434783637401296e-06,
+      "loss": 0.6512,
+      "step": 27160
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.43411765685533e-06,
+      "loss": 0.5695,
+      "step": 27161
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.433451694428656e-06,
+      "loss": 0.5852,
+      "step": 27162
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.432785750124664e-06,
+      "loss": 0.5524,
+      "step": 27163
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.432119823946726e-06,
+      "loss": 0.5573,
+      "step": 27164
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4314539158982335e-06,
+      "loss": 0.6302,
+      "step": 27165
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.430788025982568e-06,
+      "loss": 0.5866,
+      "step": 27166
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.430122154203115e-06,
+      "loss": 0.5836,
+      "step": 27167
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.429456300563256e-06,
+      "loss": 0.6264,
+      "step": 27168
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.428790465066379e-06,
+      "loss": 0.5257,
+      "step": 27169
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.428124647715861e-06,
+      "loss": 0.6197,
+      "step": 27170
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4274588485150855e-06,
+      "loss": 0.5482,
+      "step": 27171
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.426793067467441e-06,
+      "loss": 0.6259,
+      "step": 27172
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.426127304576306e-06,
+      "loss": 0.6029,
+      "step": 27173
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.425461559845065e-06,
+      "loss": 0.5649,
+      "step": 27174
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.424795833277102e-06,
+      "loss": 0.6717,
+      "step": 27175
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.424130124875797e-06,
+      "loss": 0.5544,
+      "step": 27176
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.423464434644533e-06,
+      "loss": 0.5754,
+      "step": 27177
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.422798762586695e-06,
+      "loss": 0.5419,
+      "step": 27178
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.422133108705666e-06,
+      "loss": 0.5841,
+      "step": 27179
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.421467473004824e-06,
+      "loss": 0.6159,
+      "step": 27180
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.420801855487553e-06,
+      "loss": 0.5852,
+      "step": 27181
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.420136256157236e-06,
+      "loss": 0.5967,
+      "step": 27182
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.419470675017256e-06,
+      "loss": 0.6274,
+      "step": 27183
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.418805112070993e-06,
+      "loss": 0.6094,
+      "step": 27184
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.418139567321836e-06,
+      "loss": 0.6292,
+      "step": 27185
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.417474040773155e-06,
+      "loss": 0.599,
+      "step": 27186
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.416808532428338e-06,
+      "loss": 0.5897,
+      "step": 27187
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4161430422907664e-06,
+      "loss": 0.5823,
+      "step": 27188
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.415477570363821e-06,
+      "loss": 0.5586,
+      "step": 27189
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.414812116650889e-06,
+      "loss": 0.6034,
+      "step": 27190
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.414146681155342e-06,
+      "loss": 0.6236,
+      "step": 27191
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.413481263880565e-06,
+      "loss": 0.5933,
+      "step": 27192
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.412815864829945e-06,
+      "loss": 0.582,
+      "step": 27193
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.412150484006853e-06,
+      "loss": 0.6151,
+      "step": 27194
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.411485121414675e-06,
+      "loss": 0.5839,
+      "step": 27195
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.410819777056798e-06,
+      "loss": 0.6406,
+      "step": 27196
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.410154450936592e-06,
+      "loss": 0.6009,
+      "step": 27197
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4094891430574436e-06,
+      "loss": 0.5955,
+      "step": 27198
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.40882385342273e-06,
+      "loss": 0.6338,
+      "step": 27199
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.408158582035836e-06,
+      "loss": 0.6472,
+      "step": 27200
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.407493328900144e-06,
+      "loss": 0.6292,
+      "step": 27201
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.406828094019027e-06,
+      "loss": 0.6104,
+      "step": 27202
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.406162877395867e-06,
+      "loss": 0.5565,
+      "step": 27203
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.405497679034047e-06,
+      "loss": 0.6513,
+      "step": 27204
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.4048324989369455e-06,
+      "loss": 0.6061,
+      "step": 27205
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.404167337107947e-06,
+      "loss": 0.5472,
+      "step": 27206
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.403502193550424e-06,
+      "loss": 0.6006,
+      "step": 27207
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.40283706826776e-06,
+      "loss": 0.6376,
+      "step": 27208
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.402171961263334e-06,
+      "loss": 0.5838,
+      "step": 27209
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.401506872540527e-06,
+      "loss": 0.568,
+      "step": 27210
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.400841802102717e-06,
+      "loss": 0.6492,
+      "step": 27211
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.400176749953284e-06,
+      "loss": 0.5807,
+      "step": 27212
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.399511716095606e-06,
+      "loss": 0.5985,
+      "step": 27213
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.398846700533066e-06,
+      "loss": 0.6153,
+      "step": 27214
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.398181703269037e-06,
+      "loss": 0.5899,
+      "step": 27215
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.397516724306904e-06,
+      "loss": 0.6078,
+      "step": 27216
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.396851763650046e-06,
+      "loss": 0.5753,
+      "step": 27217
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.396186821301835e-06,
+      "loss": 0.6753,
+      "step": 27218
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.395521897265655e-06,
+      "loss": 0.5812,
+      "step": 27219
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.394856991544884e-06,
+      "loss": 0.6473,
+      "step": 27220
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3941921041429e-06,
+      "loss": 0.5922,
+      "step": 27221
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3935272350630824e-06,
+      "loss": 0.6087,
+      "step": 27222
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.392862384308813e-06,
+      "loss": 0.6168,
+      "step": 27223
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.392197551883463e-06,
+      "loss": 0.6157,
+      "step": 27224
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.391532737790412e-06,
+      "loss": 0.6075,
+      "step": 27225
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.390867942033041e-06,
+      "loss": 0.636,
+      "step": 27226
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.390203164614727e-06,
+      "loss": 0.6032,
+      "step": 27227
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.389538405538851e-06,
+      "loss": 0.6322,
+      "step": 27228
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.388873664808785e-06,
+      "loss": 0.5944,
+      "step": 27229
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.388208942427909e-06,
+      "loss": 0.6238,
+      "step": 27230
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.387544238399603e-06,
+      "loss": 0.5811,
+      "step": 27231
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.386879552727241e-06,
+      "loss": 0.6222,
+      "step": 27232
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.386214885414201e-06,
+      "loss": 0.5421,
+      "step": 27233
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.385550236463865e-06,
+      "loss": 0.598,
+      "step": 27234
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.384885605879607e-06,
+      "loss": 0.6247,
+      "step": 27235
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.384220993664801e-06,
+      "loss": 0.6019,
+      "step": 27236
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3835563998228265e-06,
+      "loss": 0.5628,
+      "step": 27237
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.382891824357063e-06,
+      "loss": 0.6238,
+      "step": 27238
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3822272672708875e-06,
+      "loss": 0.5802,
+      "step": 27239
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.381562728567672e-06,
+      "loss": 0.5905,
+      "step": 27240
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.380898208250794e-06,
+      "loss": 0.5272,
+      "step": 27241
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.380233706323634e-06,
+      "loss": 0.6009,
+      "step": 27242
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.379569222789567e-06,
+      "loss": 0.5504,
+      "step": 27243
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.378904757651972e-06,
+      "loss": 0.616,
+      "step": 27244
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.378240310914219e-06,
+      "loss": 0.6303,
+      "step": 27245
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3775758825796874e-06,
+      "loss": 0.6206,
+      "step": 27246
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.376911472651754e-06,
+      "loss": 0.6005,
+      "step": 27247
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.376247081133793e-06,
+      "loss": 0.6091,
+      "step": 27248
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3755827080291845e-06,
+      "loss": 0.5644,
+      "step": 27249
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.374918353341303e-06,
+      "loss": 0.5847,
+      "step": 27250
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.37425401707352e-06,
+      "loss": 0.601,
+      "step": 27251
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.373589699229217e-06,
+      "loss": 0.6143,
+      "step": 27252
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.372925399811763e-06,
+      "loss": 0.5752,
+      "step": 27253
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.372261118824539e-06,
+      "loss": 0.5819,
+      "step": 27254
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.3715968562709215e-06,
+      "loss": 0.5967,
+      "step": 27255
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.370932612154279e-06,
+      "loss": 0.6109,
+      "step": 27256
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.370268386477993e-06,
+      "loss": 0.6657,
+      "step": 27257
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3696041792454346e-06,
+      "loss": 0.6001,
+      "step": 27258
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.368939990459979e-06,
+      "loss": 0.5831,
+      "step": 27259
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.368275820125005e-06,
+      "loss": 0.5482,
+      "step": 27260
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.367611668243889e-06,
+      "loss": 0.5687,
+      "step": 27261
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.366947534819997e-06,
+      "loss": 0.5539,
+      "step": 27262
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.366283419856709e-06,
+      "loss": 0.5776,
+      "step": 27263
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.365619323357399e-06,
+      "loss": 0.6092,
+      "step": 27264
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.364955245325441e-06,
+      "loss": 0.6099,
+      "step": 27265
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.364291185764215e-06,
+      "loss": 0.6033,
+      "step": 27266
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.363627144677087e-06,
+      "loss": 0.6548,
+      "step": 27267
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.362963122067433e-06,
+      "loss": 0.5774,
+      "step": 27268
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.362299117938629e-06,
+      "loss": 0.6248,
+      "step": 27269
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.36163513229405e-06,
+      "loss": 0.6268,
+      "step": 27270
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.360971165137069e-06,
+      "loss": 0.6167,
+      "step": 27271
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.360307216471058e-06,
+      "loss": 0.5911,
+      "step": 27272
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.359643286299393e-06,
+      "loss": 0.6923,
+      "step": 27273
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3589793746254436e-06,
+      "loss": 0.5947,
+      "step": 27274
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.358315481452589e-06,
+      "loss": 0.6311,
+      "step": 27275
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.357651606784197e-06,
+      "loss": 0.6143,
+      "step": 27276
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3569877506236484e-06,
+      "loss": 0.6398,
+      "step": 27277
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.356323912974309e-06,
+      "loss": 0.5948,
+      "step": 27278
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.355660093839555e-06,
+      "loss": 0.5938,
+      "step": 27279
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.35499629322276e-06,
+      "loss": 0.6225,
+      "step": 27280
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.354332511127295e-06,
+      "loss": 0.583,
+      "step": 27281
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.353668747556538e-06,
+      "loss": 0.5886,
+      "step": 27282
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.353005002513854e-06,
+      "loss": 0.5787,
+      "step": 27283
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.35234127600262e-06,
+      "loss": 0.5361,
+      "step": 27284
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3516775680262085e-06,
+      "loss": 0.6362,
+      "step": 27285
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.351013878587993e-06,
+      "loss": 0.5984,
+      "step": 27286
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.350350207691345e-06,
+      "loss": 0.5762,
+      "step": 27287
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.349686555339636e-06,
+      "loss": 0.6227,
+      "step": 27288
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.349022921536237e-06,
+      "loss": 0.5445,
+      "step": 27289
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.348359306284523e-06,
+      "loss": 0.6002,
+      "step": 27290
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.347695709587867e-06,
+      "loss": 0.6012,
+      "step": 27291
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.347032131449635e-06,
+      "loss": 0.624,
+      "step": 27292
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.346368571873205e-06,
+      "loss": 0.6476,
+      "step": 27293
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.345705030861946e-06,
+      "loss": 0.6242,
+      "step": 27294
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.345041508419229e-06,
+      "loss": 0.5889,
+      "step": 27295
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.344378004548426e-06,
+      "loss": 0.5942,
+      "step": 27296
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3437145192529086e-06,
+      "loss": 0.6564,
+      "step": 27297
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3430510525360535e-06,
+      "loss": 0.5858,
+      "step": 27298
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3423876044012225e-06,
+      "loss": 0.5837,
+      "step": 27299
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.341724174851791e-06,
+      "loss": 0.6314,
+      "step": 27300
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.341060763891129e-06,
+      "loss": 0.618,
+      "step": 27301
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.340397371522611e-06,
+      "loss": 0.6545,
+      "step": 27302
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3397339977496055e-06,
+      "loss": 0.6688,
+      "step": 27303
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.339070642575487e-06,
+      "loss": 0.5406,
+      "step": 27304
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.338407306003619e-06,
+      "loss": 0.6157,
+      "step": 27305
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.337743988037375e-06,
+      "loss": 0.6475,
+      "step": 27306
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.337080688680127e-06,
+      "loss": 0.6448,
+      "step": 27307
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.336417407935247e-06,
+      "loss": 0.6074,
+      "step": 27308
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.335754145806103e-06,
+      "loss": 0.6107,
+      "step": 27309
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3350909022960625e-06,
+      "loss": 0.7087,
+      "step": 27310
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.334427677408501e-06,
+      "loss": 0.5937,
+      "step": 27311
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.333764471146784e-06,
+      "loss": 0.5778,
+      "step": 27312
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.333101283514284e-06,
+      "loss": 0.6141,
+      "step": 27313
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.33243811451437e-06,
+      "loss": 0.5778,
+      "step": 27314
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.331774964150415e-06,
+      "loss": 0.616,
+      "step": 27315
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.331111832425782e-06,
+      "loss": 0.6629,
+      "step": 27316
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.330448719343844e-06,
+      "loss": 0.5385,
+      "step": 27317
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.329785624907972e-06,
+      "loss": 0.6678,
+      "step": 27318
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.329122549121534e-06,
+      "loss": 0.6027,
+      "step": 27319
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.328459491987904e-06,
+      "loss": 0.5868,
+      "step": 27320
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.327796453510441e-06,
+      "loss": 0.6124,
+      "step": 27321
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.32713343369252e-06,
+      "loss": 0.6194,
+      "step": 27322
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.32647043253751e-06,
+      "loss": 0.5776,
+      "step": 27323
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.32580745004878e-06,
+      "loss": 0.5537,
+      "step": 27324
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.325144486229697e-06,
+      "loss": 0.5905,
+      "step": 27325
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.324481541083636e-06,
+      "loss": 0.6009,
+      "step": 27326
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.323818614613957e-06,
+      "loss": 0.6088,
+      "step": 27327
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.323155706824031e-06,
+      "loss": 0.6346,
+      "step": 27328
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.322492817717232e-06,
+      "loss": 0.6256,
+      "step": 27329
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.32182994729692e-06,
+      "loss": 0.6305,
+      "step": 27330
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.32116709556647e-06,
+      "loss": 0.5903,
+      "step": 27331
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.320504262529247e-06,
+      "loss": 0.5605,
+      "step": 27332
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.319841448188617e-06,
+      "loss": 0.5677,
+      "step": 27333
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.31917865254795e-06,
+      "loss": 0.5686,
+      "step": 27334
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.318515875610616e-06,
+      "loss": 0.5902,
+      "step": 27335
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.317853117379984e-06,
+      "loss": 0.5924,
+      "step": 27336
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.317190377859415e-06,
+      "loss": 0.6096,
+      "step": 27337
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.316527657052281e-06,
+      "loss": 0.592,
+      "step": 27338
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.315864954961948e-06,
+      "loss": 0.6102,
+      "step": 27339
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.315202271591783e-06,
+      "loss": 0.488,
+      "step": 27340
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.314539606945156e-06,
+      "loss": 0.563,
+      "step": 27341
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.313876961025436e-06,
+      "loss": 0.6128,
+      "step": 27342
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.313214333835982e-06,
+      "loss": 0.5515,
+      "step": 27343
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3125517253801675e-06,
+      "loss": 0.6399,
+      "step": 27344
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3118891356613554e-06,
+      "loss": 0.615,
+      "step": 27345
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.311226564682919e-06,
+      "loss": 0.5651,
+      "step": 27346
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.310564012448219e-06,
+      "loss": 0.5337,
+      "step": 27347
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.309901478960623e-06,
+      "loss": 0.5986,
+      "step": 27348
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.309238964223496e-06,
+      "loss": 0.5883,
+      "step": 27349
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.308576468240211e-06,
+      "loss": 0.6234,
+      "step": 27350
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3079139910141275e-06,
+      "loss": 0.628,
+      "step": 27351
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.307251532548614e-06,
+      "loss": 0.5625,
+      "step": 27352
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.306589092847042e-06,
+      "loss": 0.5285,
+      "step": 27353
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.305926671912768e-06,
+      "loss": 0.5575,
+      "step": 27354
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.30526426974916e-06,
+      "loss": 0.5818,
+      "step": 27355
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.304601886359589e-06,
+      "loss": 0.5961,
+      "step": 27356
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.303939521747417e-06,
+      "loss": 0.6511,
+      "step": 27357
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.303277175916016e-06,
+      "loss": 0.5206,
+      "step": 27358
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.302614848868741e-06,
+      "loss": 0.5957,
+      "step": 27359
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.3019525406089635e-06,
+      "loss": 0.611,
+      "step": 27360
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.301290251140047e-06,
+      "loss": 0.5961,
+      "step": 27361
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.300627980465359e-06,
+      "loss": 0.5866,
+      "step": 27362
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.299965728588268e-06,
+      "loss": 0.5638,
+      "step": 27363
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.29930349551213e-06,
+      "loss": 0.5838,
+      "step": 27364
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2986412812403145e-06,
+      "loss": 0.6014,
+      "step": 27365
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.297979085776187e-06,
+      "loss": 0.6831,
+      "step": 27366
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.297316909123115e-06,
+      "loss": 0.587,
+      "step": 27367
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.296654751284457e-06,
+      "loss": 0.6359,
+      "step": 27368
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.295992612263582e-06,
+      "loss": 0.6491,
+      "step": 27369
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2953304920638525e-06,
+      "loss": 0.6265,
+      "step": 27370
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.294668390688636e-06,
+      "loss": 0.6707,
+      "step": 27371
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.294006308141292e-06,
+      "loss": 0.6208,
+      "step": 27372
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.293344244425187e-06,
+      "loss": 0.5949,
+      "step": 27373
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2926821995436895e-06,
+      "loss": 0.5695,
+      "step": 27374
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.292020173500156e-06,
+      "loss": 0.5519,
+      "step": 27375
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.291358166297954e-06,
+      "loss": 0.6021,
+      "step": 27376
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.290696177940448e-06,
+      "loss": 0.5918,
+      "step": 27377
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2900342084309995e-06,
+      "loss": 0.5746,
+      "step": 27378
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.289372257772973e-06,
+      "loss": 0.608,
+      "step": 27379
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.288710325969738e-06,
+      "loss": 0.5797,
+      "step": 27380
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.28804841302465e-06,
+      "loss": 0.5883,
+      "step": 27381
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.287386518941073e-06,
+      "loss": 0.5625,
+      "step": 27382
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.286724643722373e-06,
+      "loss": 0.6055,
+      "step": 27383
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.286062787371914e-06,
+      "loss": 0.6211,
+      "step": 27384
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.285400949893059e-06,
+      "loss": 0.6038,
+      "step": 27385
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2847391312891685e-06,
+      "loss": 0.594,
+      "step": 27386
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.284077331563605e-06,
+      "loss": 0.6003,
+      "step": 27387
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.283415550719734e-06,
+      "loss": 0.6333,
+      "step": 27388
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.282753788760916e-06,
+      "loss": 0.594,
+      "step": 27389
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2820920456905176e-06,
+      "loss": 0.6199,
+      "step": 27390
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.281430321511897e-06,
+      "loss": 0.5836,
+      "step": 27391
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2807686162284166e-06,
+      "loss": 0.6144,
+      "step": 27392
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.280106929843439e-06,
+      "loss": 0.622,
+      "step": 27393
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.279445262360328e-06,
+      "loss": 0.5671,
+      "step": 27394
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.278783613782446e-06,
+      "loss": 0.5703,
+      "step": 27395
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.278121984113159e-06,
+      "loss": 0.5844,
+      "step": 27396
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.277460373355818e-06,
+      "loss": 0.5728,
+      "step": 27397
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2767987815137936e-06,
+      "loss": 0.5516,
+      "step": 27398
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.276137208590442e-06,
+      "loss": 0.6267,
+      "step": 27399
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2754756545891295e-06,
+      "loss": 0.6431,
+      "step": 27400
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.2748141195132195e-06,
+      "loss": 0.5899,
+      "step": 27401
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.274152603366066e-06,
+      "loss": 0.6084,
+      "step": 27402
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.273491106151035e-06,
+      "loss": 0.5912,
+      "step": 27403
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.272829627871486e-06,
+      "loss": 0.5765,
+      "step": 27404
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2721681685307835e-06,
+      "loss": 0.6419,
+      "step": 27405
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.271506728132286e-06,
+      "loss": 0.62,
+      "step": 27406
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.270845306679355e-06,
+      "loss": 0.6338,
+      "step": 27407
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.270183904175349e-06,
+      "loss": 0.6103,
+      "step": 27408
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.269522520623634e-06,
+      "loss": 0.5691,
+      "step": 27409
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2688611560275645e-06,
+      "loss": 0.617,
+      "step": 27410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.268199810390505e-06,
+      "loss": 0.6264,
+      "step": 27411
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.267538483715818e-06,
+      "loss": 0.6089,
+      "step": 27412
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.266877176006858e-06,
+      "loss": 0.6119,
+      "step": 27413
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.266215887266989e-06,
+      "loss": 0.582,
+      "step": 27414
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.26555461749957e-06,
+      "loss": 0.5814,
+      "step": 27415
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.26489336670796e-06,
+      "loss": 0.5991,
+      "step": 27416
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.264232134895523e-06,
+      "loss": 0.5879,
+      "step": 27417
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.26357092206562e-06,
+      "loss": 0.5861,
+      "step": 27418
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.262909728221604e-06,
+      "loss": 0.6286,
+      "step": 27419
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.262248553366837e-06,
+      "loss": 0.5362,
+      "step": 27420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.261587397504679e-06,
+      "loss": 0.64,
+      "step": 27421
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.260926260638491e-06,
+      "loss": 0.5975,
+      "step": 27422
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.260265142771638e-06,
+      "loss": 0.5868,
+      "step": 27423
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.259604043907466e-06,
+      "loss": 0.5851,
+      "step": 27424
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.258942964049341e-06,
+      "loss": 0.5332,
+      "step": 27425
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.258281903200625e-06,
+      "loss": 0.5771,
+      "step": 27426
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.257620861364675e-06,
+      "loss": 0.6426,
+      "step": 27427
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2569598385448495e-06,
+      "loss": 0.5533,
+      "step": 27428
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.256298834744505e-06,
+      "loss": 0.6199,
+      "step": 27429
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.255637849967005e-06,
+      "loss": 0.5442,
+      "step": 27430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.254976884215704e-06,
+      "loss": 0.6064,
+      "step": 27431
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.254315937493962e-06,
+      "loss": 0.629,
+      "step": 27432
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2536550098051375e-06,
+      "loss": 0.6015,
+      "step": 27433
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.252994101152593e-06,
+      "loss": 0.6626,
+      "step": 27434
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.252333211539678e-06,
+      "loss": 0.5627,
+      "step": 27435
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.251672340969758e-06,
+      "loss": 0.5957,
+      "step": 27436
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.251011489446186e-06,
+      "loss": 0.5397,
+      "step": 27437
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.250350656972324e-06,
+      "loss": 0.6203,
+      "step": 27438
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.249689843551532e-06,
+      "loss": 0.6139,
+      "step": 27439
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.249029049187161e-06,
+      "loss": 0.6358,
+      "step": 27440
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.24836827388257e-06,
+      "loss": 0.5289,
+      "step": 27441
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.24770751764112e-06,
+      "loss": 0.6375,
+      "step": 27442
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.247046780466166e-06,
+      "loss": 0.6227,
+      "step": 27443
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2463860623610695e-06,
+      "loss": 0.5966,
+      "step": 27444
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.245725363329185e-06,
+      "loss": 0.6075,
+      "step": 27445
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.245064683373867e-06,
+      "loss": 0.6111,
+      "step": 27446
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.244404022498476e-06,
+      "loss": 0.5948,
+      "step": 27447
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.243743380706367e-06,
+      "loss": 0.543,
+      "step": 27448
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.243082758000898e-06,
+      "loss": 0.5822,
+      "step": 27449
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.242422154385428e-06,
+      "loss": 0.6242,
+      "step": 27450
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.24176156986331e-06,
+      "loss": 0.6207,
+      "step": 27451
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.241101004437902e-06,
+      "loss": 0.5955,
+      "step": 27452
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.24044045811256e-06,
+      "loss": 0.5717,
+      "step": 27453
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2397799308906405e-06,
+      "loss": 0.5908,
+      "step": 27454
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.239119422775505e-06,
+      "loss": 0.6127,
+      "step": 27455
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.238458933770501e-06,
+      "loss": 0.5815,
+      "step": 27456
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.237798463878988e-06,
+      "loss": 0.6073,
+      "step": 27457
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.237138013104324e-06,
+      "loss": 0.6456,
+      "step": 27458
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.236477581449862e-06,
+      "loss": 0.5615,
+      "step": 27459
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.235817168918961e-06,
+      "loss": 0.5703,
+      "step": 27460
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.23515677551498e-06,
+      "loss": 0.5536,
+      "step": 27461
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.234496401241266e-06,
+      "loss": 0.6248,
+      "step": 27462
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.233836046101177e-06,
+      "loss": 0.6095,
+      "step": 27463
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.23317571009807e-06,
+      "loss": 0.5421,
+      "step": 27464
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2325153932353035e-06,
+      "loss": 0.6474,
+      "step": 27465
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.231855095516231e-06,
+      "loss": 0.5881,
+      "step": 27466
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.231194816944202e-06,
+      "loss": 0.5839,
+      "step": 27467
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2305345575225785e-06,
+      "loss": 0.6671,
+      "step": 27468
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.229874317254711e-06,
+      "loss": 0.5695,
+      "step": 27469
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2292140961439565e-06,
+      "loss": 0.5751,
+      "step": 27470
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.228553894193667e-06,
+      "loss": 0.5896,
+      "step": 27471
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.227893711407207e-06,
+      "loss": 0.5953,
+      "step": 27472
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.227233547787919e-06,
+      "loss": 0.6422,
+      "step": 27473
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.226573403339161e-06,
+      "loss": 0.6028,
+      "step": 27474
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.22591327806429e-06,
+      "loss": 0.6486,
+      "step": 27475
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2252531719666586e-06,
+      "loss": 0.6303,
+      "step": 27476
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.224593085049624e-06,
+      "loss": 0.611,
+      "step": 27477
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.223933017316535e-06,
+      "loss": 0.588,
+      "step": 27478
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.223272968770748e-06,
+      "loss": 0.6136,
+      "step": 27479
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.222612939415618e-06,
+      "loss": 0.549,
+      "step": 27480
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.221952929254496e-06,
+      "loss": 0.6206,
+      "step": 27481
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.221292938290743e-06,
+      "loss": 0.5505,
+      "step": 27482
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.220632966527702e-06,
+      "loss": 0.5968,
+      "step": 27483
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.219973013968733e-06,
+      "loss": 0.6319,
+      "step": 27484
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.219313080617188e-06,
+      "loss": 0.603,
+      "step": 27485
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.218653166476421e-06,
+      "loss": 0.6202,
+      "step": 27486
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.217993271549785e-06,
+      "loss": 0.6184,
+      "step": 27487
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2173333958406335e-06,
+      "loss": 0.5593,
+      "step": 27488
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.216673539352318e-06,
+      "loss": 0.5731,
+      "step": 27489
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.216013702088192e-06,
+      "loss": 0.611,
+      "step": 27490
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.215353884051607e-06,
+      "loss": 0.6188,
+      "step": 27491
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.214694085245918e-06,
+      "loss": 0.6596,
+      "step": 27492
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.214034305674481e-06,
+      "loss": 0.6128,
+      "step": 27493
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.213374545340641e-06,
+      "loss": 0.552,
+      "step": 27494
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2127148042477535e-06,
+      "loss": 0.5777,
+      "step": 27495
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.212055082399171e-06,
+      "loss": 0.5831,
+      "step": 27496
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.211395379798246e-06,
+      "loss": 0.6014,
+      "step": 27497
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2107356964483316e-06,
+      "loss": 0.5308,
+      "step": 27498
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.210076032352781e-06,
+      "loss": 0.6336,
+      "step": 27499
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.2094163875149414e-06,
+      "loss": 0.5855,
+      "step": 27500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.208756761938166e-06,
+      "loss": 0.6647,
+      "step": 27501
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.20809715562581e-06,
+      "loss": 0.5481,
+      "step": 27502
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.207437568581223e-06,
+      "loss": 0.5235,
+      "step": 27503
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.206778000807756e-06,
+      "loss": 0.6166,
+      "step": 27504
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.206118452308761e-06,
+      "loss": 0.6247,
+      "step": 27505
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.205458923087587e-06,
+      "loss": 0.5451,
+      "step": 27506
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.204799413147591e-06,
+      "loss": 0.6376,
+      "step": 27507
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.204139922492119e-06,
+      "loss": 0.5862,
+      "step": 27508
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.203480451124522e-06,
+      "loss": 0.581,
+      "step": 27509
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.202820999048158e-06,
+      "loss": 0.6243,
+      "step": 27510
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.202161566266367e-06,
+      "loss": 0.5871,
+      "step": 27511
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.201502152782507e-06,
+      "loss": 0.6045,
+      "step": 27512
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.200842758599927e-06,
+      "loss": 0.6337,
+      "step": 27513
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.200183383721976e-06,
+      "loss": 0.6984,
+      "step": 27514
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.199524028152012e-06,
+      "loss": 0.6181,
+      "step": 27515
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.198864691893374e-06,
+      "loss": 0.6149,
+      "step": 27516
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.1982053749494185e-06,
+      "loss": 0.6311,
+      "step": 27517
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.197546077323495e-06,
+      "loss": 0.6191,
+      "step": 27518
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.196886799018953e-06,
+      "loss": 0.6228,
+      "step": 27519
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.196227540039147e-06,
+      "loss": 0.599,
+      "step": 27520
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.195568300387419e-06,
+      "loss": 0.6112,
+      "step": 27521
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.194909080067122e-06,
+      "loss": 0.595,
+      "step": 27522
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.1942498790816075e-06,
+      "loss": 0.5989,
+      "step": 27523
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.193590697434225e-06,
+      "loss": 0.5321,
+      "step": 27524
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.1929315351283214e-06,
+      "loss": 0.6351,
+      "step": 27525
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.192272392167249e-06,
+      "loss": 0.5755,
+      "step": 27526
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.191613268554355e-06,
+      "loss": 0.6315,
+      "step": 27527
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.19095416429299e-06,
+      "loss": 0.5753,
+      "step": 27528
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.1902950793865014e-06,
+      "loss": 0.5522,
+      "step": 27529
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.189636013838238e-06,
+      "loss": 0.5738,
+      "step": 27530
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.188976967651555e-06,
+      "loss": 0.5564,
+      "step": 27531
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.188317940829792e-06,
+      "loss": 0.6665,
+      "step": 27532
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.187658933376301e-06,
+      "loss": 0.7069,
+      "step": 27533
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.186999945294434e-06,
+      "loss": 0.5927,
+      "step": 27534
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.186340976587535e-06,
+      "loss": 0.6133,
+      "step": 27535
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.185682027258955e-06,
+      "loss": 0.633,
+      "step": 27536
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.185023097312044e-06,
+      "loss": 0.6371,
+      "step": 27537
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.184364186750145e-06,
+      "loss": 0.6099,
+      "step": 27538
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.18370529557661e-06,
+      "loss": 0.5449,
+      "step": 27539
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.183046423794785e-06,
+      "loss": 0.5769,
+      "step": 27540
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.182387571408019e-06,
+      "loss": 0.5886,
+      "step": 27541
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.181728738419664e-06,
+      "loss": 0.5775,
+      "step": 27542
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.181069924833058e-06,
+      "loss": 0.5936,
+      "step": 27543
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.180411130651557e-06,
+      "loss": 0.5956,
+      "step": 27544
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.179752355878506e-06,
+      "loss": 0.56,
+      "step": 27545
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.1790936005172495e-06,
+      "loss": 0.5849,
+      "step": 27546
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.178434864571141e-06,
+      "loss": 0.6254,
+      "step": 27547
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.177776148043524e-06,
+      "loss": 0.513,
+      "step": 27548
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.177117450937743e-06,
+      "loss": 0.5541,
+      "step": 27549
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.176458773257148e-06,
+      "loss": 0.6826,
+      "step": 27550
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.175800115005086e-06,
+      "loss": 0.5563,
+      "step": 27551
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.175141476184903e-06,
+      "loss": 0.623,
+      "step": 27552
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.17448285679995e-06,
+      "loss": 0.6213,
+      "step": 27553
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.173824256853567e-06,
+      "loss": 0.6066,
+      "step": 27554
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.173165676349103e-06,
+      "loss": 0.6216,
+      "step": 27555
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.172507115289905e-06,
+      "loss": 0.5922,
+      "step": 27556
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.17184857367932e-06,
+      "loss": 0.5566,
+      "step": 27557
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1711900515206965e-06,
+      "loss": 0.6491,
+      "step": 27558
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.170531548817374e-06,
+      "loss": 0.5559,
+      "step": 27559
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.169873065572703e-06,
+      "loss": 0.5927,
+      "step": 27560
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.169214601790028e-06,
+      "loss": 0.602,
+      "step": 27561
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.168556157472695e-06,
+      "loss": 0.575,
+      "step": 27562
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.167897732624053e-06,
+      "loss": 0.5742,
+      "step": 27563
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.167239327247444e-06,
+      "loss": 0.6395,
+      "step": 27564
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.166580941346213e-06,
+      "loss": 0.6036,
+      "step": 27565
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.165922574923709e-06,
+      "loss": 0.5595,
+      "step": 27566
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.165264227983274e-06,
+      "loss": 0.6054,
+      "step": 27567
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.164605900528253e-06,
+      "loss": 0.6261,
+      "step": 27568
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.163947592561997e-06,
+      "loss": 0.5997,
+      "step": 27569
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.163289304087844e-06,
+      "loss": 0.6561,
+      "step": 27570
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.162631035109141e-06,
+      "loss": 0.5626,
+      "step": 27571
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.161972785629233e-06,
+      "loss": 0.5621,
+      "step": 27572
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.161314555651465e-06,
+      "loss": 0.5851,
+      "step": 27573
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.160656345179187e-06,
+      "loss": 0.5998,
+      "step": 27574
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1599981542157335e-06,
+      "loss": 0.5835,
+      "step": 27575
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.159339982764455e-06,
+      "loss": 0.6208,
+      "step": 27576
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.158681830828694e-06,
+      "loss": 0.6577,
+      "step": 27577
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.158023698411797e-06,
+      "loss": 0.5382,
+      "step": 27578
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.157365585517106e-06,
+      "loss": 0.5484,
+      "step": 27579
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1567074921479685e-06,
+      "loss": 0.6097,
+      "step": 27580
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.156049418307723e-06,
+      "loss": 0.6168,
+      "step": 27581
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.155391363999716e-06,
+      "loss": 0.5548,
+      "step": 27582
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1547333292272935e-06,
+      "loss": 0.5943,
+      "step": 27583
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.154075313993795e-06,
+      "loss": 0.6007,
+      "step": 27584
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.153417318302568e-06,
+      "loss": 0.6005,
+      "step": 27585
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.152759342156953e-06,
+      "loss": 0.6007,
+      "step": 27586
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.152101385560296e-06,
+      "loss": 0.6114,
+      "step": 27587
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.151443448515937e-06,
+      "loss": 0.6171,
+      "step": 27588
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.150785531027219e-06,
+      "loss": 0.5742,
+      "step": 27589
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.150127633097489e-06,
+      "loss": 0.6152,
+      "step": 27590
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.149469754730092e-06,
+      "loss": 0.5534,
+      "step": 27591
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.148811895928362e-06,
+      "loss": 0.6072,
+      "step": 27592
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.148154056695648e-06,
+      "loss": 0.6222,
+      "step": 27593
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.14749623703529e-06,
+      "loss": 0.6132,
+      "step": 27594
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.146838436950632e-06,
+      "loss": 0.618,
+      "step": 27595
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.14618065644502e-06,
+      "loss": 0.5608,
+      "step": 27596
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.14552289552179e-06,
+      "loss": 0.5696,
+      "step": 27597
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.144865154184286e-06,
+      "loss": 0.6391,
+      "step": 27598
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.144207432435851e-06,
+      "loss": 0.5752,
+      "step": 27599
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.143549730279828e-06,
+      "loss": 0.5643,
+      "step": 27600
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.142892047719562e-06,
+      "loss": 0.6407,
+      "step": 27601
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.142234384758388e-06,
+      "loss": 0.5757,
+      "step": 27602
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.141576741399649e-06,
+      "loss": 0.5626,
+      "step": 27603
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1409191176466924e-06,
+      "loss": 0.6079,
+      "step": 27604
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.140261513502852e-06,
+      "loss": 0.582,
+      "step": 27605
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.139603928971475e-06,
+      "loss": 0.6041,
+      "step": 27606
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.138946364055902e-06,
+      "loss": 0.6345,
+      "step": 27607
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1382888187594735e-06,
+      "loss": 0.5524,
+      "step": 27608
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1376312930855285e-06,
+      "loss": 0.5529,
+      "step": 27609
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.13697378703741e-06,
+      "loss": 0.588,
+      "step": 27610
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.136316300618459e-06,
+      "loss": 0.5731,
+      "step": 27611
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1356588338320214e-06,
+      "loss": 0.5931,
+      "step": 27612
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.135001386681427e-06,
+      "loss": 0.5964,
+      "step": 27613
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.134343959170024e-06,
+      "loss": 0.6338,
+      "step": 27614
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.133686551301153e-06,
+      "loss": 0.6199,
+      "step": 27615
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.13302916307815e-06,
+      "loss": 0.6252,
+      "step": 27616
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.132371794504359e-06,
+      "loss": 0.5636,
+      "step": 27617
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.131714445583124e-06,
+      "loss": 0.5688,
+      "step": 27618
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.131057116317777e-06,
+      "loss": 0.5756,
+      "step": 27619
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.130399806711661e-06,
+      "loss": 0.6323,
+      "step": 27620
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.129742516768118e-06,
+      "loss": 0.661,
+      "step": 27621
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.129085246490487e-06,
+      "loss": 0.6192,
+      "step": 27622
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.128427995882109e-06,
+      "loss": 0.5691,
+      "step": 27623
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.12777076494632e-06,
+      "loss": 0.5439,
+      "step": 27624
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.127113553686463e-06,
+      "loss": 0.6526,
+      "step": 27625
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.126456362105876e-06,
+      "loss": 0.6596,
+      "step": 27626
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.125799190207897e-06,
+      "loss": 0.5858,
+      "step": 27627
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1251420379958684e-06,
+      "loss": 0.5913,
+      "step": 27628
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.124484905473131e-06,
+      "loss": 0.6141,
+      "step": 27629
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.1238277926430155e-06,
+      "loss": 0.6326,
+      "step": 27630
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.123170699508868e-06,
+      "loss": 0.6436,
+      "step": 27631
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.122513626074025e-06,
+      "loss": 0.604,
+      "step": 27632
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.121856572341826e-06,
+      "loss": 0.5769,
+      "step": 27633
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.121199538315613e-06,
+      "loss": 0.6789,
+      "step": 27634
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.120542523998718e-06,
+      "loss": 0.5959,
+      "step": 27635
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.119885529394483e-06,
+      "loss": 0.6229,
+      "step": 27636
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.119228554506244e-06,
+      "loss": 0.6005,
+      "step": 27637
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.118571599337342e-06,
+      "loss": 0.6274,
+      "step": 27638
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.117914663891119e-06,
+      "loss": 0.6579,
+      "step": 27639
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.117257748170904e-06,
+      "loss": 0.6011,
+      "step": 27640
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.116600852180039e-06,
+      "loss": 0.5866,
+      "step": 27641
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.115943975921864e-06,
+      "loss": 0.5869,
+      "step": 27642
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.115287119399716e-06,
+      "loss": 0.5881,
+      "step": 27643
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.11463028261693e-06,
+      "loss": 0.6178,
+      "step": 27644
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.113973465576848e-06,
+      "loss": 0.6305,
+      "step": 27645
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.113316668282805e-06,
+      "loss": 0.582,
+      "step": 27646
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.112659890738135e-06,
+      "loss": 0.5773,
+      "step": 27647
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.112003132946179e-06,
+      "loss": 0.6469,
+      "step": 27648
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.111346394910274e-06,
+      "loss": 0.5847,
+      "step": 27649
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.110689676633761e-06,
+      "loss": 0.6205,
+      "step": 27650
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.110032978119969e-06,
+      "loss": 0.6034,
+      "step": 27651
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.109376299372238e-06,
+      "loss": 0.6512,
+      "step": 27652
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.108719640393906e-06,
+      "loss": 0.7287,
+      "step": 27653
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.108063001188308e-06,
+      "loss": 0.5916,
+      "step": 27654
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.107406381758783e-06,
+      "loss": 0.58,
+      "step": 27655
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.10674978210867e-06,
+      "loss": 0.5951,
+      "step": 27656
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.106093202241296e-06,
+      "loss": 0.5603,
+      "step": 27657
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.105436642160005e-06,
+      "loss": 0.6118,
+      "step": 27658
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.104780101868129e-06,
+      "loss": 0.6429,
+      "step": 27659
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.10412358136901e-06,
+      "loss": 0.6393,
+      "step": 27660
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.103467080665979e-06,
+      "loss": 0.6203,
+      "step": 27661
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.102810599762371e-06,
+      "loss": 0.6004,
+      "step": 27662
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.102154138661524e-06,
+      "loss": 0.6399,
+      "step": 27663
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.101497697366775e-06,
+      "loss": 0.5997,
+      "step": 27664
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.100841275881456e-06,
+      "loss": 0.6296,
+      "step": 27665
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.100184874208907e-06,
+      "loss": 0.6125,
+      "step": 27666
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.09952849235246e-06,
+      "loss": 0.6067,
+      "step": 27667
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.09887213031545e-06,
+      "loss": 0.6146,
+      "step": 27668
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0982157881012115e-06,
+      "loss": 0.5872,
+      "step": 27669
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.097559465713082e-06,
+      "loss": 0.6255,
+      "step": 27670
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.096903163154396e-06,
+      "loss": 0.6271,
+      "step": 27671
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.096246880428493e-06,
+      "loss": 0.6342,
+      "step": 27672
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.095590617538697e-06,
+      "loss": 0.584,
+      "step": 27673
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0949343744883495e-06,
+      "loss": 0.5868,
+      "step": 27674
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.094278151280783e-06,
+      "loss": 0.5767,
+      "step": 27675
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0936219479193345e-06,
+      "loss": 0.5412,
+      "step": 27676
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.092965764407341e-06,
+      "loss": 0.6323,
+      "step": 27677
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0923096007481285e-06,
+      "loss": 0.6286,
+      "step": 27678
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.091653456945035e-06,
+      "loss": 0.5934,
+      "step": 27679
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.090997333001395e-06,
+      "loss": 0.6404,
+      "step": 27680
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.090341228920544e-06,
+      "loss": 0.5181,
+      "step": 27681
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.089685144705813e-06,
+      "loss": 0.5942,
+      "step": 27682
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.089029080360538e-06,
+      "loss": 0.6235,
+      "step": 27683
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.088373035888054e-06,
+      "loss": 0.6167,
+      "step": 27684
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.087717011291687e-06,
+      "loss": 0.6183,
+      "step": 27685
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.087061006574777e-06,
+      "loss": 0.587,
+      "step": 27686
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.086405021740656e-06,
+      "loss": 0.5576,
+      "step": 27687
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.08574905679266e-06,
+      "loss": 0.5788,
+      "step": 27688
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.085093111734117e-06,
+      "loss": 0.5806,
+      "step": 27689
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.084437186568361e-06,
+      "loss": 0.6036,
+      "step": 27690
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0837812812987285e-06,
+      "loss": 0.6199,
+      "step": 27691
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.083125395928548e-06,
+      "loss": 0.5675,
+      "step": 27692
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.082469530461161e-06,
+      "loss": 0.5828,
+      "step": 27693
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.081813684899889e-06,
+      "loss": 0.6241,
+      "step": 27694
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.081157859248068e-06,
+      "loss": 0.5491,
+      "step": 27695
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.080502053509032e-06,
+      "loss": 0.6631,
+      "step": 27696
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.079846267686111e-06,
+      "loss": 0.6952,
+      "step": 27697
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.079190501782641e-06,
+      "loss": 0.5933,
+      "step": 27698
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.078534755801957e-06,
+      "loss": 0.6067,
+      "step": 27699
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.077879029747382e-06,
+      "loss": 0.5915,
+      "step": 27700
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0772233236222525e-06,
+      "loss": 0.5792,
+      "step": 27701
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.076567637429902e-06,
+      "loss": 0.6299,
+      "step": 27702
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.075911971173658e-06,
+      "loss": 0.5819,
+      "step": 27703
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.075256324856856e-06,
+      "loss": 0.64,
+      "step": 27704
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.074600698482827e-06,
+      "loss": 0.6464,
+      "step": 27705
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.073945092054898e-06,
+      "loss": 0.6191,
+      "step": 27706
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.073289505576405e-06,
+      "loss": 0.6059,
+      "step": 27707
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0726339390506785e-06,
+      "loss": 0.6377,
+      "step": 27708
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.071978392481049e-06,
+      "loss": 0.6071,
+      "step": 27709
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.07132286587085e-06,
+      "loss": 0.528,
+      "step": 27710
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0706673592234075e-06,
+      "loss": 0.5875,
+      "step": 27711
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.070011872542054e-06,
+      "loss": 0.6206,
+      "step": 27712
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.069356405830123e-06,
+      "loss": 0.6264,
+      "step": 27713
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.068700959090941e-06,
+      "loss": 0.5675,
+      "step": 27714
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.068045532327846e-06,
+      "loss": 0.5748,
+      "step": 27715
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.06739012554416e-06,
+      "loss": 0.5572,
+      "step": 27716
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.066734738743214e-06,
+      "loss": 0.6717,
+      "step": 27717
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.066079371928341e-06,
+      "loss": 0.619,
+      "step": 27718
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.065424025102873e-06,
+      "loss": 0.6223,
+      "step": 27719
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.06476869827014e-06,
+      "loss": 0.5608,
+      "step": 27720
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.064113391433468e-06,
+      "loss": 0.6396,
+      "step": 27721
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.063458104596187e-06,
+      "loss": 0.6252,
+      "step": 27722
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.06280283776163e-06,
+      "loss": 0.634,
+      "step": 27723
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.062147590933123e-06,
+      "loss": 0.5846,
+      "step": 27724
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.061492364113996e-06,
+      "loss": 0.6137,
+      "step": 27725
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.060837157307586e-06,
+      "loss": 0.6156,
+      "step": 27726
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0601819705172115e-06,
+      "loss": 0.6,
+      "step": 27727
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0595268037462065e-06,
+      "loss": 0.6474,
+      "step": 27728
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.058871656997899e-06,
+      "loss": 0.6768,
+      "step": 27729
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.058216530275619e-06,
+      "loss": 0.5872,
+      "step": 27730
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0575614235827e-06,
+      "loss": 0.6197,
+      "step": 27731
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.056906336922461e-06,
+      "loss": 0.6177,
+      "step": 27732
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.056251270298236e-06,
+      "loss": 0.5315,
+      "step": 27733
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.055596223713354e-06,
+      "loss": 0.5703,
+      "step": 27734
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.05494119717114e-06,
+      "loss": 0.6495,
+      "step": 27735
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0542861906749284e-06,
+      "loss": 0.6032,
+      "step": 27736
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.053631204228048e-06,
+      "loss": 0.5982,
+      "step": 27737
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.052976237833816e-06,
+      "loss": 0.6253,
+      "step": 27738
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.052321291495571e-06,
+      "loss": 0.6325,
+      "step": 27739
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.051666365216639e-06,
+      "loss": 0.5903,
+      "step": 27740
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.051011459000343e-06,
+      "loss": 0.6089,
+      "step": 27741
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.050356572850018e-06,
+      "loss": 0.6147,
+      "step": 27742
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.049701706768985e-06,
+      "loss": 0.5558,
+      "step": 27743
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0490468607605765e-06,
+      "loss": 0.5454,
+      "step": 27744
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.048392034828116e-06,
+      "loss": 0.631,
+      "step": 27745
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.047737228974935e-06,
+      "loss": 0.6322,
+      "step": 27746
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.047082443204355e-06,
+      "loss": 0.592,
+      "step": 27747
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.046427677519713e-06,
+      "loss": 0.5756,
+      "step": 27748
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0457729319243255e-06,
+      "loss": 0.5584,
+      "step": 27749
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.045118206421524e-06,
+      "loss": 0.6313,
+      "step": 27750
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.044463501014635e-06,
+      "loss": 0.5648,
+      "step": 27751
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.043808815706985e-06,
+      "loss": 0.6172,
+      "step": 27752
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.043154150501905e-06,
+      "loss": 0.595,
+      "step": 27753
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0424995054027146e-06,
+      "loss": 0.5618,
+      "step": 27754
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.041844880412742e-06,
+      "loss": 0.559,
+      "step": 27755
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.041190275535316e-06,
+      "loss": 0.5834,
+      "step": 27756
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.04053569077376e-06,
+      "loss": 0.5396,
+      "step": 27757
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.039881126131406e-06,
+      "loss": 0.6014,
+      "step": 27758
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0392265816115725e-06,
+      "loss": 0.5668,
+      "step": 27759
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.038572057217589e-06,
+      "loss": 0.5733,
+      "step": 27760
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.037917552952782e-06,
+      "loss": 0.577,
+      "step": 27761
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.037263068820475e-06,
+      "loss": 0.6299,
+      "step": 27762
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.036608604823995e-06,
+      "loss": 0.5656,
+      "step": 27763
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.035954160966668e-06,
+      "loss": 0.6202,
+      "step": 27764
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.035299737251821e-06,
+      "loss": 0.5832,
+      "step": 27765
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.034645333682774e-06,
+      "loss": 0.6207,
+      "step": 27766
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.033990950262856e-06,
+      "loss": 0.5805,
+      "step": 27767
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.033336586995391e-06,
+      "loss": 0.5984,
+      "step": 27768
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.032682243883708e-06,
+      "loss": 0.6047,
+      "step": 27769
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.032027920931125e-06,
+      "loss": 0.5529,
+      "step": 27770
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.03137361814097e-06,
+      "loss": 0.6026,
+      "step": 27771
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.030719335516567e-06,
+      "loss": 0.5898,
+      "step": 27772
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.030065073061243e-06,
+      "loss": 0.5756,
+      "step": 27773
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.029410830778321e-06,
+      "loss": 0.602,
+      "step": 27774
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.028756608671129e-06,
+      "loss": 0.5591,
+      "step": 27775
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.028102406742983e-06,
+      "loss": 0.5613,
+      "step": 27776
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0274482249972124e-06,
+      "loss": 0.5654,
+      "step": 27777
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.02679406343714e-06,
+      "loss": 0.6623,
+      "step": 27778
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.026139922066094e-06,
+      "loss": 0.5313,
+      "step": 27779
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.025485800887394e-06,
+      "loss": 0.5718,
+      "step": 27780
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.024831699904363e-06,
+      "loss": 0.5933,
+      "step": 27781
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.024177619120328e-06,
+      "loss": 0.5246,
+      "step": 27782
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.02352355853861e-06,
+      "loss": 0.6262,
+      "step": 27783
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.022869518162533e-06,
+      "loss": 0.6124,
+      "step": 27784
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.022215497995424e-06,
+      "loss": 0.6698,
+      "step": 27785
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.021561498040599e-06,
+      "loss": 0.6048,
+      "step": 27786
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.020907518301385e-06,
+      "loss": 0.652,
+      "step": 27787
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.020253558781106e-06,
+      "loss": 0.6321,
+      "step": 27788
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.019599619483083e-06,
+      "loss": 0.5882,
+      "step": 27789
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.018945700410641e-06,
+      "loss": 0.5826,
+      "step": 27790
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.018291801567106e-06,
+      "loss": 0.6243,
+      "step": 27791
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.017637922955793e-06,
+      "loss": 0.631,
+      "step": 27792
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.016984064580027e-06,
+      "loss": 0.6495,
+      "step": 27793
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.016330226443131e-06,
+      "loss": 0.598,
+      "step": 27794
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.015676408548429e-06,
+      "loss": 0.6347,
+      "step": 27795
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0150226108992446e-06,
+      "loss": 0.6331,
+      "step": 27796
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.014368833498895e-06,
+      "loss": 0.5336,
+      "step": 27797
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.013715076350704e-06,
+      "loss": 0.599,
+      "step": 27798
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.013061339457993e-06,
+      "loss": 0.5652,
+      "step": 27799
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.012407622824088e-06,
+      "loss": 0.6558,
+      "step": 27800
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.011753926452306e-06,
+      "loss": 0.6444,
+      "step": 27801
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.011100250345971e-06,
+      "loss": 0.6211,
+      "step": 27802
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.010446594508406e-06,
+      "loss": 0.6039,
+      "step": 27803
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.009792958942926e-06,
+      "loss": 0.6085,
+      "step": 27804
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.009139343652858e-06,
+      "loss": 0.5863,
+      "step": 27805
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.008485748641521e-06,
+      "loss": 0.6068,
+      "step": 27806
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.007832173912242e-06,
+      "loss": 0.5829,
+      "step": 27807
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.007178619468332e-06,
+      "loss": 0.635,
+      "step": 27808
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.006525085313116e-06,
+      "loss": 0.6318,
+      "step": 27809
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0058715714499175e-06,
+      "loss": 0.5564,
+      "step": 27810
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.0052180778820535e-06,
+      "loss": 0.6129,
+      "step": 27811
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.004564604612847e-06,
+      "loss": 0.5923,
+      "step": 27812
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.003911151645622e-06,
+      "loss": 0.6322,
+      "step": 27813
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.003257718983691e-06,
+      "loss": 0.6327,
+      "step": 27814
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.002604306630377e-06,
+      "loss": 0.5994,
+      "step": 27815
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.001950914589002e-06,
+      "loss": 0.5458,
+      "step": 27816
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.001297542862887e-06,
+      "loss": 0.6023,
+      "step": 27817
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.000644191455352e-06,
+      "loss": 0.6318,
+      "step": 27818
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.999990860369711e-06,
+      "loss": 0.5798,
+      "step": 27819
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.9993375496092895e-06,
+      "loss": 0.5994,
+      "step": 27820
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.9986842591774056e-06,
+      "loss": 0.5714,
+      "step": 27821
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.998030989077377e-06,
+      "loss": 0.601,
+      "step": 27822
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.997377739312529e-06,
+      "loss": 0.5921,
+      "step": 27823
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.996724509886175e-06,
+      "loss": 0.6434,
+      "step": 27824
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.996071300801635e-06,
+      "loss": 0.6732,
+      "step": 27825
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.995418112062228e-06,
+      "loss": 0.5453,
+      "step": 27826
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.994764943671276e-06,
+      "loss": 0.6206,
+      "step": 27827
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.994111795632096e-06,
+      "loss": 0.6063,
+      "step": 27828
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.9934586679480115e-06,
+      "loss": 0.6425,
+      "step": 27829
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.992805560622332e-06,
+      "loss": 0.6405,
+      "step": 27830
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.992152473658382e-06,
+      "loss": 0.6207,
+      "step": 27831
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.991499407059478e-06,
+      "loss": 0.6369,
+      "step": 27832
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.990846360828939e-06,
+      "loss": 0.5835,
+      "step": 27833
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.9901933349700895e-06,
+      "loss": 0.6217,
+      "step": 27834
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.989540329486237e-06,
+      "loss": 0.6078,
+      "step": 27835
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.988887344380705e-06,
+      "loss": 0.5626,
+      "step": 27836
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.98823437965681e-06,
+      "loss": 0.6145,
+      "step": 27837
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.987581435317875e-06,
+      "loss": 0.605,
+      "step": 27838
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.986928511367211e-06,
+      "loss": 0.6508,
+      "step": 27839
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.98627560780814e-06,
+      "loss": 0.5732,
+      "step": 27840
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.985622724643978e-06,
+      "loss": 0.6244,
+      "step": 27841
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.98496986187804e-06,
+      "loss": 0.5419,
+      "step": 27842
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.984317019513648e-06,
+      "loss": 0.599,
+      "step": 27843
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.983664197554116e-06,
+      "loss": 0.5912,
+      "step": 27844
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.983011396002767e-06,
+      "loss": 0.6151,
+      "step": 27845
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9823586148629086e-06,
+      "loss": 0.5747,
+      "step": 27846
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.981705854137864e-06,
+      "loss": 0.5653,
+      "step": 27847
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.981053113830948e-06,
+      "loss": 0.6536,
+      "step": 27848
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.980400393945478e-06,
+      "loss": 0.6126,
+      "step": 27849
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.979747694484774e-06,
+      "loss": 0.547,
+      "step": 27850
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.979095015452147e-06,
+      "loss": 0.5729,
+      "step": 27851
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.978442356850915e-06,
+      "loss": 0.5732,
+      "step": 27852
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.977789718684395e-06,
+      "loss": 0.6104,
+      "step": 27853
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.977137100955903e-06,
+      "loss": 0.6411,
+      "step": 27854
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.976484503668757e-06,
+      "loss": 0.6376,
+      "step": 27855
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.975831926826274e-06,
+      "loss": 0.6447,
+      "step": 27856
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.975179370431765e-06,
+      "loss": 0.5453,
+      "step": 27857
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.974526834488547e-06,
+      "loss": 0.5424,
+      "step": 27858
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.97387431899994e-06,
+      "loss": 0.5466,
+      "step": 27859
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9732218239692555e-06,
+      "loss": 0.6167,
+      "step": 27860
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9725693493998104e-06,
+      "loss": 0.5764,
+      "step": 27861
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.97191689529492e-06,
+      "loss": 0.6179,
+      "step": 27862
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9712644616578994e-06,
+      "loss": 0.6045,
+      "step": 27863
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9706120484920615e-06,
+      "loss": 0.5347,
+      "step": 27864
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.969959655800726e-06,
+      "loss": 0.6479,
+      "step": 27865
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.969307283587205e-06,
+      "loss": 0.6092,
+      "step": 27866
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.968654931854818e-06,
+      "loss": 0.6116,
+      "step": 27867
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9680026006068716e-06,
+      "loss": 0.5924,
+      "step": 27868
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.967350289846685e-06,
+      "loss": 0.5893,
+      "step": 27869
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.966697999577573e-06,
+      "loss": 0.6158,
+      "step": 27870
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.966045729802849e-06,
+      "loss": 0.5554,
+      "step": 27871
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.965393480525832e-06,
+      "loss": 0.6068,
+      "step": 27872
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.964741251749828e-06,
+      "loss": 0.6329,
+      "step": 27873
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.964089043478156e-06,
+      "loss": 0.5515,
+      "step": 27874
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.963436855714131e-06,
+      "loss": 0.5923,
+      "step": 27875
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.962784688461065e-06,
+      "loss": 0.5809,
+      "step": 27876
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9621325417222746e-06,
+      "loss": 0.6111,
+      "step": 27877
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9614804155010685e-06,
+      "loss": 0.6633,
+      "step": 27878
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.960828309800763e-06,
+      "loss": 0.7011,
+      "step": 27879
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.960176224624673e-06,
+      "loss": 0.5762,
+      "step": 27880
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.959524159976109e-06,
+      "loss": 0.6288,
+      "step": 27881
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.958872115858387e-06,
+      "loss": 0.5923,
+      "step": 27882
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.958220092274822e-06,
+      "loss": 0.6086,
+      "step": 27883
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.957568089228723e-06,
+      "loss": 0.591,
+      "step": 27884
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.956916106723402e-06,
+      "loss": 0.5749,
+      "step": 27885
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.956264144762174e-06,
+      "loss": 0.5779,
+      "step": 27886
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.955612203348353e-06,
+      "loss": 0.5964,
+      "step": 27887
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.954960282485256e-06,
+      "loss": 0.6277,
+      "step": 27888
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.954308382176184e-06,
+      "loss": 0.5852,
+      "step": 27889
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.953656502424459e-06,
+      "loss": 0.6412,
+      "step": 27890
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.953004643233388e-06,
+      "loss": 0.5754,
+      "step": 27891
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.952352804606285e-06,
+      "loss": 0.6461,
+      "step": 27892
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.951700986546465e-06,
+      "loss": 0.5379,
+      "step": 27893
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.951049189057241e-06,
+      "loss": 0.6,
+      "step": 27894
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.950397412141917e-06,
+      "loss": 0.5533,
+      "step": 27895
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9497456558038115e-06,
+      "loss": 0.5566,
+      "step": 27896
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.949093920046235e-06,
+      "loss": 0.5347,
+      "step": 27897
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.948442204872498e-06,
+      "loss": 0.6006,
+      "step": 27898
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.947790510285913e-06,
+      "loss": 0.6231,
+      "step": 27899
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9471388362897895e-06,
+      "loss": 0.5504,
+      "step": 27900
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.946487182887444e-06,
+      "loss": 0.652,
+      "step": 27901
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.94583555008218e-06,
+      "loss": 0.6602,
+      "step": 27902
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.945183937877314e-06,
+      "loss": 0.543,
+      "step": 27903
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.944532346276157e-06,
+      "loss": 0.5625,
+      "step": 27904
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.943880775282022e-06,
+      "loss": 0.6404,
+      "step": 27905
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.943229224898213e-06,
+      "loss": 0.5995,
+      "step": 27906
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.942577695128045e-06,
+      "loss": 0.5991,
+      "step": 27907
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.941926185974826e-06,
+      "loss": 0.642,
+      "step": 27908
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.94127469744187e-06,
+      "loss": 0.6187,
+      "step": 27909
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.94062322953249e-06,
+      "loss": 0.6149,
+      "step": 27910
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.939971782249987e-06,
+      "loss": 0.5909,
+      "step": 27911
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.939320355597676e-06,
+      "loss": 0.6248,
+      "step": 27912
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.938668949578869e-06,
+      "loss": 0.6244,
+      "step": 27913
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.938017564196874e-06,
+      "loss": 0.5998,
+      "step": 27914
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.937366199455006e-06,
+      "loss": 0.5574,
+      "step": 27915
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.936714855356565e-06,
+      "loss": 0.6114,
+      "step": 27916
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.936063531904866e-06,
+      "loss": 0.6288,
+      "step": 27917
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.935412229103221e-06,
+      "loss": 0.5878,
+      "step": 27918
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.934760946954935e-06,
+      "loss": 0.6468,
+      "step": 27919
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9341096854633185e-06,
+      "loss": 0.5782,
+      "step": 27920
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9334584446316855e-06,
+      "loss": 0.6203,
+      "step": 27921
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.932807224463337e-06,
+      "loss": 0.6029,
+      "step": 27922
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.932156024961588e-06,
+      "loss": 0.6049,
+      "step": 27923
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.931504846129743e-06,
+      "loss": 0.6637,
+      "step": 27924
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9308536879711145e-06,
+      "loss": 0.601,
+      "step": 27925
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.930202550489014e-06,
+      "loss": 0.6056,
+      "step": 27926
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.929551433686742e-06,
+      "loss": 0.6461,
+      "step": 27927
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.928900337567611e-06,
+      "loss": 0.6031,
+      "step": 27928
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.928249262134931e-06,
+      "loss": 0.6545,
+      "step": 27929
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.927598207392007e-06,
+      "loss": 0.5451,
+      "step": 27930
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.92694717334215e-06,
+      "loss": 0.5704,
+      "step": 27931
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.926296159988673e-06,
+      "loss": 0.6053,
+      "step": 27932
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9256451673348725e-06,
+      "loss": 0.6834,
+      "step": 27933
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.924994195384062e-06,
+      "loss": 0.6295,
+      "step": 27934
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9243432441395495e-06,
+      "loss": 0.6422,
+      "step": 27935
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.923692313604645e-06,
+      "loss": 0.5182,
+      "step": 27936
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9230414037826545e-06,
+      "loss": 0.6181,
+      "step": 27937
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.922390514676881e-06,
+      "loss": 0.6528,
+      "step": 27938
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.921739646290639e-06,
+      "loss": 0.5912,
+      "step": 27939
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.921088798627228e-06,
+      "loss": 0.5804,
+      "step": 27940
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.920437971689961e-06,
+      "loss": 0.5894,
+      "step": 27941
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.919787165482147e-06,
+      "loss": 0.5594,
+      "step": 27942
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.919136380007086e-06,
+      "loss": 0.6045,
+      "step": 27943
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.918485615268088e-06,
+      "loss": 0.5861,
+      "step": 27944
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.91783487126846e-06,
+      "loss": 0.6466,
+      "step": 27945
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.917184148011508e-06,
+      "loss": 0.6151,
+      "step": 27946
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.916533445500538e-06,
+      "loss": 0.6046,
+      "step": 27947
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.915882763738864e-06,
+      "loss": 0.5983,
+      "step": 27948
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.91523210272978e-06,
+      "loss": 0.5833,
+      "step": 27949
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.914581462476598e-06,
+      "loss": 0.6341,
+      "step": 27950
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.913930842982624e-06,
+      "loss": 0.6537,
+      "step": 27951
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.913280244251165e-06,
+      "loss": 0.596,
+      "step": 27952
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.9126296662855295e-06,
+      "loss": 0.6151,
+      "step": 27953
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.911979109089016e-06,
+      "loss": 0.6113,
+      "step": 27954
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.911328572664934e-06,
+      "loss": 0.6291,
+      "step": 27955
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.910678057016588e-06,
+      "loss": 0.6127,
+      "step": 27956
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.910027562147287e-06,
+      "loss": 0.5901,
+      "step": 27957
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.909377088060332e-06,
+      "loss": 0.6091,
+      "step": 27958
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.908726634759032e-06,
+      "loss": 0.5645,
+      "step": 27959
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.90807620224669e-06,
+      "loss": 0.5868,
+      "step": 27960
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.907425790526609e-06,
+      "loss": 0.5751,
+      "step": 27961
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.906775399602096e-06,
+      "loss": 0.5862,
+      "step": 27962
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.906125029476456e-06,
+      "loss": 0.5986,
+      "step": 27963
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.905474680152998e-06,
+      "loss": 0.5469,
+      "step": 27964
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.904824351635018e-06,
+      "loss": 0.6044,
+      "step": 27965
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.904174043925824e-06,
+      "loss": 0.5487,
+      "step": 27966
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.903523757028722e-06,
+      "loss": 0.6456,
+      "step": 27967
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.902873490947017e-06,
+      "loss": 0.54,
+      "step": 27968
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.902223245684013e-06,
+      "loss": 0.6527,
+      "step": 27969
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.90157302124301e-06,
+      "loss": 0.6432,
+      "step": 27970
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.900922817627315e-06,
+      "loss": 0.6339,
+      "step": 27971
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.900272634840232e-06,
+      "loss": 0.6102,
+      "step": 27972
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.899622472885063e-06,
+      "loss": 0.6566,
+      "step": 27973
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.898972331765116e-06,
+      "loss": 0.5789,
+      "step": 27974
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8983222114836915e-06,
+      "loss": 0.5877,
+      "step": 27975
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.897672112044091e-06,
+      "loss": 0.6207,
+      "step": 27976
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.897022033449622e-06,
+      "loss": 0.6005,
+      "step": 27977
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.896371975703584e-06,
+      "loss": 0.5652,
+      "step": 27978
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.895721938809282e-06,
+      "loss": 0.6137,
+      "step": 27979
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.895071922770019e-06,
+      "loss": 0.6021,
+      "step": 27980
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8944219275891e-06,
+      "loss": 0.555,
+      "step": 27981
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.893771953269822e-06,
+      "loss": 0.6304,
+      "step": 27982
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.893121999815492e-06,
+      "loss": 0.6002,
+      "step": 27983
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8924720672294124e-06,
+      "loss": 0.564,
+      "step": 27984
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8918221555148835e-06,
+      "loss": 0.5811,
+      "step": 27985
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8911722646752155e-06,
+      "loss": 0.5617,
+      "step": 27986
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.890522394713699e-06,
+      "loss": 0.6372,
+      "step": 27987
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.889872545633642e-06,
+      "loss": 0.6432,
+      "step": 27988
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.889222717438347e-06,
+      "loss": 0.6051,
+      "step": 27989
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.888572910131114e-06,
+      "loss": 0.5975,
+      "step": 27990
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.887923123715251e-06,
+      "loss": 0.5631,
+      "step": 27991
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.8872733581940515e-06,
+      "loss": 0.5819,
+      "step": 27992
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8866236135708186e-06,
+      "loss": 0.6311,
+      "step": 27993
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.885973889848857e-06,
+      "loss": 0.5791,
+      "step": 27994
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.885324187031469e-06,
+      "loss": 0.6188,
+      "step": 27995
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.884674505121952e-06,
+      "loss": 0.5808,
+      "step": 27996
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.88402484412361e-06,
+      "loss": 0.6108,
+      "step": 27997
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.883375204039745e-06,
+      "loss": 0.5296,
+      "step": 27998
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.882725584873653e-06,
+      "loss": 0.5626,
+      "step": 27999
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.882075986628639e-06,
+      "loss": 0.6122,
+      "step": 28000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.881426409308002e-06,
+      "loss": 0.6463,
+      "step": 28001
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.880776852915046e-06,
+      "loss": 0.6178,
+      "step": 28002
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.880127317453068e-06,
+      "loss": 0.6404,
+      "step": 28003
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.879477802925368e-06,
+      "loss": 0.5974,
+      "step": 28004
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.878828309335249e-06,
+      "loss": 0.5923,
+      "step": 28005
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.87817883668601e-06,
+      "loss": 0.5686,
+      "step": 28006
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8775293849809556e-06,
+      "loss": 0.6287,
+      "step": 28007
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.876879954223378e-06,
+      "loss": 0.6211,
+      "step": 28008
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8762305444165805e-06,
+      "loss": 0.5976,
+      "step": 28009
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.875581155563863e-06,
+      "loss": 0.5999,
+      "step": 28010
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8749317876685255e-06,
+      "loss": 0.5848,
+      "step": 28011
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.87428244073387e-06,
+      "loss": 0.5963,
+      "step": 28012
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.873633114763193e-06,
+      "loss": 0.6368,
+      "step": 28013
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.872983809759793e-06,
+      "loss": 0.6043,
+      "step": 28014
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.872334525726972e-06,
+      "loss": 0.6265,
+      "step": 28015
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.87168526266803e-06,
+      "loss": 0.5697,
+      "step": 28016
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.87103602058626e-06,
+      "loss": 0.6277,
+      "step": 28017
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.870386799484969e-06,
+      "loss": 0.6038,
+      "step": 28018
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.869737599367452e-06,
+      "loss": 0.6479,
+      "step": 28019
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.869088420237006e-06,
+      "loss": 0.5627,
+      "step": 28020
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.868439262096931e-06,
+      "loss": 0.6007,
+      "step": 28021
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.867790124950525e-06,
+      "loss": 0.5758,
+      "step": 28022
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.86714100880109e-06,
+      "loss": 0.5612,
+      "step": 28023
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.866491913651923e-06,
+      "loss": 0.5305,
+      "step": 28024
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.865842839506321e-06,
+      "loss": 0.5917,
+      "step": 28025
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.86519378636758e-06,
+      "loss": 0.5944,
+      "step": 28026
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8645447542389995e-06,
+      "loss": 0.6665,
+      "step": 28027
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.86389574312388e-06,
+      "loss": 0.648,
+      "step": 28028
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.86324675302552e-06,
+      "loss": 0.6778,
+      "step": 28029
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.862597783947211e-06,
+      "loss": 0.544,
+      "step": 28030
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.861948835892255e-06,
+      "loss": 0.5948,
+      "step": 28031
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.861299908863949e-06,
+      "loss": 0.575,
+      "step": 28032
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.860651002865592e-06,
+      "loss": 0.5973,
+      "step": 28033
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.860002117900479e-06,
+      "loss": 0.5583,
+      "step": 28034
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.859353253971907e-06,
+      "loss": 0.6153,
+      "step": 28035
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.858704411083171e-06,
+      "loss": 0.5586,
+      "step": 28036
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.858055589237575e-06,
+      "loss": 0.6105,
+      "step": 28037
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.857406788438409e-06,
+      "loss": 0.6147,
+      "step": 28038
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.856758008688973e-06,
+      "loss": 0.61,
+      "step": 28039
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8561092499925665e-06,
+      "loss": 0.6127,
+      "step": 28040
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8554605123524775e-06,
+      "loss": 0.5383,
+      "step": 28041
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.854811795772009e-06,
+      "loss": 0.6034,
+      "step": 28042
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.854163100254455e-06,
+      "loss": 0.7149,
+      "step": 28043
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.853514425803113e-06,
+      "loss": 0.6043,
+      "step": 28044
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.852865772421281e-06,
+      "loss": 0.6007,
+      "step": 28045
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.852217140112251e-06,
+      "loss": 0.6713,
+      "step": 28046
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.851568528879319e-06,
+      "loss": 0.6332,
+      "step": 28047
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.850919938725782e-06,
+      "loss": 0.6333,
+      "step": 28048
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.850271369654936e-06,
+      "loss": 0.612,
+      "step": 28049
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.849622821670077e-06,
+      "loss": 0.5879,
+      "step": 28050
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.848974294774504e-06,
+      "loss": 0.5971,
+      "step": 28051
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.848325788971503e-06,
+      "loss": 0.5922,
+      "step": 28052
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8476773042643765e-06,
+      "loss": 0.5821,
+      "step": 28053
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.847028840656418e-06,
+      "loss": 0.6246,
+      "step": 28054
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.846380398150922e-06,
+      "loss": 0.5923,
+      "step": 28055
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.845731976751183e-06,
+      "loss": 0.5881,
+      "step": 28056
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.845083576460499e-06,
+      "loss": 0.6323,
+      "step": 28057
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8444351972821585e-06,
+      "loss": 0.589,
+      "step": 28058
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.84378683921946e-06,
+      "loss": 0.6494,
+      "step": 28059
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.843138502275698e-06,
+      "loss": 0.565,
+      "step": 28060
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.842490186454171e-06,
+      "loss": 0.5876,
+      "step": 28061
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.841841891758164e-06,
+      "loss": 0.5738,
+      "step": 28062
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8411936181909776e-06,
+      "loss": 0.6311,
+      "step": 28063
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8405453657559044e-06,
+      "loss": 0.5435,
+      "step": 28064
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.839897134456237e-06,
+      "loss": 0.6116,
+      "step": 28065
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.839248924295271e-06,
+      "loss": 0.583,
+      "step": 28066
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.838600735276305e-06,
+      "loss": 0.5964,
+      "step": 28067
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.837952567402624e-06,
+      "loss": 0.6199,
+      "step": 28068
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.837304420677521e-06,
+      "loss": 0.6078,
+      "step": 28069
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.836656295104297e-06,
+      "loss": 0.5846,
+      "step": 28070
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.83600819068624e-06,
+      "loss": 0.6304,
+      "step": 28071
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.835360107426651e-06,
+      "loss": 0.5708,
+      "step": 28072
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.834712045328809e-06,
+      "loss": 0.596,
+      "step": 28073
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8340640043960184e-06,
+      "loss": 0.5953,
+      "step": 28074
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.833415984631565e-06,
+      "loss": 0.6408,
+      "step": 28075
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.83276798603875e-06,
+      "loss": 0.5832,
+      "step": 28076
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.832120008620856e-06,
+      "loss": 0.6087,
+      "step": 28077
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.831472052381191e-06,
+      "loss": 0.6009,
+      "step": 28078
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.830824117323028e-06,
+      "loss": 0.6251,
+      "step": 28079
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.830176203449672e-06,
+      "loss": 0.6131,
+      "step": 28080
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.829528310764409e-06,
+      "loss": 0.5941,
+      "step": 28081
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8288804392705365e-06,
+      "loss": 0.5777,
+      "step": 28082
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.828232588971346e-06,
+      "loss": 0.5791,
+      "step": 28083
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.827584759870127e-06,
+      "loss": 0.593,
+      "step": 28084
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.826936951970171e-06,
+      "loss": 0.6201,
+      "step": 28085
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.826289165274767e-06,
+      "loss": 0.6064,
+      "step": 28086
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.825641399787214e-06,
+      "loss": 0.6074,
+      "step": 28087
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.824993655510795e-06,
+      "loss": 0.6004,
+      "step": 28088
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.824345932448815e-06,
+      "loss": 0.6116,
+      "step": 28089
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.823698230604549e-06,
+      "loss": 0.593,
+      "step": 28090
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.823050549981297e-06,
+      "loss": 0.5457,
+      "step": 28091
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.822402890582346e-06,
+      "loss": 0.553,
+      "step": 28092
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.821755252410992e-06,
+      "loss": 0.5707,
+      "step": 28093
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.821107635470524e-06,
+      "loss": 0.6368,
+      "step": 28094
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.820460039764233e-06,
+      "loss": 0.6418,
+      "step": 28095
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.819812465295402e-06,
+      "loss": 0.6365,
+      "step": 28096
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.819164912067335e-06,
+      "loss": 0.566,
+      "step": 28097
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.818517380083311e-06,
+      "loss": 0.6232,
+      "step": 28098
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.817869869346631e-06,
+      "loss": 0.5902,
+      "step": 28099
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.817222379860572e-06,
+      "loss": 0.5909,
+      "step": 28100
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.816574911628435e-06,
+      "loss": 0.648,
+      "step": 28101
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.815927464653502e-06,
+      "loss": 0.6483,
+      "step": 28102
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8152800389390705e-06,
+      "loss": 0.5795,
+      "step": 28103
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.814632634488429e-06,
+      "loss": 0.615,
+      "step": 28104
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.813985251304862e-06,
+      "loss": 0.6032,
+      "step": 28105
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.813337889391663e-06,
+      "loss": 0.5964,
+      "step": 28106
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8126905487521156e-06,
+      "loss": 0.5615,
+      "step": 28107
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.812043229389519e-06,
+      "loss": 0.5693,
+      "step": 28108
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.811395931307153e-06,
+      "loss": 0.6055,
+      "step": 28109
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.810748654508318e-06,
+      "loss": 0.6213,
+      "step": 28110
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8101013989962885e-06,
+      "loss": 0.5646,
+      "step": 28111
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.809454164774366e-06,
+      "loss": 0.558,
+      "step": 28112
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.808806951845829e-06,
+      "loss": 0.5647,
+      "step": 28113
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.808159760213975e-06,
+      "loss": 0.6499,
+      "step": 28114
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.807512589882085e-06,
+      "loss": 0.5566,
+      "step": 28115
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.80686544085346e-06,
+      "loss": 0.5671,
+      "step": 28116
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.806218313131371e-06,
+      "loss": 0.6518,
+      "step": 28117
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.805571206719121e-06,
+      "loss": 0.5998,
+      "step": 28118
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.804924121619986e-06,
+      "loss": 0.5498,
+      "step": 28119
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.804277057837264e-06,
+      "loss": 0.5492,
+      "step": 28120
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.803630015374239e-06,
+      "loss": 0.548,
+      "step": 28121
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8029829942342005e-06,
+      "loss": 0.6486,
+      "step": 28122
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.802335994420429e-06,
+      "loss": 0.5518,
+      "step": 28123
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8016890159362225e-06,
+      "loss": 0.6381,
+      "step": 28124
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.801042058784862e-06,
+      "loss": 0.6147,
+      "step": 28125
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.8003951229696396e-06,
+      "loss": 0.5794,
+      "step": 28126
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.799748208493837e-06,
+      "loss": 0.6502,
+      "step": 28127
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.799101315360741e-06,
+      "loss": 0.5672,
+      "step": 28128
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.798454443573644e-06,
+      "loss": 0.616,
+      "step": 28129
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.797807593135829e-06,
+      "loss": 0.5204,
+      "step": 28130
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.797160764050587e-06,
+      "loss": 0.618,
+      "step": 28131
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.796513956321201e-06,
+      "loss": 0.5765,
+      "step": 28132
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.795867169950959e-06,
+      "loss": 0.6051,
+      "step": 28133
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.795220404943145e-06,
+      "loss": 0.5841,
+      "step": 28134
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.79457366130105e-06,
+      "loss": 0.6234,
+      "step": 28135
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.793926939027954e-06,
+      "loss": 0.5993,
+      "step": 28136
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.793280238127155e-06,
+      "loss": 0.5562,
+      "step": 28137
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.792633558601923e-06,
+      "loss": 0.6041,
+      "step": 28138
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 5.791986900455557e-06,
+      "loss": 0.6146,
+      "step": 28139
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.791340263691333e-06,
+      "loss": 0.6025,
+      "step": 28140
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.790693648312547e-06,
+      "loss": 0.5605,
+      "step": 28141
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.790047054322476e-06,
+      "loss": 0.5732,
+      "step": 28142
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.789400481724412e-06,
+      "loss": 0.6292,
+      "step": 28143
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7887539305216324e-06,
+      "loss": 0.5927,
+      "step": 28144
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.788107400717431e-06,
+      "loss": 0.5305,
+      "step": 28145
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7874608923150896e-06,
+      "loss": 0.5983,
+      "step": 28146
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.78681440531789e-06,
+      "loss": 0.5976,
+      "step": 28147
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.786167939729127e-06,
+      "loss": 0.6692,
+      "step": 28148
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.785521495552071e-06,
+      "loss": 0.5617,
+      "step": 28149
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.78487507279002e-06,
+      "loss": 0.6297,
+      "step": 28150
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.784228671446248e-06,
+      "loss": 0.5583,
+      "step": 28151
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.783582291524048e-06,
+      "loss": 0.647,
+      "step": 28152
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7829359330267035e-06,
+      "loss": 0.6218,
+      "step": 28153
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.782289595957495e-06,
+      "loss": 0.5973,
+      "step": 28154
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.781643280319704e-06,
+      "loss": 0.6385,
+      "step": 28155
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.780996986116623e-06,
+      "loss": 0.5875,
+      "step": 28156
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.780350713351528e-06,
+      "loss": 0.6168,
+      "step": 28157
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.779704462027711e-06,
+      "loss": 0.6513,
+      "step": 28158
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.779058232148451e-06,
+      "loss": 0.6315,
+      "step": 28159
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7784120237170326e-06,
+      "loss": 0.6202,
+      "step": 28160
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.777765836736736e-06,
+      "loss": 0.5937,
+      "step": 28161
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.777119671210851e-06,
+      "loss": 0.6064,
+      "step": 28162
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.776473527142659e-06,
+      "loss": 0.6373,
+      "step": 28163
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.775827404535439e-06,
+      "loss": 0.6221,
+      "step": 28164
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.775181303392475e-06,
+      "loss": 0.6394,
+      "step": 28165
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.774535223717056e-06,
+      "loss": 0.5551,
+      "step": 28166
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.77388916551246e-06,
+      "loss": 0.5814,
+      "step": 28167
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.773243128781969e-06,
+      "loss": 0.5435,
+      "step": 28168
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.772597113528871e-06,
+      "loss": 0.6122,
+      "step": 28169
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.771951119756443e-06,
+      "loss": 0.6532,
+      "step": 28170
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7713051474679716e-06,
+      "loss": 0.5732,
+      "step": 28171
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.770659196666732e-06,
+      "loss": 0.6159,
+      "step": 28172
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.770013267356016e-06,
+      "loss": 0.5913,
+      "step": 28173
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7693673595390976e-06,
+      "loss": 0.5859,
+      "step": 28174
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.768721473219271e-06,
+      "loss": 0.6148,
+      "step": 28175
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.768075608399802e-06,
+      "loss": 0.5507,
+      "step": 28176
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.767429765083983e-06,
+      "loss": 0.5816,
+      "step": 28177
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7667839432750895e-06,
+      "loss": 0.6343,
+      "step": 28178
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.76613814297641e-06,
+      "loss": 0.6471,
+      "step": 28179
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.765492364191223e-06,
+      "loss": 0.674,
+      "step": 28180
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7648466069228095e-06,
+      "loss": 0.5503,
+      "step": 28181
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7642008711744465e-06,
+      "loss": 0.5659,
+      "step": 28182
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7635551569494245e-06,
+      "loss": 0.6181,
+      "step": 28183
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.762909464251019e-06,
+      "loss": 0.6091,
+      "step": 28184
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.762263793082505e-06,
+      "loss": 0.5993,
+      "step": 28185
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.76161814344718e-06,
+      "loss": 0.6144,
+      "step": 28186
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.760972515348307e-06,
+      "loss": 0.5815,
+      "step": 28187
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7603269087891776e-06,
+      "loss": 0.5661,
+      "step": 28188
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.759681323773064e-06,
+      "loss": 0.6538,
+      "step": 28189
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.759035760303255e-06,
+      "loss": 0.5878,
+      "step": 28190
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.75839021838303e-06,
+      "loss": 0.5659,
+      "step": 28191
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7577446980156634e-06,
+      "loss": 0.6074,
+      "step": 28192
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.757099199204437e-06,
+      "loss": 0.6265,
+      "step": 28193
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.756453721952635e-06,
+      "loss": 0.5817,
+      "step": 28194
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.75580826626353e-06,
+      "loss": 0.5597,
+      "step": 28195
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.75516283214041e-06,
+      "loss": 0.5955,
+      "step": 28196
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.754517419586552e-06,
+      "loss": 0.662,
+      "step": 28197
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.753872028605233e-06,
+      "loss": 0.6093,
+      "step": 28198
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.753226659199731e-06,
+      "loss": 0.5882,
+      "step": 28199
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.752581311373331e-06,
+      "loss": 0.5673,
+      "step": 28200
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.75193598512931e-06,
+      "loss": 0.6,
+      "step": 28201
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7512906804709465e-06,
+      "loss": 0.6482,
+      "step": 28202
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.750645397401514e-06,
+      "loss": 0.663,
+      "step": 28203
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.750000135924302e-06,
+      "loss": 0.5689,
+      "step": 28204
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.749354896042584e-06,
+      "loss": 0.5504,
+      "step": 28205
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.748709677759634e-06,
+      "loss": 0.5873,
+      "step": 28206
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.748064481078739e-06,
+      "loss": 0.586,
+      "step": 28207
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7474193060031744e-06,
+      "loss": 0.5959,
+      "step": 28208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.746774152536218e-06,
+      "loss": 0.5904,
+      "step": 28209
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.746129020681143e-06,
+      "loss": 0.6044,
+      "step": 28210
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.745483910441236e-06,
+      "loss": 0.6216,
+      "step": 28211
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.744838821819767e-06,
+      "loss": 0.6256,
+      "step": 28212
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.744193754820028e-06,
+      "loss": 0.6004,
+      "step": 28213
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7435487094452765e-06,
+      "loss": 0.5916,
+      "step": 28214
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.742903685698807e-06,
+      "loss": 0.6231,
+      "step": 28215
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7422586835838854e-06,
+      "loss": 0.6091,
+      "step": 28216
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.741613703103798e-06,
+      "loss": 0.5972,
+      "step": 28217
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.740968744261819e-06,
+      "loss": 0.594,
+      "step": 28218
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.740323807061225e-06,
+      "loss": 0.6402,
+      "step": 28219
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.739678891505289e-06,
+      "loss": 0.567,
+      "step": 28220
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.739033997597296e-06,
+      "loss": 0.6476,
+      "step": 28221
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.73838912534052e-06,
+      "loss": 0.5933,
+      "step": 28222
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.737744274738233e-06,
+      "loss": 0.6446,
+      "step": 28223
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.737099445793719e-06,
+      "loss": 0.6482,
+      "step": 28224
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.736454638510252e-06,
+      "loss": 0.5764,
+      "step": 28225
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.735809852891106e-06,
+      "loss": 0.5658,
+      "step": 28226
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.735165088939556e-06,
+      "loss": 0.5924,
+      "step": 28227
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.734520346658885e-06,
+      "loss": 0.649,
+      "step": 28228
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.733875626052365e-06,
+      "loss": 0.5826,
+      "step": 28229
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.733230927123273e-06,
+      "loss": 0.5916,
+      "step": 28230
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7325862498748795e-06,
+      "loss": 0.5734,
+      "step": 28231
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.73194159431047e-06,
+      "loss": 0.5957,
+      "step": 28232
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.73129696043331e-06,
+      "loss": 0.5894,
+      "step": 28233
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.730652348246686e-06,
+      "loss": 0.5954,
+      "step": 28234
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.730007757753866e-06,
+      "loss": 0.5121,
+      "step": 28235
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.729363188958128e-06,
+      "loss": 0.5689,
+      "step": 28236
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.728718641862743e-06,
+      "loss": 0.607,
+      "step": 28237
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.728074116470992e-06,
+      "loss": 0.6446,
+      "step": 28238
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.727429612786148e-06,
+      "loss": 0.6547,
+      "step": 28239
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7267851308114855e-06,
+      "loss": 0.5832,
+      "step": 28240
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7261406705502754e-06,
+      "loss": 0.6253,
+      "step": 28241
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.725496232005801e-06,
+      "loss": 0.6675,
+      "step": 28242
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7248518151813305e-06,
+      "loss": 0.5617,
+      "step": 28243
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7242074200801365e-06,
+      "loss": 0.6327,
+      "step": 28244
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.723563046705501e-06,
+      "loss": 0.6025,
+      "step": 28245
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7229186950606954e-06,
+      "loss": 0.6139,
+      "step": 28246
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.72227436514899e-06,
+      "loss": 0.5759,
+      "step": 28247
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7216300569736595e-06,
+      "loss": 0.6138,
+      "step": 28248
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.720985770537981e-06,
+      "loss": 0.5759,
+      "step": 28249
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7203415058452264e-06,
+      "loss": 0.5476,
+      "step": 28250
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.719697262898676e-06,
+      "loss": 0.636,
+      "step": 28251
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.71905304170159e-06,
+      "loss": 0.5456,
+      "step": 28252
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.718408842257252e-06,
+      "loss": 0.5876,
+      "step": 28253
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.71776466456893e-06,
+      "loss": 0.6191,
+      "step": 28254
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.717120508639903e-06,
+      "loss": 0.6176,
+      "step": 28255
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.716476374473441e-06,
+      "loss": 0.5434,
+      "step": 28256
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.715832262072818e-06,
+      "loss": 0.656,
+      "step": 28257
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7151881714413e-06,
+      "loss": 0.5991,
+      "step": 28258
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.714544102582171e-06,
+      "loss": 0.5621,
+      "step": 28259
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.713900055498699e-06,
+      "loss": 0.607,
+      "step": 28260
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.713256030194152e-06,
+      "loss": 0.6036,
+      "step": 28261
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7126120266718085e-06,
+      "loss": 0.6165,
+      "step": 28262
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.711968044934942e-06,
+      "loss": 0.5607,
+      "step": 28263
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.71132408498682e-06,
+      "loss": 0.5803,
+      "step": 28264
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7106801468307115e-06,
+      "loss": 0.5813,
+      "step": 28265
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7100362304699e-06,
+      "loss": 0.5839,
+      "step": 28266
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.709392335907649e-06,
+      "loss": 0.6142,
+      "step": 28267
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7087484631472315e-06,
+      "loss": 0.6038,
+      "step": 28268
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.708104612191916e-06,
+      "loss": 0.616,
+      "step": 28269
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.707460783044982e-06,
+      "loss": 0.6321,
+      "step": 28270
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.706816975709695e-06,
+      "loss": 0.5624,
+      "step": 28271
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7061731901893305e-06,
+      "loss": 0.6161,
+      "step": 28272
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.705529426487157e-06,
+      "loss": 0.6609,
+      "step": 28273
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.704885684606447e-06,
+      "loss": 0.5934,
+      "step": 28274
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.704241964550467e-06,
+      "loss": 0.6375,
+      "step": 28275
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.703598266322494e-06,
+      "loss": 0.5622,
+      "step": 28276
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.7029545899257975e-06,
+      "loss": 0.5814,
+      "step": 28277
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.702310935363648e-06,
+      "loss": 0.6353,
+      "step": 28278
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.701667302639311e-06,
+      "loss": 0.5727,
+      "step": 28279
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.701023691756065e-06,
+      "loss": 0.597,
+      "step": 28280
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.700380102717176e-06,
+      "loss": 0.6183,
+      "step": 28281
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.699736535525912e-06,
+      "loss": 0.6389,
+      "step": 28282
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.699092990185549e-06,
+      "loss": 0.6112,
+      "step": 28283
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.698449466699354e-06,
+      "loss": 0.674,
+      "step": 28284
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.697805965070597e-06,
+      "loss": 0.6143,
+      "step": 28285
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 5.697162485302545e-06,
+      "loss": 0.5937,
+      "step": 28286
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.696519027398475e-06,
+      "loss": 0.6041,
+      "step": 28287
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.695875591361646e-06,
+      "loss": 0.6153,
+      "step": 28288
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.695232177195343e-06,
+      "loss": 0.6234,
+      "step": 28289
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.694588784902818e-06,
+      "loss": 0.5913,
+      "step": 28290
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.693945414487352e-06,
+      "loss": 0.6243,
+      "step": 28291
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.693302065952207e-06,
+      "loss": 0.5661,
+      "step": 28292
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.692658739300659e-06,
+      "loss": 0.617,
+      "step": 28293
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.692015434535973e-06,
+      "loss": 0.5488,
+      "step": 28294
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.691372151661418e-06,
+      "loss": 0.6529,
+      "step": 28295
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.690728890680261e-06,
+      "loss": 0.5729,
+      "step": 28296
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.690085651595774e-06,
+      "loss": 0.5652,
+      "step": 28297
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.689442434411226e-06,
+      "loss": 0.6585,
+      "step": 28298
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.688799239129879e-06,
+      "loss": 0.6318,
+      "step": 28299
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.688156065755008e-06,
+      "loss": 0.6031,
+      "step": 28300
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.68751291428988e-06,
+      "loss": 0.6444,
+      "step": 28301
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.686869784737762e-06,
+      "loss": 0.6016,
+      "step": 28302
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.686226677101918e-06,
+      "loss": 0.6228,
+      "step": 28303
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6855835913856225e-06,
+      "loss": 0.6289,
+      "step": 28304
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.684940527592142e-06,
+      "loss": 0.6381,
+      "step": 28305
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.684297485724741e-06,
+      "loss": 0.6049,
+      "step": 28306
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.683654465786686e-06,
+      "loss": 0.6378,
+      "step": 28307
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.683011467781248e-06,
+      "loss": 0.5742,
+      "step": 28308
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.682368491711692e-06,
+      "loss": 0.5849,
+      "step": 28309
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6817255375812916e-06,
+      "loss": 0.6063,
+      "step": 28310
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.681082605393302e-06,
+      "loss": 0.6,
+      "step": 28311
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.680439695151e-06,
+      "loss": 0.6124,
+      "step": 28312
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6797968068576444e-06,
+      "loss": 0.6296,
+      "step": 28313
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6791539405165105e-06,
+      "loss": 0.6122,
+      "step": 28314
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6785110961308585e-06,
+      "loss": 0.5958,
+      "step": 28315
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.677868273703964e-06,
+      "loss": 0.6053,
+      "step": 28316
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6772254732390784e-06,
+      "loss": 0.6224,
+      "step": 28317
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.676582694739481e-06,
+      "loss": 0.629,
+      "step": 28318
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.675939938208433e-06,
+      "loss": 0.5402,
+      "step": 28319
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.675297203649196e-06,
+      "loss": 0.657,
+      "step": 28320
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.674654491065045e-06,
+      "loss": 0.6024,
+      "step": 28321
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.674011800459241e-06,
+      "loss": 0.5874,
+      "step": 28322
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.673369131835049e-06,
+      "loss": 0.6427,
+      "step": 28323
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.672726485195732e-06,
+      "loss": 0.5812,
+      "step": 28324
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.672083860544564e-06,
+      "loss": 0.6544,
+      "step": 28325
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.671441257884801e-06,
+      "loss": 0.5981,
+      "step": 28326
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.67079867721972e-06,
+      "loss": 0.6186,
+      "step": 28327
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.670156118552572e-06,
+      "loss": 0.6248,
+      "step": 28328
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.669513581886632e-06,
+      "loss": 0.5588,
+      "step": 28329
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.668871067225158e-06,
+      "loss": 0.6658,
+      "step": 28330
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.668228574571422e-06,
+      "loss": 0.5935,
+      "step": 28331
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.667586103928686e-06,
+      "loss": 0.6024,
+      "step": 28332
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6669436553002134e-06,
+      "loss": 0.5738,
+      "step": 28333
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.666301228689265e-06,
+      "loss": 0.5554,
+      "step": 28334
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.665658824099114e-06,
+      "loss": 0.5762,
+      "step": 28335
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.665016441533015e-06,
+      "loss": 0.5688,
+      "step": 28336
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.664374080994245e-06,
+      "loss": 0.607,
+      "step": 28337
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.663731742486052e-06,
+      "loss": 0.6153,
+      "step": 28338
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.663089426011712e-06,
+      "loss": 0.6145,
+      "step": 28339
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.662447131574484e-06,
+      "loss": 0.5666,
+      "step": 28340
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.661804859177631e-06,
+      "loss": 0.6427,
+      "step": 28341
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.66116260882442e-06,
+      "loss": 0.5464,
+      "step": 28342
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.660520380518113e-06,
+      "loss": 0.5742,
+      "step": 28343
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6598781742619725e-06,
+      "loss": 0.5898,
+      "step": 28344
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.659235990059259e-06,
+      "loss": 0.6301,
+      "step": 28345
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.658593827913241e-06,
+      "loss": 0.6447,
+      "step": 28346
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6579516878271764e-06,
+      "loss": 0.595,
+      "step": 28347
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.657309569804339e-06,
+      "loss": 0.6289,
+      "step": 28348
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6566674738479745e-06,
+      "loss": 0.6517,
+      "step": 28349
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6560253999613604e-06,
+      "loss": 0.6301,
+      "step": 28350
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.655383348147747e-06,
+      "loss": 0.5769,
+      "step": 28351
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.654741318410408e-06,
+      "loss": 0.6384,
+      "step": 28352
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.654099310752599e-06,
+      "loss": 0.546,
+      "step": 28353
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.653457325177589e-06,
+      "loss": 0.6289,
+      "step": 28354
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.65281536168863e-06,
+      "loss": 0.6212,
+      "step": 28355
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.652173420288991e-06,
+      "loss": 0.6602,
+      "step": 28356
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.65153150098193e-06,
+      "loss": 0.5768,
+      "step": 28357
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.650889603770715e-06,
+      "loss": 0.608,
+      "step": 28358
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.650247728658602e-06,
+      "loss": 0.5789,
+      "step": 28359
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6496058756488555e-06,
+      "loss": 0.5867,
+      "step": 28360
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.648964044744734e-06,
+      "loss": 0.6567,
+      "step": 28361
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.648322235949499e-06,
+      "loss": 0.67,
+      "step": 28362
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6476804492664154e-06,
+      "loss": 0.6095,
+      "step": 28363
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.647038684698738e-06,
+      "loss": 0.6047,
+      "step": 28364
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.646396942249741e-06,
+      "loss": 0.6334,
+      "step": 28365
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6457552219226665e-06,
+      "loss": 0.6257,
+      "step": 28366
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.64511352372079e-06,
+      "loss": 0.6211,
+      "step": 28367
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6444718476473635e-06,
+      "loss": 0.5731,
+      "step": 28368
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.643830193705654e-06,
+      "loss": 0.6357,
+      "step": 28369
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.64318856189892e-06,
+      "loss": 0.5679,
+      "step": 28370
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.64254695223042e-06,
+      "loss": 0.6448,
+      "step": 28371
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6419053647034114e-06,
+      "loss": 0.6064,
+      "step": 28372
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.641263799321162e-06,
+      "loss": 0.6052,
+      "step": 28373
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.640622256086924e-06,
+      "loss": 0.5976,
+      "step": 28374
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6399807350039695e-06,
+      "loss": 0.6043,
+      "step": 28375
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.639339236075541e-06,
+      "loss": 0.6326,
+      "step": 28376
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.638697759304911e-06,
+      "loss": 0.6066,
+      "step": 28377
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.638056304695334e-06,
+      "loss": 0.6011,
+      "step": 28378
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.637414872250069e-06,
+      "loss": 0.6259,
+      "step": 28379
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.636773461972378e-06,
+      "loss": 0.5728,
+      "step": 28380
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.63613207386552e-06,
+      "loss": 0.5236,
+      "step": 28381
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6354907079327535e-06,
+      "loss": 0.5995,
+      "step": 28382
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.634849364177332e-06,
+      "loss": 0.6547,
+      "step": 28383
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.634208042602524e-06,
+      "loss": 0.5966,
+      "step": 28384
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6335667432115786e-06,
+      "loss": 0.6143,
+      "step": 28385
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.632925466007768e-06,
+      "loss": 0.5884,
+      "step": 28386
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.632284210994334e-06,
+      "loss": 0.6009,
+      "step": 28387
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6316429781745475e-06,
+      "loss": 0.6186,
+      "step": 28388
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6310017675516574e-06,
+      "loss": 0.6688,
+      "step": 28389
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.630360579128931e-06,
+      "loss": 0.5913,
+      "step": 28390
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.629719412909618e-06,
+      "loss": 0.6222,
+      "step": 28391
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.629078268896989e-06,
+      "loss": 0.6203,
+      "step": 28392
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.628437147094285e-06,
+      "loss": 0.6535,
+      "step": 28393
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.627796047504778e-06,
+      "loss": 0.5459,
+      "step": 28394
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6271549701317145e-06,
+      "loss": 0.6482,
+      "step": 28395
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.62651391497836e-06,
+      "loss": 0.5595,
+      "step": 28396
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.62587288204797e-06,
+      "loss": 0.5991,
+      "step": 28397
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.625231871343801e-06,
+      "loss": 0.534,
+      "step": 28398
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.624590882869111e-06,
+      "loss": 0.5922,
+      "step": 28399
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6239499166271515e-06,
+      "loss": 0.6503,
+      "step": 28400
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.623308972621188e-06,
+      "loss": 0.5371,
+      "step": 28401
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.622668050854473e-06,
+      "loss": 0.5787,
+      "step": 28402
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.622027151330263e-06,
+      "loss": 0.5892,
+      "step": 28403
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.621386274051812e-06,
+      "loss": 0.6162,
+      "step": 28404
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6207454190223845e-06,
+      "loss": 0.6462,
+      "step": 28405
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.620104586245227e-06,
+      "loss": 0.5802,
+      "step": 28406
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.619463775723605e-06,
+      "loss": 0.5963,
+      "step": 28407
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.618822987460771e-06,
+      "loss": 0.5508,
+      "step": 28408
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.61818222145998e-06,
+      "loss": 0.6046,
+      "step": 28409
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.617541477724484e-06,
+      "loss": 0.5904,
+      "step": 28410
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.616900756257547e-06,
+      "loss": 0.6154,
+      "step": 28411
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.616260057062418e-06,
+      "loss": 0.5673,
+      "step": 28412
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.615619380142363e-06,
+      "loss": 0.5395,
+      "step": 28413
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.614978725500621e-06,
+      "loss": 0.6146,
+      "step": 28414
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.614338093140462e-06,
+      "loss": 0.5809,
+      "step": 28415
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.613697483065131e-06,
+      "loss": 0.6352,
+      "step": 28416
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.613056895277892e-06,
+      "loss": 0.5931,
+      "step": 28417
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.612416329781995e-06,
+      "loss": 0.6118,
+      "step": 28418
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6117757865806955e-06,
+      "loss": 0.6236,
+      "step": 28419
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.611135265677249e-06,
+      "loss": 0.6184,
+      "step": 28420
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.610494767074906e-06,
+      "loss": 0.5956,
+      "step": 28421
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.609854290776927e-06,
+      "loss": 0.5675,
+      "step": 28422
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.60921383678656e-06,
+      "loss": 0.6251,
+      "step": 28423
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6085734051070725e-06,
+      "loss": 0.6027,
+      "step": 28424
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.607932995741703e-06,
+      "loss": 0.584,
+      "step": 28425
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.607292608693713e-06,
+      "loss": 0.628,
+      "step": 28426
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.606652243966354e-06,
+      "loss": 0.594,
+      "step": 28427
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.606011901562885e-06,
+      "loss": 0.5907,
+      "step": 28428
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.605371581486556e-06,
+      "loss": 0.6103,
+      "step": 28429
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.604731283740621e-06,
+      "loss": 0.6165,
+      "step": 28430
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.604091008328331e-06,
+      "loss": 0.6044,
+      "step": 28431
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.6034507552529435e-06,
+      "loss": 0.6298,
+      "step": 28432
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 5.602810524517709e-06,
+      "loss": 0.5689,
+      "step": 28433
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.602170316125884e-06,
+      "loss": 0.6134,
+      "step": 28434
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.601530130080721e-06,
+      "loss": 0.656,
+      "step": 28435
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.60088996638547e-06,
+      "loss": 0.5781,
+      "step": 28436
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.6002498250433825e-06,
+      "loss": 0.5595,
+      "step": 28437
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5996097060577185e-06,
+      "loss": 0.5939,
+      "step": 28438
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.598969609431726e-06,
+      "loss": 0.6352,
+      "step": 28439
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5983295351686565e-06,
+      "loss": 0.5434,
+      "step": 28440
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.597689483271765e-06,
+      "loss": 0.583,
+      "step": 28441
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5970494537442985e-06,
+      "loss": 0.6065,
+      "step": 28442
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.596409446589517e-06,
+      "loss": 0.5879,
+      "step": 28443
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.595769461810665e-06,
+      "loss": 0.5299,
+      "step": 28444
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.595129499411002e-06,
+      "loss": 0.6287,
+      "step": 28445
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.594489559393776e-06,
+      "loss": 0.607,
+      "step": 28446
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.593849641762239e-06,
+      "loss": 0.6104,
+      "step": 28447
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5932097465196366e-06,
+      "loss": 0.5988,
+      "step": 28448
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.592569873669232e-06,
+      "loss": 0.6077,
+      "step": 28449
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.591930023214267e-06,
+      "loss": 0.5476,
+      "step": 28450
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.591290195158003e-06,
+      "loss": 0.5899,
+      "step": 28451
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.590650389503678e-06,
+      "loss": 0.6174,
+      "step": 28452
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5900106062545524e-06,
+      "loss": 0.5448,
+      "step": 28453
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5893708454138715e-06,
+      "loss": 0.6058,
+      "step": 28454
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.588731106984893e-06,
+      "loss": 0.6148,
+      "step": 28455
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5880913909708635e-06,
+      "loss": 0.6413,
+      "step": 28456
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.587451697375032e-06,
+      "loss": 0.5893,
+      "step": 28457
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.586812026200654e-06,
+      "loss": 0.6109,
+      "step": 28458
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5861723774509706e-06,
+      "loss": 0.6182,
+      "step": 28459
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.585532751129241e-06,
+      "loss": 0.6383,
+      "step": 28460
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.58489314723871e-06,
+      "loss": 0.6007,
+      "step": 28461
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.584253565782638e-06,
+      "loss": 0.6285,
+      "step": 28462
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.583614006764259e-06,
+      "loss": 0.6057,
+      "step": 28463
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.582974470186833e-06,
+      "loss": 0.5944,
+      "step": 28464
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5823349560536046e-06,
+      "loss": 0.5982,
+      "step": 28465
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.581695464367829e-06,
+      "loss": 0.6372,
+      "step": 28466
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.581055995132753e-06,
+      "loss": 0.6234,
+      "step": 28467
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.580416548351627e-06,
+      "loss": 0.611,
+      "step": 28468
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5797771240276945e-06,
+      "loss": 0.5757,
+      "step": 28469
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.579137722164212e-06,
+      "loss": 0.5751,
+      "step": 28470
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.578498342764422e-06,
+      "loss": 0.6131,
+      "step": 28471
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5778589858315816e-06,
+      "loss": 0.5886,
+      "step": 28472
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.577219651368933e-06,
+      "loss": 0.5914,
+      "step": 28473
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.576580339379727e-06,
+      "loss": 0.5534,
+      "step": 28474
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.575941049867208e-06,
+      "loss": 0.6633,
+      "step": 28475
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.575301782834632e-06,
+      "loss": 0.584,
+      "step": 28476
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.574662538285245e-06,
+      "loss": 0.6242,
+      "step": 28477
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5740233162222926e-06,
+      "loss": 0.6303,
+      "step": 28478
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.573384116649023e-06,
+      "loss": 0.622,
+      "step": 28479
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5727449395686815e-06,
+      "loss": 0.5762,
+      "step": 28480
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.572105784984525e-06,
+      "loss": 0.6429,
+      "step": 28481
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.571466652899791e-06,
+      "loss": 0.6064,
+      "step": 28482
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.570827543317735e-06,
+      "loss": 0.6407,
+      "step": 28483
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.570188456241602e-06,
+      "loss": 0.5638,
+      "step": 28484
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.569549391674639e-06,
+      "loss": 0.6433,
+      "step": 28485
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5689103496200894e-06,
+      "loss": 0.5728,
+      "step": 28486
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.568271330081207e-06,
+      "loss": 0.5866,
+      "step": 28487
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.567632333061232e-06,
+      "loss": 0.5549,
+      "step": 28488
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5669933585634235e-06,
+      "loss": 0.6538,
+      "step": 28489
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.566354406591013e-06,
+      "loss": 0.5845,
+      "step": 28490
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.565715477147258e-06,
+      "loss": 0.5463,
+      "step": 28491
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5650765702353945e-06,
+      "loss": 0.5642,
+      "step": 28492
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.564437685858684e-06,
+      "loss": 0.6665,
+      "step": 28493
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.563798824020361e-06,
+      "loss": 0.6256,
+      "step": 28494
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.563159984723677e-06,
+      "loss": 0.5441,
+      "step": 28495
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.562521167971873e-06,
+      "loss": 0.6249,
+      "step": 28496
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.561882373768202e-06,
+      "loss": 0.5806,
+      "step": 28497
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.561243602115907e-06,
+      "loss": 0.6847,
+      "step": 28498
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.560604853018228e-06,
+      "loss": 0.6213,
+      "step": 28499
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.559966126478424e-06,
+      "loss": 0.6181,
+      "step": 28500
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.559327422499725e-06,
+      "loss": 0.5803,
+      "step": 28501
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.558688741085387e-06,
+      "loss": 0.6304,
+      "step": 28502
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.558050082238648e-06,
+      "loss": 0.5805,
+      "step": 28503
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.557411445962763e-06,
+      "loss": 0.5477,
+      "step": 28504
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.556772832260971e-06,
+      "loss": 0.5543,
+      "step": 28505
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.556134241136517e-06,
+      "loss": 0.6078,
+      "step": 28506
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5554956725926435e-06,
+      "loss": 0.5997,
+      "step": 28507
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.554857126632601e-06,
+      "loss": 0.6244,
+      "step": 28508
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.554218603259628e-06,
+      "loss": 0.609,
+      "step": 28509
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.553580102476977e-06,
+      "loss": 0.6211,
+      "step": 28510
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.552941624287886e-06,
+      "loss": 0.5894,
+      "step": 28511
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.552303168695601e-06,
+      "loss": 0.5897,
+      "step": 28512
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.551664735703364e-06,
+      "loss": 0.6284,
+      "step": 28513
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.551026325314423e-06,
+      "loss": 0.5919,
+      "step": 28514
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.550387937532021e-06,
+      "loss": 0.5931,
+      "step": 28515
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.549749572359402e-06,
+      "loss": 0.7134,
+      "step": 28516
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.549111229799804e-06,
+      "loss": 0.6599,
+      "step": 28517
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.548472909856478e-06,
+      "loss": 0.6133,
+      "step": 28518
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.547834612532666e-06,
+      "loss": 0.5842,
+      "step": 28519
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5471963378316065e-06,
+      "loss": 0.6806,
+      "step": 28520
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5465580857565544e-06,
+      "loss": 0.5611,
+      "step": 28521
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.545919856310737e-06,
+      "loss": 0.557,
+      "step": 28522
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.545281649497409e-06,
+      "loss": 0.588,
+      "step": 28523
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.544643465319806e-06,
+      "loss": 0.5864,
+      "step": 28524
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.544005303781177e-06,
+      "loss": 0.6315,
+      "step": 28525
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.54336716488476e-06,
+      "loss": 0.5604,
+      "step": 28526
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.542729048633806e-06,
+      "loss": 0.5727,
+      "step": 28527
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5420909550315445e-06,
+      "loss": 0.5724,
+      "step": 28528
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.541452884081226e-06,
+      "loss": 0.6181,
+      "step": 28529
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.54081483578609e-06,
+      "loss": 0.6376,
+      "step": 28530
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.540176810149382e-06,
+      "loss": 0.6191,
+      "step": 28531
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.539538807174343e-06,
+      "loss": 0.6019,
+      "step": 28532
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.538900826864213e-06,
+      "loss": 0.5687,
+      "step": 28533
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.538262869222231e-06,
+      "loss": 0.6651,
+      "step": 28534
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.537624934251644e-06,
+      "loss": 0.5951,
+      "step": 28535
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.536987021955693e-06,
+      "loss": 0.579,
+      "step": 28536
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.536349132337615e-06,
+      "loss": 0.5346,
+      "step": 28537
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.535711265400656e-06,
+      "loss": 0.5879,
+      "step": 28538
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.535073421148057e-06,
+      "loss": 0.5887,
+      "step": 28539
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.534435599583057e-06,
+      "loss": 0.5825,
+      "step": 28540
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5337978007088945e-06,
+      "loss": 0.6186,
+      "step": 28541
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.533160024528817e-06,
+      "loss": 0.5971,
+      "step": 28542
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.532522271046062e-06,
+      "loss": 0.6315,
+      "step": 28543
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5318845402638675e-06,
+      "loss": 0.6066,
+      "step": 28544
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.531246832185474e-06,
+      "loss": 0.6143,
+      "step": 28545
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.530609146814127e-06,
+      "loss": 0.5966,
+      "step": 28546
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5299714841530604e-06,
+      "loss": 0.6658,
+      "step": 28547
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5293338442055225e-06,
+      "loss": 0.5999,
+      "step": 28548
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.528696226974748e-06,
+      "loss": 0.6128,
+      "step": 28549
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.528058632463977e-06,
+      "loss": 0.6149,
+      "step": 28550
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.527421060676448e-06,
+      "loss": 0.5886,
+      "step": 28551
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5267835116154035e-06,
+      "loss": 0.5918,
+      "step": 28552
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.526145985284084e-06,
+      "loss": 0.6099,
+      "step": 28553
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.525508481685726e-06,
+      "loss": 0.5835,
+      "step": 28554
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.524871000823566e-06,
+      "loss": 0.5332,
+      "step": 28555
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.524233542700851e-06,
+      "loss": 0.5586,
+      "step": 28556
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.523596107320817e-06,
+      "loss": 0.6091,
+      "step": 28557
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.522958694686698e-06,
+      "loss": 0.6008,
+      "step": 28558
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.522321304801744e-06,
+      "loss": 0.5322,
+      "step": 28559
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.521683937669178e-06,
+      "loss": 0.6283,
+      "step": 28560
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.521046593292253e-06,
+      "loss": 0.5935,
+      "step": 28561
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.520409271674198e-06,
+      "loss": 0.6255,
+      "step": 28562
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.51977197281826e-06,
+      "loss": 0.5312,
+      "step": 28563
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.519134696727666e-06,
+      "loss": 0.5681,
+      "step": 28564
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.518497443405671e-06,
+      "loss": 0.5953,
+      "step": 28565
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.517860212855495e-06,
+      "loss": 0.5716,
+      "step": 28566
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5172230050803875e-06,
+      "loss": 0.557,
+      "step": 28567
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.51658582008358e-06,
+      "loss": 0.5586,
+      "step": 28568
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.515948657868315e-06,
+      "loss": 0.6164,
+      "step": 28569
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.515311518437829e-06,
+      "loss": 0.5529,
+      "step": 28570
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.514674401795358e-06,
+      "loss": 0.5914,
+      "step": 28571
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.514037307944137e-06,
+      "loss": 0.6356,
+      "step": 28572
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.513400236887409e-06,
+      "loss": 0.6129,
+      "step": 28573
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.512763188628409e-06,
+      "loss": 0.6606,
+      "step": 28574
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.512126163170369e-06,
+      "loss": 0.6178,
+      "step": 28575
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.511489160516534e-06,
+      "loss": 0.6326,
+      "step": 28576
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.510852180670138e-06,
+      "loss": 0.5923,
+      "step": 28577
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5102152236344176e-06,
+      "loss": 0.6102,
+      "step": 28578
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.5095782894126025e-06,
+      "loss": 0.5478,
+      "step": 28579
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 5.508941378007942e-06,
+      "loss": 0.5827,
+      "step": 28580
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.508304489423663e-06,
+      "loss": 0.6121,
+      "step": 28581
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.507667623663005e-06,
+      "loss": 0.5771,
+      "step": 28582
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.507030780729199e-06,
+      "loss": 0.5588,
+      "step": 28583
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.506393960625491e-06,
+      "loss": 0.5984,
+      "step": 28584
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.5057571633551054e-06,
+      "loss": 0.5813,
+      "step": 28585
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.505120388921293e-06,
+      "loss": 0.6247,
+      "step": 28586
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.504483637327273e-06,
+      "loss": 0.6089,
+      "step": 28587
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.503846908576292e-06,
+      "loss": 0.5557,
+      "step": 28588
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.5032102026715765e-06,
+      "loss": 0.6626,
+      "step": 28589
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.50257351961637e-06,
+      "loss": 0.6204,
+      "step": 28590
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.501936859413906e-06,
+      "loss": 0.5894,
+      "step": 28591
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.501300222067418e-06,
+      "loss": 0.6185,
+      "step": 28592
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.500663607580137e-06,
+      "loss": 0.5893,
+      "step": 28593
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.500027015955306e-06,
+      "loss": 0.6502,
+      "step": 28594
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.499390447196156e-06,
+      "loss": 0.5932,
+      "step": 28595
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.498753901305916e-06,
+      "loss": 0.6495,
+      "step": 28596
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.49811737828783e-06,
+      "loss": 0.6518,
+      "step": 28597
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4974808781451295e-06,
+      "loss": 0.6348,
+      "step": 28598
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.496844400881046e-06,
+      "loss": 0.6115,
+      "step": 28599
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.496207946498812e-06,
+      "loss": 0.6098,
+      "step": 28600
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4955715150016674e-06,
+      "loss": 0.5584,
+      "step": 28601
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.494935106392839e-06,
+      "loss": 0.6247,
+      "step": 28602
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.494298720675573e-06,
+      "loss": 0.6093,
+      "step": 28603
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.493662357853087e-06,
+      "loss": 0.6217,
+      "step": 28604
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.493026017928626e-06,
+      "loss": 0.6426,
+      "step": 28605
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.492389700905416e-06,
+      "loss": 0.6921,
+      "step": 28606
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.491753406786697e-06,
+      "loss": 0.5663,
+      "step": 28607
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.491117135575701e-06,
+      "loss": 0.635,
+      "step": 28608
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4904808872756576e-06,
+      "loss": 0.6033,
+      "step": 28609
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.489844661889799e-06,
+      "loss": 0.6126,
+      "step": 28610
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.489208459421364e-06,
+      "loss": 0.5607,
+      "step": 28611
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.488572279873582e-06,
+      "loss": 0.6079,
+      "step": 28612
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.487936123249684e-06,
+      "loss": 0.6575,
+      "step": 28613
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.487299989552902e-06,
+      "loss": 0.5956,
+      "step": 28614
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4866638787864734e-06,
+      "loss": 0.6288,
+      "step": 28615
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.486027790953628e-06,
+      "loss": 0.5944,
+      "step": 28616
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.485391726057593e-06,
+      "loss": 0.6244,
+      "step": 28617
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.484755684101608e-06,
+      "loss": 0.5802,
+      "step": 28618
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.484119665088903e-06,
+      "loss": 0.6187,
+      "step": 28619
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.483483669022708e-06,
+      "loss": 0.6082,
+      "step": 28620
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.48284769590625e-06,
+      "loss": 0.5596,
+      "step": 28621
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.482211745742771e-06,
+      "loss": 0.5669,
+      "step": 28622
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.481575818535494e-06,
+      "loss": 0.5725,
+      "step": 28623
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.480939914287661e-06,
+      "loss": 0.5925,
+      "step": 28624
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.480304033002488e-06,
+      "loss": 0.5714,
+      "step": 28625
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.479668174683217e-06,
+      "loss": 0.5597,
+      "step": 28626
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.479032339333072e-06,
+      "loss": 0.6351,
+      "step": 28627
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.478396526955292e-06,
+      "loss": 0.638,
+      "step": 28628
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4777607375530995e-06,
+      "loss": 0.627,
+      "step": 28629
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.477124971129737e-06,
+      "loss": 0.6149,
+      "step": 28630
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.476489227688418e-06,
+      "loss": 0.5945,
+      "step": 28631
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.475853507232387e-06,
+      "loss": 0.5669,
+      "step": 28632
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.47521780976487e-06,
+      "loss": 0.6058,
+      "step": 28633
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.474582135289092e-06,
+      "loss": 0.6763,
+      "step": 28634
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.473946483808291e-06,
+      "loss": 0.6446,
+      "step": 28635
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.473310855325693e-06,
+      "loss": 0.5842,
+      "step": 28636
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.47267524984453e-06,
+      "loss": 0.5913,
+      "step": 28637
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.472039667368025e-06,
+      "loss": 0.6088,
+      "step": 28638
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4714041078994164e-06,
+      "loss": 0.5479,
+      "step": 28639
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.470768571441926e-06,
+      "loss": 0.6211,
+      "step": 28640
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4701330579987945e-06,
+      "loss": 0.6456,
+      "step": 28641
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4694975675732355e-06,
+      "loss": 0.6571,
+      "step": 28642
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4688621001684905e-06,
+      "loss": 0.607,
+      "step": 28643
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.468226655787781e-06,
+      "loss": 0.5978,
+      "step": 28644
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.467591234434342e-06,
+      "loss": 0.6232,
+      "step": 28645
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.466955836111398e-06,
+      "loss": 0.6694,
+      "step": 28646
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.466320460822181e-06,
+      "loss": 0.6108,
+      "step": 28647
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.465685108569913e-06,
+      "loss": 0.6333,
+      "step": 28648
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.465049779357832e-06,
+      "loss": 0.5691,
+      "step": 28649
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4644144731891594e-06,
+      "loss": 0.5672,
+      "step": 28650
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.463779190067127e-06,
+      "loss": 0.5677,
+      "step": 28651
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.463143929994955e-06,
+      "loss": 0.6672,
+      "step": 28652
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.462508692975882e-06,
+      "loss": 0.5303,
+      "step": 28653
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.461873479013131e-06,
+      "loss": 0.6574,
+      "step": 28654
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4612382881099265e-06,
+      "loss": 0.5973,
+      "step": 28655
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.460603120269503e-06,
+      "loss": 0.5892,
+      "step": 28656
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.459967975495086e-06,
+      "loss": 0.564,
+      "step": 28657
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.459332853789901e-06,
+      "loss": 0.5641,
+      "step": 28658
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.45869775515717e-06,
+      "loss": 0.5741,
+      "step": 28659
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.45806267960013e-06,
+      "loss": 0.6158,
+      "step": 28660
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.457427627122001e-06,
+      "loss": 0.6,
+      "step": 28661
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.456792597726019e-06,
+      "loss": 0.6523,
+      "step": 28662
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.456157591415396e-06,
+      "loss": 0.6074,
+      "step": 28663
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.455522608193373e-06,
+      "loss": 0.5616,
+      "step": 28664
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.454887648063165e-06,
+      "loss": 0.6109,
+      "step": 28665
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.454252711028009e-06,
+      "loss": 0.6086,
+      "step": 28666
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.453617797091122e-06,
+      "loss": 0.5677,
+      "step": 28667
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.452982906255742e-06,
+      "loss": 0.5451,
+      "step": 28668
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.452348038525079e-06,
+      "loss": 0.6007,
+      "step": 28669
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.451713193902373e-06,
+      "loss": 0.5837,
+      "step": 28670
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.451078372390843e-06,
+      "loss": 0.638,
+      "step": 28671
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4504435739937136e-06,
+      "loss": 0.5873,
+      "step": 28672
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4498087987142155e-06,
+      "loss": 0.6325,
+      "step": 28673
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.449174046555572e-06,
+      "loss": 0.5832,
+      "step": 28674
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.448539317521006e-06,
+      "loss": 0.6045,
+      "step": 28675
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.447904611613744e-06,
+      "loss": 0.6651,
+      "step": 28676
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.447269928837013e-06,
+      "loss": 0.6252,
+      "step": 28677
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.446635269194038e-06,
+      "loss": 0.5917,
+      "step": 28678
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.446000632688042e-06,
+      "loss": 0.6179,
+      "step": 28679
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.445366019322248e-06,
+      "loss": 0.5787,
+      "step": 28680
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4447314290998865e-06,
+      "loss": 0.5558,
+      "step": 28681
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.444096862024174e-06,
+      "loss": 0.5759,
+      "step": 28682
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.443462318098345e-06,
+      "loss": 0.5764,
+      "step": 28683
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.442827797325616e-06,
+      "loss": 0.607,
+      "step": 28684
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.442193299709215e-06,
+      "loss": 0.5957,
+      "step": 28685
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.441558825252362e-06,
+      "loss": 0.6072,
+      "step": 28686
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.440924373958285e-06,
+      "loss": 0.5597,
+      "step": 28687
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.440289945830204e-06,
+      "loss": 0.6514,
+      "step": 28688
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.439655540871354e-06,
+      "loss": 0.6206,
+      "step": 28689
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.439021159084941e-06,
+      "loss": 0.605,
+      "step": 28690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4383868004742e-06,
+      "loss": 0.5649,
+      "step": 28691
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.437752465042352e-06,
+      "loss": 0.6168,
+      "step": 28692
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.437118152792618e-06,
+      "loss": 0.5918,
+      "step": 28693
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.436483863728224e-06,
+      "loss": 0.6172,
+      "step": 28694
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.435849597852394e-06,
+      "loss": 0.5692,
+      "step": 28695
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.435215355168349e-06,
+      "loss": 0.6043,
+      "step": 28696
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4345811356793055e-06,
+      "loss": 0.5664,
+      "step": 28697
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.433946939388497e-06,
+      "loss": 0.6249,
+      "step": 28698
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.433312766299139e-06,
+      "loss": 0.6259,
+      "step": 28699
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4326786164144626e-06,
+      "loss": 0.6482,
+      "step": 28700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4320444897376765e-06,
+      "loss": 0.6027,
+      "step": 28701
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.431410386272013e-06,
+      "loss": 0.5698,
+      "step": 28702
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.430776306020689e-06,
+      "loss": 0.5578,
+      "step": 28703
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.430142248986931e-06,
+      "loss": 0.5708,
+      "step": 28704
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.429508215173961e-06,
+      "loss": 0.5823,
+      "step": 28705
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.428874204584997e-06,
+      "loss": 0.555,
+      "step": 28706
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.428240217223257e-06,
+      "loss": 0.6135,
+      "step": 28707
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.427606253091973e-06,
+      "loss": 0.6329,
+      "step": 28708
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.426972312194356e-06,
+      "loss": 0.6187,
+      "step": 28709
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.426338394533636e-06,
+      "loss": 0.6047,
+      "step": 28710
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.425704500113029e-06,
+      "loss": 0.6329,
+      "step": 28711
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.425070628935758e-06,
+      "loss": 0.6321,
+      "step": 28712
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.424436781005043e-06,
+      "loss": 0.5769,
+      "step": 28713
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.4238029563241e-06,
+      "loss": 0.5877,
+      "step": 28714
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.42316915489616e-06,
+      "loss": 0.5916,
+      "step": 28715
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.422535376724436e-06,
+      "loss": 0.6162,
+      "step": 28716
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.421901621812151e-06,
+      "loss": 0.5208,
+      "step": 28717
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.421267890162523e-06,
+      "loss": 0.6212,
+      "step": 28718
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.420634181778774e-06,
+      "loss": 0.6603,
+      "step": 28719
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.420000496664122e-06,
+      "loss": 0.6047,
+      "step": 28720
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.419366834821793e-06,
+      "loss": 0.5944,
+      "step": 28721
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.418733196255002e-06,
+      "loss": 0.594,
+      "step": 28722
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.418099580966969e-06,
+      "loss": 0.5426,
+      "step": 28723
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.417465988960911e-06,
+      "loss": 0.6612,
+      "step": 28724
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.416832420240052e-06,
+      "loss": 0.5733,
+      "step": 28725
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.416198874807608e-06,
+      "loss": 0.5999,
+      "step": 28726
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 5.415565352666807e-06,
+      "loss": 0.6207,
+      "step": 28727
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.414931853820853e-06,
+      "loss": 0.5723,
+      "step": 28728
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.414298378272977e-06,
+      "loss": 0.6541,
+      "step": 28729
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.4136649260263906e-06,
+      "loss": 0.5888,
+      "step": 28730
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.41303149708432e-06,
+      "loss": 0.598,
+      "step": 28731
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.412398091449979e-06,
+      "loss": 0.5785,
+      "step": 28732
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.411764709126587e-06,
+      "loss": 0.624,
+      "step": 28733
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.4111313501173625e-06,
+      "loss": 0.5751,
+      "step": 28734
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.410498014425518e-06,
+      "loss": 0.5621,
+      "step": 28735
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.4098647020542835e-06,
+      "loss": 0.6322,
+      "step": 28736
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.409231413006865e-06,
+      "loss": 0.5921,
+      "step": 28737
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.408598147286495e-06,
+      "loss": 0.6196,
+      "step": 28738
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.407964904896375e-06,
+      "loss": 0.6295,
+      "step": 28739
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.407331685839734e-06,
+      "loss": 0.5497,
+      "step": 28740
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.40669849011978e-06,
+      "loss": 0.6035,
+      "step": 28741
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.406065317739742e-06,
+      "loss": 0.64,
+      "step": 28742
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.405432168702831e-06,
+      "loss": 0.6455,
+      "step": 28743
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.4047990430122645e-06,
+      "loss": 0.5974,
+      "step": 28744
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.404165940671255e-06,
+      "loss": 0.6465,
+      "step": 28745
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.403532861683029e-06,
+      "loss": 0.6015,
+      "step": 28746
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.4028998060507965e-06,
+      "loss": 0.6374,
+      "step": 28747
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.402266773777779e-06,
+      "loss": 0.5997,
+      "step": 28748
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.40163376486719e-06,
+      "loss": 0.6293,
+      "step": 28749
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.401000779322248e-06,
+      "loss": 0.6313,
+      "step": 28750
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.400367817146166e-06,
+      "loss": 0.5894,
+      "step": 28751
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.399734878342159e-06,
+      "loss": 0.5815,
+      "step": 28752
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.399101962913451e-06,
+      "loss": 0.6583,
+      "step": 28753
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.398469070863253e-06,
+      "loss": 0.5979,
+      "step": 28754
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3978362021947795e-06,
+      "loss": 0.5548,
+      "step": 28755
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.397203356911247e-06,
+      "loss": 0.6265,
+      "step": 28756
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.396570535015875e-06,
+      "loss": 0.6228,
+      "step": 28757
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.395937736511871e-06,
+      "loss": 0.6351,
+      "step": 28758
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.395304961402461e-06,
+      "loss": 0.61,
+      "step": 28759
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.394672209690854e-06,
+      "loss": 0.5816,
+      "step": 28760
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3940394813802686e-06,
+      "loss": 0.6339,
+      "step": 28761
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3934067764739115e-06,
+      "loss": 0.6406,
+      "step": 28762
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.392774094975008e-06,
+      "loss": 0.6459,
+      "step": 28763
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.392141436886765e-06,
+      "loss": 0.6048,
+      "step": 28764
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.391508802212408e-06,
+      "loss": 0.5867,
+      "step": 28765
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.390876190955138e-06,
+      "loss": 0.6251,
+      "step": 28766
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.390243603118179e-06,
+      "loss": 0.631,
+      "step": 28767
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.389611038704739e-06,
+      "loss": 0.6578,
+      "step": 28768
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.388978497718038e-06,
+      "loss": 0.5852,
+      "step": 28769
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.388345980161288e-06,
+      "loss": 0.6568,
+      "step": 28770
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.387713486037703e-06,
+      "loss": 0.5585,
+      "step": 28771
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.387081015350497e-06,
+      "loss": 0.5656,
+      "step": 28772
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.386448568102879e-06,
+      "loss": 0.5873,
+      "step": 28773
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.38581614429807e-06,
+      "loss": 0.6214,
+      "step": 28774
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.385183743939279e-06,
+      "loss": 0.6169,
+      "step": 28775
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.384551367029727e-06,
+      "loss": 0.5929,
+      "step": 28776
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.383919013572613e-06,
+      "loss": 0.6373,
+      "step": 28777
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.383286683571163e-06,
+      "loss": 0.5941,
+      "step": 28778
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.382654377028582e-06,
+      "loss": 0.6124,
+      "step": 28779
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.38202209394809e-06,
+      "loss": 0.5808,
+      "step": 28780
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.381389834332895e-06,
+      "loss": 0.5933,
+      "step": 28781
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3807575981862125e-06,
+      "loss": 0.5978,
+      "step": 28782
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3801253855112475e-06,
+      "loss": 0.6078,
+      "step": 28783
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3794931963112255e-06,
+      "loss": 0.5772,
+      "step": 28784
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.378861030589345e-06,
+      "loss": 0.6158,
+      "step": 28785
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3782288883488295e-06,
+      "loss": 0.642,
+      "step": 28786
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.377596769592888e-06,
+      "loss": 0.5574,
+      "step": 28787
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3769646743247295e-06,
+      "loss": 0.576,
+      "step": 28788
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.376332602547564e-06,
+      "loss": 0.5904,
+      "step": 28789
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.37570055426461e-06,
+      "loss": 0.6078,
+      "step": 28790
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.375068529479077e-06,
+      "loss": 0.6018,
+      "step": 28791
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.374436528194173e-06,
+      "loss": 0.592,
+      "step": 28792
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.373804550413114e-06,
+      "loss": 0.5729,
+      "step": 28793
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.373172596139105e-06,
+      "loss": 0.6122,
+      "step": 28794
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.372540665375363e-06,
+      "loss": 0.6549,
+      "step": 28795
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.371908758125095e-06,
+      "loss": 0.5725,
+      "step": 28796
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.371276874391521e-06,
+      "loss": 0.5906,
+      "step": 28797
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.370645014177838e-06,
+      "loss": 0.5257,
+      "step": 28798
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.370013177487265e-06,
+      "loss": 0.6421,
+      "step": 28799
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.36938136432301e-06,
+      "loss": 0.5519,
+      "step": 28800
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.368749574688286e-06,
+      "loss": 0.6285,
+      "step": 28801
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3681178085863e-06,
+      "loss": 0.5872,
+      "step": 28802
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.36748606602027e-06,
+      "loss": 0.6292,
+      "step": 28803
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.366854346993393e-06,
+      "loss": 0.6636,
+      "step": 28804
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.366222651508889e-06,
+      "loss": 0.5967,
+      "step": 28805
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.365590979569961e-06,
+      "loss": 0.5659,
+      "step": 28806
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.364959331179826e-06,
+      "loss": 0.6066,
+      "step": 28807
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.364327706341692e-06,
+      "loss": 0.6441,
+      "step": 28808
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.363696105058765e-06,
+      "loss": 0.615,
+      "step": 28809
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.363064527334253e-06,
+      "loss": 0.6182,
+      "step": 28810
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.362432973171372e-06,
+      "loss": 0.6051,
+      "step": 28811
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.361801442573327e-06,
+      "loss": 0.613,
+      "step": 28812
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.361169935543322e-06,
+      "loss": 0.6062,
+      "step": 28813
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.36053845208458e-06,
+      "loss": 0.5511,
+      "step": 28814
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.359906992200292e-06,
+      "loss": 0.5759,
+      "step": 28815
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.359275555893682e-06,
+      "loss": 0.5332,
+      "step": 28816
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.358644143167946e-06,
+      "loss": 0.631,
+      "step": 28817
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.358012754026303e-06,
+      "loss": 0.5722,
+      "step": 28818
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.357381388471956e-06,
+      "loss": 0.5222,
+      "step": 28819
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.356750046508114e-06,
+      "loss": 0.5721,
+      "step": 28820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.356118728137981e-06,
+      "loss": 0.6552,
+      "step": 28821
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.355487433364773e-06,
+      "loss": 0.5726,
+      "step": 28822
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.35485616219169e-06,
+      "loss": 0.5791,
+      "step": 28823
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.354224914621946e-06,
+      "loss": 0.6491,
+      "step": 28824
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.353593690658747e-06,
+      "loss": 0.5929,
+      "step": 28825
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.352962490305298e-06,
+      "loss": 0.5374,
+      "step": 28826
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.352331313564805e-06,
+      "loss": 0.6045,
+      "step": 28827
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3517001604404805e-06,
+      "loss": 0.6233,
+      "step": 28828
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.351069030935529e-06,
+      "loss": 0.579,
+      "step": 28829
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.350437925053158e-06,
+      "loss": 0.5636,
+      "step": 28830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3498068427965685e-06,
+      "loss": 0.5902,
+      "step": 28831
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.349175784168977e-06,
+      "loss": 0.5916,
+      "step": 28832
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.348544749173585e-06,
+      "loss": 0.5838,
+      "step": 28833
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.347913737813597e-06,
+      "loss": 0.6183,
+      "step": 28834
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.347282750092227e-06,
+      "loss": 0.6243,
+      "step": 28835
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.34665178601267e-06,
+      "loss": 0.5707,
+      "step": 28836
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.346020845578142e-06,
+      "loss": 0.5623,
+      "step": 28837
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.34538992879184e-06,
+      "loss": 0.5926,
+      "step": 28838
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.344759035656979e-06,
+      "loss": 0.5965,
+      "step": 28839
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.344128166176757e-06,
+      "loss": 0.5495,
+      "step": 28840
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.343497320354392e-06,
+      "loss": 0.6442,
+      "step": 28841
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.342866498193072e-06,
+      "loss": 0.64,
+      "step": 28842
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.342235699696016e-06,
+      "loss": 0.5979,
+      "step": 28843
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.341604924866419e-06,
+      "loss": 0.5989,
+      "step": 28844
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.340974173707498e-06,
+      "loss": 0.5992,
+      "step": 28845
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.340343446222449e-06,
+      "loss": 0.5871,
+      "step": 28846
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.339712742414483e-06,
+      "loss": 0.5324,
+      "step": 28847
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.339082062286793e-06,
+      "loss": 0.6296,
+      "step": 28848
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.338451405842599e-06,
+      "loss": 0.5885,
+      "step": 28849
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.337820773085099e-06,
+      "loss": 0.5765,
+      "step": 28850
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.337190164017492e-06,
+      "loss": 0.6181,
+      "step": 28851
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.336559578642995e-06,
+      "loss": 0.6167,
+      "step": 28852
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.335929016964796e-06,
+      "loss": 0.6029,
+      "step": 28853
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3352984789861125e-06,
+      "loss": 0.6005,
+      "step": 28854
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.334667964710141e-06,
+      "loss": 0.5626,
+      "step": 28855
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.33403747414009e-06,
+      "loss": 0.6417,
+      "step": 28856
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.333407007279161e-06,
+      "loss": 0.5795,
+      "step": 28857
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.3327765641305575e-06,
+      "loss": 0.5722,
+      "step": 28858
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.332146144697481e-06,
+      "loss": 0.5453,
+      "step": 28859
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.331515748983139e-06,
+      "loss": 0.6063,
+      "step": 28860
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.33088537699073e-06,
+      "loss": 0.6214,
+      "step": 28861
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.330255028723468e-06,
+      "loss": 0.5934,
+      "step": 28862
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.329624704184538e-06,
+      "loss": 0.619,
+      "step": 28863
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.328994403377158e-06,
+      "loss": 0.6238,
+      "step": 28864
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.328364126304521e-06,
+      "loss": 0.6448,
+      "step": 28865
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.327733872969838e-06,
+      "loss": 0.6317,
+      "step": 28866
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.327103643376308e-06,
+      "loss": 0.6051,
+      "step": 28867
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.326473437527133e-06,
+      "loss": 0.5657,
+      "step": 28868
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.325843255425511e-06,
+      "loss": 0.6241,
+      "step": 28869
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.325213097074652e-06,
+      "loss": 0.5895,
+      "step": 28870
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.324582962477753e-06,
+      "loss": 0.6205,
+      "step": 28871
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.323952851638016e-06,
+      "loss": 0.6174,
+      "step": 28872
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.323322764558651e-06,
+      "loss": 0.608,
+      "step": 28873
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.322692701242844e-06,
+      "loss": 0.6426,
+      "step": 28874
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.32206266169381e-06,
+      "loss": 0.5393,
+      "step": 28875
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.321432645914741e-06,
+      "loss": 0.6357,
+      "step": 28876
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.320802653908846e-06,
+      "loss": 0.5953,
+      "step": 28877
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3201726856793215e-06,
+      "loss": 0.6071,
+      "step": 28878
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.319542741229375e-06,
+      "loss": 0.5641,
+      "step": 28879
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3189128205621965e-06,
+      "loss": 0.5986,
+      "step": 28880
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.318282923680997e-06,
+      "loss": 0.6543,
+      "step": 28881
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.317653050588969e-06,
+      "loss": 0.656,
+      "step": 28882
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.317023201289322e-06,
+      "loss": 0.6624,
+      "step": 28883
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.316393375785249e-06,
+      "loss": 0.5631,
+      "step": 28884
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.315763574079955e-06,
+      "loss": 0.5553,
+      "step": 28885
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.315133796176634e-06,
+      "loss": 0.5506,
+      "step": 28886
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3145040420784934e-06,
+      "loss": 0.5807,
+      "step": 28887
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3138743117887306e-06,
+      "loss": 0.5688,
+      "step": 28888
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.313244605310544e-06,
+      "loss": 0.6384,
+      "step": 28889
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.31261492264713e-06,
+      "loss": 0.6137,
+      "step": 28890
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.311985263801698e-06,
+      "loss": 0.6411,
+      "step": 28891
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.311355628777442e-06,
+      "loss": 0.6034,
+      "step": 28892
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.310726017577554e-06,
+      "loss": 0.6261,
+      "step": 28893
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.310096430205247e-06,
+      "loss": 0.5406,
+      "step": 28894
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.309466866663713e-06,
+      "loss": 0.637,
+      "step": 28895
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.308837326956151e-06,
+      "loss": 0.6128,
+      "step": 28896
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.308207811085757e-06,
+      "loss": 0.5703,
+      "step": 28897
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.307578319055737e-06,
+      "loss": 0.5939,
+      "step": 28898
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.306948850869283e-06,
+      "loss": 0.6312,
+      "step": 28899
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.306319406529603e-06,
+      "loss": 0.5828,
+      "step": 28900
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3056899860398815e-06,
+      "loss": 0.6114,
+      "step": 28901
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.305060589403327e-06,
+      "loss": 0.554,
+      "step": 28902
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.304431216623132e-06,
+      "loss": 0.6054,
+      "step": 28903
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3038018677024995e-06,
+      "loss": 0.5858,
+      "step": 28904
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.303172542644626e-06,
+      "loss": 0.6494,
+      "step": 28905
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.302543241452707e-06,
+      "loss": 0.5949,
+      "step": 28906
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3019139641299385e-06,
+      "loss": 0.614,
+      "step": 28907
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.3012847106795264e-06,
+      "loss": 0.5526,
+      "step": 28908
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.300655481104661e-06,
+      "loss": 0.5758,
+      "step": 28909
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.300026275408537e-06,
+      "loss": 0.5647,
+      "step": 28910
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2993970935943604e-06,
+      "loss": 0.5771,
+      "step": 28911
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.298767935665325e-06,
+      "loss": 0.6017,
+      "step": 28912
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.298138801624626e-06,
+      "loss": 0.6164,
+      "step": 28913
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.297509691475457e-06,
+      "loss": 0.5917,
+      "step": 28914
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.296880605221022e-06,
+      "loss": 0.5922,
+      "step": 28915
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.29625154286451e-06,
+      "loss": 0.6533,
+      "step": 28916
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.295622504409128e-06,
+      "loss": 0.6242,
+      "step": 28917
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.29499348985806e-06,
+      "loss": 0.5758,
+      "step": 28918
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.294364499214511e-06,
+      "loss": 0.6182,
+      "step": 28919
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.293735532481671e-06,
+      "loss": 0.6524,
+      "step": 28920
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.293106589662742e-06,
+      "loss": 0.558,
+      "step": 28921
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.292477670760917e-06,
+      "loss": 0.5618,
+      "step": 28922
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.291848775779391e-06,
+      "loss": 0.6195,
+      "step": 28923
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.291219904721358e-06,
+      "loss": 0.6323,
+      "step": 28924
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2905910575900184e-06,
+      "loss": 0.5904,
+      "step": 28925
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2899622343885655e-06,
+      "loss": 0.591,
+      "step": 28926
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.289333435120194e-06,
+      "loss": 0.6118,
+      "step": 28927
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2887046597880955e-06,
+      "loss": 0.6321,
+      "step": 28928
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.28807590839547e-06,
+      "loss": 0.5486,
+      "step": 28929
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.287447180945513e-06,
+      "loss": 0.5872,
+      "step": 28930
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.286818477441411e-06,
+      "loss": 0.5579,
+      "step": 28931
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2861897978863705e-06,
+      "loss": 0.6027,
+      "step": 28932
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2855611422835796e-06,
+      "loss": 0.6446,
+      "step": 28933
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.284932510636234e-06,
+      "loss": 0.6031,
+      "step": 28934
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.284303902947521e-06,
+      "loss": 0.5694,
+      "step": 28935
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.283675319220647e-06,
+      "loss": 0.6523,
+      "step": 28936
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2830467594587945e-06,
+      "loss": 0.6759,
+      "step": 28937
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.282418223665171e-06,
+      "loss": 0.523,
+      "step": 28938
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.281789711842955e-06,
+      "loss": 0.6266,
+      "step": 28939
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.281161223995351e-06,
+      "loss": 0.5917,
+      "step": 28940
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2805327601255455e-06,
+      "loss": 0.5871,
+      "step": 28941
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.279904320236737e-06,
+      "loss": 0.5831,
+      "step": 28942
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.279275904332119e-06,
+      "loss": 0.5272,
+      "step": 28943
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.278647512414883e-06,
+      "loss": 0.5814,
+      "step": 28944
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2780191444882155e-06,
+      "loss": 0.5465,
+      "step": 28945
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.277390800555321e-06,
+      "loss": 0.6071,
+      "step": 28946
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.276762480619386e-06,
+      "loss": 0.6193,
+      "step": 28947
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.276134184683601e-06,
+      "loss": 0.5883,
+      "step": 28948
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.275505912751167e-06,
+      "loss": 0.5489,
+      "step": 28949
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.274877664825269e-06,
+      "loss": 0.6017,
+      "step": 28950
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.274249440909102e-06,
+      "loss": 0.6241,
+      "step": 28951
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.273621241005853e-06,
+      "loss": 0.586,
+      "step": 28952
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.272993065118723e-06,
+      "loss": 0.5667,
+      "step": 28953
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2723649132509e-06,
+      "loss": 0.6462,
+      "step": 28954
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.271736785405575e-06,
+      "loss": 0.5753,
+      "step": 28955
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.271108681585937e-06,
+      "loss": 0.6656,
+      "step": 28956
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.270480601795185e-06,
+      "loss": 0.6051,
+      "step": 28957
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2698525460365e-06,
+      "loss": 0.6158,
+      "step": 28958
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.269224514313085e-06,
+      "loss": 0.6199,
+      "step": 28959
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.268596506628125e-06,
+      "loss": 0.5795,
+      "step": 28960
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.267968522984813e-06,
+      "loss": 0.6643,
+      "step": 28961
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.267340563386334e-06,
+      "loss": 0.5979,
+      "step": 28962
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.266712627835889e-06,
+      "loss": 0.6158,
+      "step": 28963
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.266084716336661e-06,
+      "loss": 0.5841,
+      "step": 28964
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.265456828891844e-06,
+      "loss": 0.5557,
+      "step": 28965
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.264828965504624e-06,
+      "loss": 0.5996,
+      "step": 28966
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2642011261781975e-06,
+      "loss": 0.6259,
+      "step": 28967
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2635733109157526e-06,
+      "loss": 0.618,
+      "step": 28968
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.262945519720475e-06,
+      "loss": 0.6416,
+      "step": 28969
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.262317752595563e-06,
+      "loss": 0.5481,
+      "step": 28970
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.261690009544201e-06,
+      "loss": 0.5758,
+      "step": 28971
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.261062290569581e-06,
+      "loss": 0.5776,
+      "step": 28972
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.260434595674886e-06,
+      "loss": 0.5918,
+      "step": 28973
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.259806924863316e-06,
+      "loss": 0.5434,
+      "step": 28974
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.259179278138051e-06,
+      "loss": 0.5537,
+      "step": 28975
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.258551655502291e-06,
+      "loss": 0.6373,
+      "step": 28976
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.257924056959212e-06,
+      "loss": 0.6275,
+      "step": 28977
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.257296482512015e-06,
+      "loss": 0.6621,
+      "step": 28978
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2566689321638775e-06,
+      "loss": 0.6073,
+      "step": 28979
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2560414059179985e-06,
+      "loss": 0.6815,
+      "step": 28980
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.255413903777562e-06,
+      "loss": 0.5866,
+      "step": 28981
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.254786425745759e-06,
+      "loss": 0.6025,
+      "step": 28982
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.25415897182577e-06,
+      "loss": 0.6065,
+      "step": 28983
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.253531542020793e-06,
+      "loss": 0.5819,
+      "step": 28984
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.252904136334014e-06,
+      "loss": 0.6187,
+      "step": 28985
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2522767547686145e-06,
+      "loss": 0.5872,
+      "step": 28986
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.251649397327792e-06,
+      "loss": 0.553,
+      "step": 28987
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.251022064014728e-06,
+      "loss": 0.5703,
+      "step": 28988
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.250394754832613e-06,
+      "loss": 0.5441,
+      "step": 28989
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2497674697846275e-06,
+      "loss": 0.6046,
+      "step": 28990
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.249140208873971e-06,
+      "loss": 0.6301,
+      "step": 28991
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2485129721038215e-06,
+      "loss": 0.6031,
+      "step": 28992
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2478857594773716e-06,
+      "loss": 0.6026,
+      "step": 28993
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.247258570997802e-06,
+      "loss": 0.6384,
+      "step": 28994
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.246631406668305e-06,
+      "loss": 0.5466,
+      "step": 28995
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.246004266492064e-06,
+      "loss": 0.5844,
+      "step": 28996
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.245377150472272e-06,
+      "loss": 0.5971,
+      "step": 28997
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.244750058612109e-06,
+      "loss": 0.5699,
+      "step": 28998
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.244122990914766e-06,
+      "loss": 0.6096,
+      "step": 28999
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.243495947383421e-06,
+      "loss": 0.6293,
+      "step": 29000
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.242868928021272e-06,
+      "loss": 0.6439,
+      "step": 29001
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.242241932831493e-06,
+      "loss": 0.6016,
+      "step": 29002
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2416149618172855e-06,
+      "loss": 0.6789,
+      "step": 29003
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.240988014981818e-06,
+      "loss": 0.5761,
+      "step": 29004
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.240361092328288e-06,
+      "loss": 0.5656,
+      "step": 29005
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2397341938598755e-06,
+      "loss": 0.6525,
+      "step": 29006
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.239107319579766e-06,
+      "loss": 0.6479,
+      "step": 29007
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.23848046949115e-06,
+      "loss": 0.6279,
+      "step": 29008
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.237853643597208e-06,
+      "loss": 0.5079,
+      "step": 29009
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.237226841901128e-06,
+      "loss": 0.62,
+      "step": 29010
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.236600064406087e-06,
+      "loss": 0.6037,
+      "step": 29011
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2359733111152835e-06,
+      "loss": 0.6113,
+      "step": 29012
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2353465820318885e-06,
+      "loss": 0.6287,
+      "step": 29013
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2347198771591e-06,
+      "loss": 0.5989,
+      "step": 29014
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.23409319650009e-06,
+      "loss": 0.6546,
+      "step": 29015
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.23346654005805e-06,
+      "loss": 0.6383,
+      "step": 29016
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.232839907836159e-06,
+      "loss": 0.5941,
+      "step": 29017
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.232213299837609e-06,
+      "loss": 0.6315,
+      "step": 29018
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.2315867160655795e-06,
+      "loss": 0.6602,
+      "step": 29019
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.230960156523255e-06,
+      "loss": 0.6399,
+      "step": 29020
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.230333621213814e-06,
+      "loss": 0.6087,
+      "step": 29021
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.229707110140447e-06,
+      "loss": 0.6284,
+      "step": 29022
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.229080623306334e-06,
+      "loss": 0.633,
+      "step": 29023
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.228454160714662e-06,
+      "loss": 0.5636,
+      "step": 29024
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.227827722368612e-06,
+      "loss": 0.5808,
+      "step": 29025
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.227201308271368e-06,
+      "loss": 0.5728,
+      "step": 29026
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.226574918426111e-06,
+      "loss": 0.5707,
+      "step": 29027
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.22594855283602e-06,
+      "loss": 0.638,
+      "step": 29028
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.225322211504288e-06,
+      "loss": 0.5617,
+      "step": 29029
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.224695894434092e-06,
+      "loss": 0.5733,
+      "step": 29030
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.224069601628614e-06,
+      "loss": 0.6525,
+      "step": 29031
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.223443333091034e-06,
+      "loss": 0.5754,
+      "step": 29032
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.22281708882454e-06,
+      "loss": 0.5897,
+      "step": 29033
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.222190868832309e-06,
+      "loss": 0.6263,
+      "step": 29034
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.221564673117528e-06,
+      "loss": 0.5618,
+      "step": 29035
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.220938501683377e-06,
+      "loss": 0.6497,
+      "step": 29036
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.220312354533037e-06,
+      "loss": 0.581,
+      "step": 29037
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.219686231669685e-06,
+      "loss": 0.6133,
+      "step": 29038
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.219060133096512e-06,
+      "loss": 0.6208,
+      "step": 29039
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.21843405881669e-06,
+      "loss": 0.6626,
+      "step": 29040
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.217808008833414e-06,
+      "loss": 0.5317,
+      "step": 29041
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.217181983149848e-06,
+      "loss": 0.6004,
+      "step": 29042
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.216555981769184e-06,
+      "loss": 0.5691,
+      "step": 29043
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.2159300046946e-06,
+      "loss": 0.655,
+      "step": 29044
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.215304051929274e-06,
+      "loss": 0.6202,
+      "step": 29045
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.214678123476393e-06,
+      "loss": 0.6497,
+      "step": 29046
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.214052219339133e-06,
+      "loss": 0.65,
+      "step": 29047
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.213426339520675e-06,
+      "loss": 0.6367,
+      "step": 29048
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.212800484024196e-06,
+      "loss": 0.6132,
+      "step": 29049
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.212174652852885e-06,
+      "loss": 0.6169,
+      "step": 29050
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.211548846009913e-06,
+      "loss": 0.6733,
+      "step": 29051
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.2109230634984694e-06,
+      "loss": 0.5695,
+      "step": 29052
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.210297305321721e-06,
+      "loss": 0.5965,
+      "step": 29053
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.209671571482859e-06,
+      "loss": 0.5757,
+      "step": 29054
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.209045861985056e-06,
+      "loss": 0.6378,
+      "step": 29055
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.208420176831497e-06,
+      "loss": 0.6544,
+      "step": 29056
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.207794516025357e-06,
+      "loss": 0.6429,
+      "step": 29057
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.207168879569819e-06,
+      "loss": 0.6036,
+      "step": 29058
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.206543267468054e-06,
+      "loss": 0.6185,
+      "step": 29059
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.205917679723252e-06,
+      "loss": 0.6272,
+      "step": 29060
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.205292116338579e-06,
+      "loss": 0.5851,
+      "step": 29061
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.204666577317228e-06,
+      "loss": 0.5348,
+      "step": 29062
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.204041062662369e-06,
+      "loss": 0.6304,
+      "step": 29063
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.203415572377183e-06,
+      "loss": 0.5724,
+      "step": 29064
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.202790106464846e-06,
+      "loss": 0.5143,
+      "step": 29065
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.202164664928535e-06,
+      "loss": 0.5757,
+      "step": 29066
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.2015392477714326e-06,
+      "loss": 0.6406,
+      "step": 29067
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.2009138549967165e-06,
+      "loss": 0.6759,
+      "step": 29068
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.200288486607561e-06,
+      "loss": 0.5916,
+      "step": 29069
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.199663142607142e-06,
+      "loss": 0.5679,
+      "step": 29070
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.199037822998645e-06,
+      "loss": 0.6252,
+      "step": 29071
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.198412527785237e-06,
+      "loss": 0.5968,
+      "step": 29072
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.19778725697011e-06,
+      "loss": 0.5989,
+      "step": 29073
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.197162010556426e-06,
+      "loss": 0.5858,
+      "step": 29074
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.196536788547371e-06,
+      "loss": 0.6598,
+      "step": 29075
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.195911590946114e-06,
+      "loss": 0.6088,
+      "step": 29076
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.195286417755844e-06,
+      "loss": 0.6129,
+      "step": 29077
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.194661268979726e-06,
+      "loss": 0.6373,
+      "step": 29078
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.194036144620948e-06,
+      "loss": 0.5872,
+      "step": 29079
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.193411044682672e-06,
+      "loss": 0.5729,
+      "step": 29080
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1927859691680885e-06,
+      "loss": 0.5833,
+      "step": 29081
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.19216091808036e-06,
+      "loss": 0.6154,
+      "step": 29082
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.191535891422676e-06,
+      "loss": 0.5337,
+      "step": 29083
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.190910889198206e-06,
+      "loss": 0.6361,
+      "step": 29084
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.190285911410127e-06,
+      "loss": 0.6173,
+      "step": 29085
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.189660958061613e-06,
+      "loss": 0.5764,
+      "step": 29086
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.189036029155836e-06,
+      "loss": 0.561,
+      "step": 29087
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.18841112469598e-06,
+      "loss": 0.6131,
+      "step": 29088
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1877862446852134e-06,
+      "loss": 0.6531,
+      "step": 29089
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.18716138912672e-06,
+      "loss": 0.5574,
+      "step": 29090
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.186536558023663e-06,
+      "loss": 0.6768,
+      "step": 29091
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1859117513792265e-06,
+      "loss": 0.6143,
+      "step": 29092
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.185286969196579e-06,
+      "loss": 0.6269,
+      "step": 29093
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.184662211478902e-06,
+      "loss": 0.599,
+      "step": 29094
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.184037478229367e-06,
+      "loss": 0.5744,
+      "step": 29095
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.183412769451149e-06,
+      "loss": 0.5816,
+      "step": 29096
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.182788085147415e-06,
+      "loss": 0.6041,
+      "step": 29097
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.182163425321351e-06,
+      "loss": 0.6241,
+      "step": 29098
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.18153878997612e-06,
+      "loss": 0.6028,
+      "step": 29099
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1809141791149065e-06,
+      "loss": 0.5494,
+      "step": 29100
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.18028959274088e-06,
+      "loss": 0.614,
+      "step": 29101
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.179665030857213e-06,
+      "loss": 0.5975,
+      "step": 29102
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.179040493467076e-06,
+      "loss": 0.6316,
+      "step": 29103
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.17841598057365e-06,
+      "loss": 0.5704,
+      "step": 29104
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1777914921801045e-06,
+      "loss": 0.5446,
+      "step": 29105
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.177167028289613e-06,
+      "loss": 0.5903,
+      "step": 29106
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.176542588905346e-06,
+      "loss": 0.6159,
+      "step": 29107
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.175918174030479e-06,
+      "loss": 0.5885,
+      "step": 29108
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.175293783668185e-06,
+      "loss": 0.646,
+      "step": 29109
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.174669417821634e-06,
+      "loss": 0.5417,
+      "step": 29110
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.174045076494008e-06,
+      "loss": 0.5853,
+      "step": 29111
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.173420759688464e-06,
+      "loss": 0.5295,
+      "step": 29112
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1727964674081876e-06,
+      "loss": 0.6564,
+      "step": 29113
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.172172199656342e-06,
+      "loss": 0.6397,
+      "step": 29114
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.171547956436107e-06,
+      "loss": 0.6234,
+      "step": 29115
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.170923737750647e-06,
+      "loss": 0.63,
+      "step": 29116
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.170299543603146e-06,
+      "loss": 0.607,
+      "step": 29117
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.169675373996759e-06,
+      "loss": 0.5488,
+      "step": 29118
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.169051228934669e-06,
+      "loss": 0.6003,
+      "step": 29119
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.168427108420042e-06,
+      "loss": 0.5877,
+      "step": 29120
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.167803012456056e-06,
+      "loss": 0.5907,
+      "step": 29121
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.167178941045879e-06,
+      "loss": 0.5857,
+      "step": 29122
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.166554894192679e-06,
+      "loss": 0.5991,
+      "step": 29123
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.165930871899629e-06,
+      "loss": 0.5954,
+      "step": 29124
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1653068741698975e-06,
+      "loss": 0.5952,
+      "step": 29125
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1646829010066615e-06,
+      "loss": 0.5429,
+      "step": 29126
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.164058952413082e-06,
+      "loss": 0.6121,
+      "step": 29127
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.163435028392345e-06,
+      "loss": 0.5273,
+      "step": 29128
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1628111289476025e-06,
+      "loss": 0.6178,
+      "step": 29129
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1621872540820386e-06,
+      "loss": 0.5631,
+      "step": 29130
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.161563403798812e-06,
+      "loss": 0.6099,
+      "step": 29131
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.160939578101104e-06,
+      "loss": 0.5691,
+      "step": 29132
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.16031577699208e-06,
+      "loss": 0.6765,
+      "step": 29133
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.159692000474907e-06,
+      "loss": 0.6217,
+      "step": 29134
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.159068248552753e-06,
+      "loss": 0.5969,
+      "step": 29135
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.158444521228796e-06,
+      "loss": 0.6285,
+      "step": 29136
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.157820818506195e-06,
+      "loss": 0.5896,
+      "step": 29137
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.157197140388133e-06,
+      "loss": 0.6517,
+      "step": 29138
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.156573486877763e-06,
+      "loss": 0.6215,
+      "step": 29139
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.155949857978264e-06,
+      "loss": 0.5689,
+      "step": 29140
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.155326253692801e-06,
+      "loss": 0.5596,
+      "step": 29141
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1547026740245455e-06,
+      "loss": 0.5817,
+      "step": 29142
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.154079118976667e-06,
+      "loss": 0.6452,
+      "step": 29143
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.15345558855233e-06,
+      "loss": 0.5983,
+      "step": 29144
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1528320827547065e-06,
+      "loss": 0.6053,
+      "step": 29145
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.152208601586956e-06,
+      "loss": 0.5758,
+      "step": 29146
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.151585145052259e-06,
+      "loss": 0.5845,
+      "step": 29147
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.150961713153774e-06,
+      "loss": 0.5616,
+      "step": 29148
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.15033830589468e-06,
+      "loss": 0.5617,
+      "step": 29149
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.149714923278129e-06,
+      "loss": 0.5586,
+      "step": 29150
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.149091565307303e-06,
+      "loss": 0.6283,
+      "step": 29151
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.148468231985357e-06,
+      "loss": 0.7281,
+      "step": 29152
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.14784492331547e-06,
+      "loss": 0.5668,
+      "step": 29153
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1472216393008e-06,
+      "loss": 0.6119,
+      "step": 29154
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.146598379944525e-06,
+      "loss": 0.6324,
+      "step": 29155
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.145975145249799e-06,
+      "loss": 0.623,
+      "step": 29156
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.145351935219796e-06,
+      "loss": 0.5889,
+      "step": 29157
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.14472874985768e-06,
+      "loss": 0.5872,
+      "step": 29158
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.144105589166623e-06,
+      "loss": 0.6299,
+      "step": 29159
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1434824531497865e-06,
+      "loss": 0.6234,
+      "step": 29160
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.142859341810338e-06,
+      "loss": 0.6189,
+      "step": 29161
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.142236255151441e-06,
+      "loss": 0.6071,
+      "step": 29162
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.141613193176267e-06,
+      "loss": 0.6121,
+      "step": 29163
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.14099015588798e-06,
+      "loss": 0.6516,
+      "step": 29164
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.140367143289744e-06,
+      "loss": 0.628,
+      "step": 29165
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.139744155384726e-06,
+      "loss": 0.6549,
+      "step": 29166
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.139121192176087e-06,
+      "loss": 0.6167,
+      "step": 29167
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 5.1384982536670015e-06,
+      "loss": 0.6157,
+      "step": 29168
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.137875339860627e-06,
+      "loss": 0.5778,
+      "step": 29169
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.137252450760135e-06,
+      "loss": 0.6099,
+      "step": 29170
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.136629586368687e-06,
+      "loss": 0.5367,
+      "step": 29171
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.136006746689448e-06,
+      "loss": 0.6129,
+      "step": 29172
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1353839317255805e-06,
+      "loss": 0.5014,
+      "step": 29173
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.134761141480256e-06,
+      "loss": 0.626,
+      "step": 29174
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.13413837595663e-06,
+      "loss": 0.6152,
+      "step": 29175
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1335156351578795e-06,
+      "loss": 0.6493,
+      "step": 29176
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1328929190871545e-06,
+      "loss": 0.6275,
+      "step": 29177
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.13227022774763e-06,
+      "loss": 0.5739,
+      "step": 29178
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.131647561142461e-06,
+      "loss": 0.6746,
+      "step": 29179
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.131024919274821e-06,
+      "loss": 0.6134,
+      "step": 29180
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.13040230214787e-06,
+      "loss": 0.6381,
+      "step": 29181
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.12977970976477e-06,
+      "loss": 0.6516,
+      "step": 29182
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.129157142128683e-06,
+      "loss": 0.5993,
+      "step": 29183
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.128534599242778e-06,
+      "loss": 0.5704,
+      "step": 29184
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.127912081110216e-06,
+      "loss": 0.6007,
+      "step": 29185
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.127289587734154e-06,
+      "loss": 0.6008,
+      "step": 29186
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.12666711911777e-06,
+      "loss": 0.5556,
+      "step": 29187
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.126044675264209e-06,
+      "loss": 0.6356,
+      "step": 29188
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1254222561766465e-06,
+      "loss": 0.5512,
+      "step": 29189
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.124799861858237e-06,
+      "loss": 0.5809,
+      "step": 29190
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.124177492312152e-06,
+      "loss": 0.6211,
+      "step": 29191
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.123555147541545e-06,
+      "loss": 0.5596,
+      "step": 29192
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.122932827549589e-06,
+      "loss": 0.6154,
+      "step": 29193
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.122310532339433e-06,
+      "loss": 0.5915,
+      "step": 29194
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.121688261914249e-06,
+      "loss": 0.6149,
+      "step": 29195
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.121066016277192e-06,
+      "loss": 0.5623,
+      "step": 29196
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.120443795431431e-06,
+      "loss": 0.6358,
+      "step": 29197
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.119821599380126e-06,
+      "loss": 0.6226,
+      "step": 29198
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.119199428126434e-06,
+      "loss": 0.5719,
+      "step": 29199
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.118577281673517e-06,
+      "loss": 0.5725,
+      "step": 29200
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.117955160024541e-06,
+      "loss": 0.5509,
+      "step": 29201
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1173330631826655e-06,
+      "loss": 0.5827,
+      "step": 29202
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.116710991151051e-06,
+      "loss": 0.5636,
+      "step": 29203
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.116088943932853e-06,
+      "loss": 0.5962,
+      "step": 29204
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.115466921531242e-06,
+      "loss": 0.5952,
+      "step": 29205
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.114844923949373e-06,
+      "loss": 0.6194,
+      "step": 29206
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.114222951190406e-06,
+      "loss": 0.5566,
+      "step": 29207
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.113601003257506e-06,
+      "loss": 0.6432,
+      "step": 29208
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.11297908015383e-06,
+      "loss": 0.6377,
+      "step": 29209
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.112357181882538e-06,
+      "loss": 0.5527,
+      "step": 29210
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.111735308446788e-06,
+      "loss": 0.6453,
+      "step": 29211
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.111113459849745e-06,
+      "loss": 0.5737,
+      "step": 29212
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.110491636094563e-06,
+      "loss": 0.6409,
+      "step": 29213
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.109869837184413e-06,
+      "loss": 0.5245,
+      "step": 29214
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.109248063122438e-06,
+      "loss": 0.556,
+      "step": 29215
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.108626313911811e-06,
+      "loss": 0.5422,
+      "step": 29216
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.108004589555682e-06,
+      "loss": 0.5963,
+      "step": 29217
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.107382890057218e-06,
+      "loss": 0.6394,
+      "step": 29218
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.106761215419573e-06,
+      "loss": 0.6173,
+      "step": 29219
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.106139565645909e-06,
+      "loss": 0.5627,
+      "step": 29220
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.105517940739378e-06,
+      "loss": 0.5901,
+      "step": 29221
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.104896340703149e-06,
+      "loss": 0.539,
+      "step": 29222
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.104274765540374e-06,
+      "loss": 0.583,
+      "step": 29223
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.103653215254209e-06,
+      "loss": 0.6071,
+      "step": 29224
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.103031689847826e-06,
+      "loss": 0.6522,
+      "step": 29225
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.102410189324363e-06,
+      "loss": 0.6793,
+      "step": 29226
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.101788713686993e-06,
+      "loss": 0.6267,
+      "step": 29227
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.1011672629388645e-06,
+      "loss": 0.5879,
+      "step": 29228
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.100545837083144e-06,
+      "loss": 0.6288,
+      "step": 29229
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.099924436122985e-06,
+      "loss": 0.6157,
+      "step": 29230
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.099303060061545e-06,
+      "loss": 0.618,
+      "step": 29231
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.098681708901978e-06,
+      "loss": 0.6101,
+      "step": 29232
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0980603826474474e-06,
+      "loss": 0.634,
+      "step": 29233
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.097439081301105e-06,
+      "loss": 0.5436,
+      "step": 29234
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.096817804866113e-06,
+      "loss": 0.6427,
+      "step": 29235
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.096196553345627e-06,
+      "loss": 0.5689,
+      "step": 29236
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.095575326742801e-06,
+      "loss": 0.5861,
+      "step": 29237
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0949541250607906e-06,
+      "loss": 0.5907,
+      "step": 29238
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0943329483027585e-06,
+      "loss": 0.6831,
+      "step": 29239
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.093711796471857e-06,
+      "loss": 0.5407,
+      "step": 29240
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.093090669571242e-06,
+      "loss": 0.5786,
+      "step": 29241
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.092469567604068e-06,
+      "loss": 0.6521,
+      "step": 29242
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.091848490573497e-06,
+      "loss": 0.5878,
+      "step": 29243
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0912274384826795e-06,
+      "loss": 0.6019,
+      "step": 29244
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.090606411334771e-06,
+      "loss": 0.5745,
+      "step": 29245
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.089985409132931e-06,
+      "loss": 0.5661,
+      "step": 29246
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.089364431880315e-06,
+      "loss": 0.6258,
+      "step": 29247
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.088743479580076e-06,
+      "loss": 0.5791,
+      "step": 29248
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.088122552235366e-06,
+      "loss": 0.6218,
+      "step": 29249
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0875016498493466e-06,
+      "loss": 0.6192,
+      "step": 29250
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.086880772425167e-06,
+      "loss": 0.6273,
+      "step": 29251
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.086259919965993e-06,
+      "loss": 0.5747,
+      "step": 29252
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0856390924749636e-06,
+      "loss": 0.5785,
+      "step": 29253
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.085018289955245e-06,
+      "loss": 0.5581,
+      "step": 29254
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.084397512409983e-06,
+      "loss": 0.5354,
+      "step": 29255
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0837767598423425e-06,
+      "loss": 0.6398,
+      "step": 29256
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.083156032255471e-06,
+      "loss": 0.5963,
+      "step": 29257
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0825353296525235e-06,
+      "loss": 0.6176,
+      "step": 29258
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.08191465203665e-06,
+      "loss": 0.6326,
+      "step": 29259
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.081293999411013e-06,
+      "loss": 0.5852,
+      "step": 29260
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.080673371778762e-06,
+      "loss": 0.5978,
+      "step": 29261
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.080052769143047e-06,
+      "loss": 0.5864,
+      "step": 29262
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.079432191507027e-06,
+      "loss": 0.5864,
+      "step": 29263
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.078811638873853e-06,
+      "loss": 0.6576,
+      "step": 29264
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.07819111124668e-06,
+      "loss": 0.5499,
+      "step": 29265
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.077570608628655e-06,
+      "loss": 0.557,
+      "step": 29266
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.076950131022937e-06,
+      "loss": 0.6019,
+      "step": 29267
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.07632967843268e-06,
+      "loss": 0.6504,
+      "step": 29268
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0757092508610325e-06,
+      "loss": 0.6328,
+      "step": 29269
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.075088848311145e-06,
+      "loss": 0.5983,
+      "step": 29270
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.074468470786178e-06,
+      "loss": 0.6334,
+      "step": 29271
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.073848118289274e-06,
+      "loss": 0.574,
+      "step": 29272
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.073227790823595e-06,
+      "loss": 0.5552,
+      "step": 29273
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.072607488392287e-06,
+      "loss": 0.6728,
+      "step": 29274
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.071987210998506e-06,
+      "loss": 0.6613,
+      "step": 29275
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.071366958645396e-06,
+      "loss": 0.638,
+      "step": 29276
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.070746731336118e-06,
+      "loss": 0.6122,
+      "step": 29277
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.070126529073818e-06,
+      "loss": 0.611,
+      "step": 29278
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.06950635186165e-06,
+      "loss": 0.7035,
+      "step": 29279
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.068886199702761e-06,
+      "loss": 0.5709,
+      "step": 29280
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.068266072600308e-06,
+      "loss": 0.6148,
+      "step": 29281
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.06764597055744e-06,
+      "loss": 0.5753,
+      "step": 29282
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.067025893577304e-06,
+      "loss": 0.5743,
+      "step": 29283
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.066405841663058e-06,
+      "loss": 0.6215,
+      "step": 29284
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.065785814817847e-06,
+      "loss": 0.5627,
+      "step": 29285
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.065165813044826e-06,
+      "loss": 0.6439,
+      "step": 29286
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.064545836347136e-06,
+      "loss": 0.6972,
+      "step": 29287
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.06392588472794e-06,
+      "loss": 0.6301,
+      "step": 29288
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.063305958190378e-06,
+      "loss": 0.6259,
+      "step": 29289
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.062686056737612e-06,
+      "loss": 0.5322,
+      "step": 29290
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.062066180372776e-06,
+      "loss": 0.6302,
+      "step": 29291
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.061446329099031e-06,
+      "loss": 0.6016,
+      "step": 29292
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.06082650291952e-06,
+      "loss": 0.5934,
+      "step": 29293
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.060206701837401e-06,
+      "loss": 0.6066,
+      "step": 29294
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.059586925855818e-06,
+      "loss": 0.5769,
+      "step": 29295
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.05896717497792e-06,
+      "loss": 0.5864,
+      "step": 29296
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.058347449206854e-06,
+      "loss": 0.6241,
+      "step": 29297
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.057727748545776e-06,
+      "loss": 0.582,
+      "step": 29298
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.057108072997829e-06,
+      "loss": 0.5375,
+      "step": 29299
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0564884225661606e-06,
+      "loss": 0.6084,
+      "step": 29300
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.055868797253926e-06,
+      "loss": 0.6175,
+      "step": 29301
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0552491970642694e-06,
+      "loss": 0.6012,
+      "step": 29302
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.054629622000341e-06,
+      "loss": 0.567,
+      "step": 29303
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.054010072065283e-06,
+      "loss": 0.5741,
+      "step": 29304
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.053390547262253e-06,
+      "loss": 0.5687,
+      "step": 29305
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.052771047594393e-06,
+      "loss": 0.5954,
+      "step": 29306
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.052151573064854e-06,
+      "loss": 0.5722,
+      "step": 29307
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.051532123676776e-06,
+      "loss": 0.6412,
+      "step": 29308
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.050912699433318e-06,
+      "loss": 0.6408,
+      "step": 29309
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.050293300337618e-06,
+      "loss": 0.6078,
+      "step": 29310
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.049673926392831e-06,
+      "loss": 0.5863,
+      "step": 29311
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.049054577602099e-06,
+      "loss": 0.5754,
+      "step": 29312
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.0484352539685725e-06,
+      "loss": 0.6139,
+      "step": 29313
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.047815955495392e-06,
+      "loss": 0.5737,
+      "step": 29314
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.047196682185713e-06,
+      "loss": 0.631,
+      "step": 29315
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.046577434042679e-06,
+      "loss": 0.6322,
+      "step": 29316
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.045958211069434e-06,
+      "loss": 0.6451,
+      "step": 29317
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.045339013269124e-06,
+      "loss": 0.5663,
+      "step": 29318
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0447198406449e-06,
+      "loss": 0.6025,
+      "step": 29319
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.044100693199906e-06,
+      "loss": 0.6328,
+      "step": 29320
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.043481570937284e-06,
+      "loss": 0.6199,
+      "step": 29321
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.042862473860188e-06,
+      "loss": 0.5888,
+      "step": 29322
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.042243401971759e-06,
+      "loss": 0.6627,
+      "step": 29323
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.041624355275143e-06,
+      "loss": 0.6577,
+      "step": 29324
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.041005333773483e-06,
+      "loss": 0.6035,
+      "step": 29325
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0403863374699304e-06,
+      "loss": 0.5339,
+      "step": 29326
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.039767366367624e-06,
+      "loss": 0.6228,
+      "step": 29327
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.039148420469719e-06,
+      "loss": 0.6194,
+      "step": 29328
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.038529499779346e-06,
+      "loss": 0.6344,
+      "step": 29329
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.037910604299663e-06,
+      "loss": 0.6284,
+      "step": 29330
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.037291734033804e-06,
+      "loss": 0.5885,
+      "step": 29331
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0366728889849235e-06,
+      "loss": 0.5418,
+      "step": 29332
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0360540691561625e-06,
+      "loss": 0.5873,
+      "step": 29333
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.035435274550664e-06,
+      "loss": 0.5998,
+      "step": 29334
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.034816505171568e-06,
+      "loss": 0.604,
+      "step": 29335
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.034197761022029e-06,
+      "loss": 0.6497,
+      "step": 29336
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.033579042105183e-06,
+      "loss": 0.6292,
+      "step": 29337
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.032960348424174e-06,
+      "loss": 0.5857,
+      "step": 29338
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0323416799821535e-06,
+      "loss": 0.5986,
+      "step": 29339
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0317230367822575e-06,
+      "loss": 0.62,
+      "step": 29340
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.031104418827632e-06,
+      "loss": 0.5308,
+      "step": 29341
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0304858261214175e-06,
+      "loss": 0.5691,
+      "step": 29342
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.029867258666763e-06,
+      "loss": 0.6073,
+      "step": 29343
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.029248716466809e-06,
+      "loss": 0.5944,
+      "step": 29344
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0286301995246975e-06,
+      "loss": 0.6313,
+      "step": 29345
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.028011707843568e-06,
+      "loss": 0.6173,
+      "step": 29346
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.027393241426572e-06,
+      "loss": 0.6191,
+      "step": 29347
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.026774800276842e-06,
+      "loss": 0.604,
+      "step": 29348
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.026156384397533e-06,
+      "loss": 0.6233,
+      "step": 29349
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.025537993791774e-06,
+      "loss": 0.6309,
+      "step": 29350
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.024919628462717e-06,
+      "loss": 0.5901,
+      "step": 29351
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.024301288413496e-06,
+      "loss": 0.5961,
+      "step": 29352
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.023682973647262e-06,
+      "loss": 0.6101,
+      "step": 29353
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.023064684167147e-06,
+      "loss": 0.6207,
+      "step": 29354
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.022446419976306e-06,
+      "loss": 0.6444,
+      "step": 29355
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.021828181077866e-06,
+      "loss": 0.5697,
+      "step": 29356
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.021209967474976e-06,
+      "loss": 0.5574,
+      "step": 29357
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.020591779170778e-06,
+      "loss": 0.5391,
+      "step": 29358
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.019973616168407e-06,
+      "loss": 0.5942,
+      "step": 29359
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.019355478471014e-06,
+      "loss": 0.6219,
+      "step": 29360
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.018737366081734e-06,
+      "loss": 0.5752,
+      "step": 29361
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.018119279003708e-06,
+      "loss": 0.6191,
+      "step": 29362
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.017501217240073e-06,
+      "loss": 0.6096,
+      "step": 29363
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.016883180793976e-06,
+      "loss": 0.5965,
+      "step": 29364
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.016265169668553e-06,
+      "loss": 0.6182,
+      "step": 29365
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.015647183866953e-06,
+      "loss": 0.5477,
+      "step": 29366
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0150292233923025e-06,
+      "loss": 0.6439,
+      "step": 29367
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.014411288247751e-06,
+      "loss": 0.6281,
+      "step": 29368
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.013793378436433e-06,
+      "loss": 0.6111,
+      "step": 29369
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0131754939614955e-06,
+      "loss": 0.6028,
+      "step": 29370
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.012557634826072e-06,
+      "loss": 0.6524,
+      "step": 29371
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0119398010333054e-06,
+      "loss": 0.6635,
+      "step": 29372
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.011321992586329e-06,
+      "loss": 0.6085,
+      "step": 29373
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0107042094882896e-06,
+      "loss": 0.5885,
+      "step": 29374
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.01008645174232e-06,
+      "loss": 0.6246,
+      "step": 29375
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.009468719351567e-06,
+      "loss": 0.611,
+      "step": 29376
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.008851012319165e-06,
+      "loss": 0.6009,
+      "step": 29377
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0082333306482514e-06,
+      "loss": 0.5594,
+      "step": 29378
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.0076156743419676e-06,
+      "loss": 0.5718,
+      "step": 29379
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.006998043403446e-06,
+      "loss": 0.6655,
+      "step": 29380
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.006380437835834e-06,
+      "loss": 0.6208,
+      "step": 29381
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.005762857642264e-06,
+      "loss": 0.5881,
+      "step": 29382
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.005145302825878e-06,
+      "loss": 0.6454,
+      "step": 29383
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.004527773389807e-06,
+      "loss": 0.5779,
+      "step": 29384
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.003910269337196e-06,
+      "loss": 0.5828,
+      "step": 29385
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.003292790671177e-06,
+      "loss": 0.5699,
+      "step": 29386
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.002675337394898e-06,
+      "loss": 0.5831,
+      "step": 29387
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.002057909511482e-06,
+      "loss": 0.6012,
+      "step": 29388
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.001440507024076e-06,
+      "loss": 0.6649,
+      "step": 29389
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.000823129935812e-06,
+      "loss": 0.6517,
+      "step": 29390
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 5.000205778249834e-06,
+      "loss": 0.6628,
+      "step": 29391
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.999588451969269e-06,
+      "loss": 0.6684,
+      "step": 29392
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.998971151097269e-06,
+      "loss": 0.5989,
+      "step": 29393
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9983538756369535e-06,
+      "loss": 0.6646,
+      "step": 29394
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.99773662559147e-06,
+      "loss": 0.5921,
+      "step": 29395
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.997119400963948e-06,
+      "loss": 0.5901,
+      "step": 29396
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.996502201757531e-06,
+      "loss": 0.5299,
+      "step": 29397
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.995885027975352e-06,
+      "loss": 0.5982,
+      "step": 29398
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.995267879620547e-06,
+      "loss": 0.6063,
+      "step": 29399
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.994650756696251e-06,
+      "loss": 0.5757,
+      "step": 29400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.994033659205597e-06,
+      "loss": 0.5728,
+      "step": 29401
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.993416587151727e-06,
+      "loss": 0.6117,
+      "step": 29402
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9927995405377705e-06,
+      "loss": 0.5939,
+      "step": 29403
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.992182519366874e-06,
+      "loss": 0.5887,
+      "step": 29404
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.991565523642157e-06,
+      "loss": 0.6472,
+      "step": 29405
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9909485533667655e-06,
+      "loss": 0.5994,
+      "step": 29406
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.990331608543828e-06,
+      "loss": 0.5809,
+      "step": 29407
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.989714689176486e-06,
+      "loss": 0.5473,
+      "step": 29408
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.989097795267871e-06,
+      "loss": 0.6469,
+      "step": 29409
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.988480926821119e-06,
+      "loss": 0.6098,
+      "step": 29410
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9878640838393575e-06,
+      "loss": 0.6086,
+      "step": 29411
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.987247266325731e-06,
+      "loss": 0.5867,
+      "step": 29412
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.986630474283366e-06,
+      "loss": 0.6057,
+      "step": 29413
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.986013707715407e-06,
+      "loss": 0.593,
+      "step": 29414
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.985396966624972e-06,
+      "loss": 0.5546,
+      "step": 29415
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.984780251015209e-06,
+      "loss": 0.5526,
+      "step": 29416
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.984163560889246e-06,
+      "loss": 0.5846,
+      "step": 29417
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.983546896250212e-06,
+      "loss": 0.5544,
+      "step": 29418
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.98293025710125e-06,
+      "loss": 0.6281,
+      "step": 29419
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.982313643445489e-06,
+      "loss": 0.6525,
+      "step": 29420
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.981697055286062e-06,
+      "loss": 0.6223,
+      "step": 29421
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.981080492626097e-06,
+      "loss": 0.6046,
+      "step": 29422
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.980463955468738e-06,
+      "loss": 0.555,
+      "step": 29423
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.979847443817105e-06,
+      "loss": 0.6124,
+      "step": 29424
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.979230957674346e-06,
+      "loss": 0.5888,
+      "step": 29425
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.978614497043578e-06,
+      "loss": 0.5813,
+      "step": 29426
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.977998061927943e-06,
+      "loss": 0.5878,
+      "step": 29427
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.977381652330567e-06,
+      "loss": 0.6128,
+      "step": 29428
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.976765268254591e-06,
+      "loss": 0.5608,
+      "step": 29429
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.976148909703136e-06,
+      "loss": 0.5613,
+      "step": 29430
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.97553257667935e-06,
+      "loss": 0.6014,
+      "step": 29431
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.974916269186344e-06,
+      "loss": 0.6064,
+      "step": 29432
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.974299987227266e-06,
+      "loss": 0.5896,
+      "step": 29433
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9736837308052365e-06,
+      "loss": 0.6087,
+      "step": 29434
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.973067499923396e-06,
+      "loss": 0.6446,
+      "step": 29435
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.972451294584871e-06,
+      "loss": 0.609,
+      "step": 29436
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9718351147927946e-06,
+      "loss": 0.4806,
+      "step": 29437
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.971218960550297e-06,
+      "loss": 0.585,
+      "step": 29438
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.970602831860503e-06,
+      "loss": 0.5948,
+      "step": 29439
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.969986728726554e-06,
+      "loss": 0.5933,
+      "step": 29440
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.969370651151576e-06,
+      "loss": 0.655,
+      "step": 29441
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.968754599138697e-06,
+      "loss": 0.6628,
+      "step": 29442
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.968138572691048e-06,
+      "loss": 0.5963,
+      "step": 29443
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.967522571811763e-06,
+      "loss": 0.6182,
+      "step": 29444
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9669065965039664e-06,
+      "loss": 0.5258,
+      "step": 29445
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.966290646770795e-06,
+      "loss": 0.6011,
+      "step": 29446
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.965674722615375e-06,
+      "loss": 0.6581,
+      "step": 29447
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9650588240408344e-06,
+      "loss": 0.6114,
+      "step": 29448
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9644429510503035e-06,
+      "loss": 0.5936,
+      "step": 29449
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.963827103646914e-06,
+      "loss": 0.6058,
+      "step": 29450
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.963211281833792e-06,
+      "loss": 0.6161,
+      "step": 29451
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.962595485614075e-06,
+      "loss": 0.5513,
+      "step": 29452
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.961979714990879e-06,
+      "loss": 0.5424,
+      "step": 29453
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.9613639699673435e-06,
+      "loss": 0.5928,
+      "step": 29454
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.960748250546589e-06,
+      "loss": 0.5664,
+      "step": 29455
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.960132556731752e-06,
+      "loss": 0.5826,
+      "step": 29456
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.959516888525957e-06,
+      "loss": 0.6739,
+      "step": 29457
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.958901245932334e-06,
+      "loss": 0.6332,
+      "step": 29458
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.958285628954009e-06,
+      "loss": 0.6361,
+      "step": 29459
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.957670037594108e-06,
+      "loss": 0.628,
+      "step": 29460
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.957054471855766e-06,
+      "loss": 0.5674,
+      "step": 29461
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.956438931742104e-06,
+      "loss": 0.578,
+      "step": 29462
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.955823417256259e-06,
+      "loss": 0.5359,
+      "step": 29463
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.955207928401346e-06,
+      "loss": 0.5754,
+      "step": 29464
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.954592465180502e-06,
+      "loss": 0.619,
+      "step": 29465
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.953977027596848e-06,
+      "loss": 0.6151,
+      "step": 29466
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.953361615653518e-06,
+      "loss": 0.6146,
+      "step": 29467
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.952746229353636e-06,
+      "loss": 0.597,
+      "step": 29468
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.952130868700329e-06,
+      "loss": 0.594,
+      "step": 29469
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.951515533696719e-06,
+      "loss": 0.609,
+      "step": 29470
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.95090022434594e-06,
+      "loss": 0.6035,
+      "step": 29471
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.950284940651113e-06,
+      "loss": 0.5915,
+      "step": 29472
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.949669682615371e-06,
+      "loss": 0.5768,
+      "step": 29473
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.949054450241836e-06,
+      "loss": 0.5829,
+      "step": 29474
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.948439243533635e-06,
+      "loss": 0.599,
+      "step": 29475
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9478240624938885e-06,
+      "loss": 0.6007,
+      "step": 29476
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.947208907125734e-06,
+      "loss": 0.609,
+      "step": 29477
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.946593777432288e-06,
+      "loss": 0.5852,
+      "step": 29478
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.945978673416681e-06,
+      "loss": 0.5805,
+      "step": 29479
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9453635950820354e-06,
+      "loss": 0.6162,
+      "step": 29480
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.944748542431473e-06,
+      "loss": 0.6627,
+      "step": 29481
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.944133515468129e-06,
+      "loss": 0.5367,
+      "step": 29482
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.943518514195119e-06,
+      "loss": 0.5963,
+      "step": 29483
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.942903538615577e-06,
+      "loss": 0.548,
+      "step": 29484
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.942288588732622e-06,
+      "loss": 0.5771,
+      "step": 29485
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.94167366454938e-06,
+      "loss": 0.6142,
+      "step": 29486
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.941058766068973e-06,
+      "loss": 0.6075,
+      "step": 29487
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.940443893294531e-06,
+      "loss": 0.6785,
+      "step": 29488
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.939829046229171e-06,
+      "loss": 0.584,
+      "step": 29489
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.939214224876029e-06,
+      "loss": 0.5816,
+      "step": 29490
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9385994292382144e-06,
+      "loss": 0.6147,
+      "step": 29491
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.937984659318863e-06,
+      "loss": 0.6442,
+      "step": 29492
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.93736991512109e-06,
+      "loss": 0.5731,
+      "step": 29493
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9367551966480255e-06,
+      "loss": 0.5458,
+      "step": 29494
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.936140503902792e-06,
+      "loss": 0.6305,
+      "step": 29495
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.935525836888512e-06,
+      "loss": 0.6383,
+      "step": 29496
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.934911195608303e-06,
+      "loss": 0.5984,
+      "step": 29497
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.934296580065299e-06,
+      "loss": 0.5981,
+      "step": 29498
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.933681990262617e-06,
+      "loss": 0.6097,
+      "step": 29499
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9330674262033765e-06,
+      "loss": 0.6463,
+      "step": 29500
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9324528878907116e-06,
+      "loss": 0.6114,
+      "step": 29501
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.931838375327731e-06,
+      "loss": 0.6062,
+      "step": 29502
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9312238885175665e-06,
+      "loss": 0.5938,
+      "step": 29503
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.930609427463336e-06,
+      "loss": 0.5278,
+      "step": 29504
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9299949921681655e-06,
+      "loss": 0.5934,
+      "step": 29505
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.929380582635176e-06,
+      "loss": 0.527,
+      "step": 29506
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.928766198867489e-06,
+      "loss": 0.5899,
+      "step": 29507
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.928151840868223e-06,
+      "loss": 0.5609,
+      "step": 29508
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.927537508640505e-06,
+      "loss": 0.5992,
+      "step": 29509
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.926923202187451e-06,
+      "loss": 0.6495,
+      "step": 29510
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.926308921512191e-06,
+      "loss": 0.5802,
+      "step": 29511
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.92569466661784e-06,
+      "loss": 0.5941,
+      "step": 29512
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9250804375075215e-06,
+      "loss": 0.6316,
+      "step": 29513
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.924466234184351e-06,
+      "loss": 0.5806,
+      "step": 29514
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.923852056651458e-06,
+      "loss": 0.5791,
+      "step": 29515
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.92323790491196e-06,
+      "loss": 0.5626,
+      "step": 29516
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.922623778968976e-06,
+      "loss": 0.6218,
+      "step": 29517
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.922009678825628e-06,
+      "loss": 0.6246,
+      "step": 29518
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.921395604485032e-06,
+      "loss": 0.6389,
+      "step": 29519
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9207815559503156e-06,
+      "loss": 0.5759,
+      "step": 29520
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.920167533224592e-06,
+      "loss": 0.5663,
+      "step": 29521
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.919553536310989e-06,
+      "loss": 0.6437,
+      "step": 29522
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.918939565212623e-06,
+      "loss": 0.6149,
+      "step": 29523
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.918325619932613e-06,
+      "loss": 0.6068,
+      "step": 29524
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.917711700474073e-06,
+      "loss": 0.5874,
+      "step": 29525
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9170978068401345e-06,
+      "loss": 0.6106,
+      "step": 29526
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.916483939033906e-06,
+      "loss": 0.6393,
+      "step": 29527
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.915870097058518e-06,
+      "loss": 0.5984,
+      "step": 29528
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.915256280917077e-06,
+      "loss": 0.5853,
+      "step": 29529
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.914642490612711e-06,
+      "loss": 0.5787,
+      "step": 29530
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9140287261485316e-06,
+      "loss": 0.5826,
+      "step": 29531
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.913414987527666e-06,
+      "loss": 0.6051,
+      "step": 29532
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.912801274753228e-06,
+      "loss": 0.581,
+      "step": 29533
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.912187587828337e-06,
+      "loss": 0.5824,
+      "step": 29534
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.911573926756108e-06,
+      "loss": 0.5895,
+      "step": 29535
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.910960291539665e-06,
+      "loss": 0.6573,
+      "step": 29536
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.910346682182123e-06,
+      "loss": 0.6493,
+      "step": 29537
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.909733098686597e-06,
+      "loss": 0.6269,
+      "step": 29538
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.909119541056216e-06,
+      "loss": 0.6969,
+      "step": 29539
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9085060092940794e-06,
+      "loss": 0.604,
+      "step": 29540
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.907892503403321e-06,
+      "loss": 0.6117,
+      "step": 29541
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.907279023387047e-06,
+      "loss": 0.5892,
+      "step": 29542
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.906665569248385e-06,
+      "loss": 0.6049,
+      "step": 29543
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.906052140990445e-06,
+      "loss": 0.5876,
+      "step": 29544
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.905438738616347e-06,
+      "loss": 0.5744,
+      "step": 29545
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.904825362129203e-06,
+      "loss": 0.6218,
+      "step": 29546
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.904212011532137e-06,
+      "loss": 0.641,
+      "step": 29547
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.903598686828259e-06,
+      "loss": 0.5545,
+      "step": 29548
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.902985388020692e-06,
+      "loss": 0.5491,
+      "step": 29549
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.902372115112549e-06,
+      "loss": 0.5741,
+      "step": 29550
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.901758868106946e-06,
+      "loss": 0.5999,
+      "step": 29551
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.901145647006995e-06,
+      "loss": 0.6348,
+      "step": 29552
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.9005324518158215e-06,
+      "loss": 0.5654,
+      "step": 29553
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.899919282536535e-06,
+      "loss": 0.5553,
+      "step": 29554
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.899306139172252e-06,
+      "loss": 0.6151,
+      "step": 29555
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.898693021726085e-06,
+      "loss": 0.6257,
+      "step": 29556
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.898079930201156e-06,
+      "loss": 0.6106,
+      "step": 29557
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.897466864600579e-06,
+      "loss": 0.5912,
+      "step": 29558
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.896853824927461e-06,
+      "loss": 0.6134,
+      "step": 29559
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.896240811184932e-06,
+      "loss": 0.6421,
+      "step": 29560
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.89562782337609e-06,
+      "loss": 0.5584,
+      "step": 29561
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.895014861504063e-06,
+      "loss": 0.6041,
+      "step": 29562
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.894401925571956e-06,
+      "loss": 0.5611,
+      "step": 29563
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.89378901558289e-06,
+      "loss": 0.5516,
+      "step": 29564
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.893176131539976e-06,
+      "loss": 0.6033,
+      "step": 29565
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.892563273446337e-06,
+      "loss": 0.6703,
+      "step": 29566
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.891950441305071e-06,
+      "loss": 0.6231,
+      "step": 29567
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.891337635119306e-06,
+      "loss": 0.5997,
+      "step": 29568
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.890724854892146e-06,
+      "loss": 0.6187,
+      "step": 29569
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.890112100626712e-06,
+      "loss": 0.6487,
+      "step": 29570
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.889499372326117e-06,
+      "loss": 0.6163,
+      "step": 29571
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.888886669993472e-06,
+      "loss": 0.5634,
+      "step": 29572
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.888273993631886e-06,
+      "loss": 0.5617,
+      "step": 29573
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.887661343244481e-06,
+      "loss": 0.6236,
+      "step": 29574
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.887048718834366e-06,
+      "loss": 0.5525,
+      "step": 29575
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.886436120404651e-06,
+      "loss": 0.6043,
+      "step": 29576
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.885823547958454e-06,
+      "loss": 0.5995,
+      "step": 29577
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.885211001498886e-06,
+      "loss": 0.587,
+      "step": 29578
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.884598481029059e-06,
+      "loss": 0.6147,
+      "step": 29579
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.883985986552081e-06,
+      "loss": 0.6439,
+      "step": 29580
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.8833735180710716e-06,
+      "loss": 0.6015,
+      "step": 29581
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.882761075589141e-06,
+      "loss": 0.5897,
+      "step": 29582
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.8821486591094e-06,
+      "loss": 0.6016,
+      "step": 29583
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.881536268634957e-06,
+      "loss": 0.6306,
+      "step": 29584
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.880923904168929e-06,
+      "loss": 0.5876,
+      "step": 29585
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.880311565714424e-06,
+      "loss": 0.6001,
+      "step": 29586
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.879699253274559e-06,
+      "loss": 0.6453,
+      "step": 29587
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.879086966852441e-06,
+      "loss": 0.5606,
+      "step": 29588
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.878474706451182e-06,
+      "loss": 0.6151,
+      "step": 29589
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.87786247207389e-06,
+      "loss": 0.5967,
+      "step": 29590
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.877250263723681e-06,
+      "loss": 0.5495,
+      "step": 29591
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.876638081403665e-06,
+      "loss": 0.559,
+      "step": 29592
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.876025925116951e-06,
+      "loss": 0.6071,
+      "step": 29593
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.875413794866646e-06,
+      "loss": 0.6138,
+      "step": 29594
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.874801690655868e-06,
+      "loss": 0.5318,
+      "step": 29595
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.874189612487723e-06,
+      "loss": 0.6104,
+      "step": 29596
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.873577560365319e-06,
+      "loss": 0.6007,
+      "step": 29597
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.872965534291776e-06,
+      "loss": 0.6095,
+      "step": 29598
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.872353534270189e-06,
+      "loss": 0.537,
+      "step": 29599
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.8717415603036785e-06,
+      "loss": 0.5826,
+      "step": 29600
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.8711296123953485e-06,
+      "loss": 0.5912,
+      "step": 29601
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.870517690548314e-06,
+      "loss": 0.6394,
+      "step": 29602
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.869905794765678e-06,
+      "loss": 0.6222,
+      "step": 29603
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.869293925050561e-06,
+      "loss": 0.6528,
+      "step": 29604
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.868682081406055e-06,
+      "loss": 0.647,
+      "step": 29605
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.868070263835284e-06,
+      "loss": 0.6421,
+      "step": 29606
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.8674584723413455e-06,
+      "loss": 0.6879,
+      "step": 29607
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.866846706927358e-06,
+      "loss": 0.6381,
+      "step": 29608
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.866234967596426e-06,
+      "loss": 0.6095,
+      "step": 29609
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.865623254351657e-06,
+      "loss": 0.5754,
+      "step": 29610
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.865011567196157e-06,
+      "loss": 0.5906,
+      "step": 29611
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.864399906133041e-06,
+      "loss": 0.5895,
+      "step": 29612
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.863788271165411e-06,
+      "loss": 0.5504,
+      "step": 29613
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.863176662296375e-06,
+      "loss": 0.5635,
+      "step": 29614
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8625650795290465e-06,
+      "loss": 0.6632,
+      "step": 29615
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.861953522866529e-06,
+      "loss": 0.6397,
+      "step": 29616
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.861341992311931e-06,
+      "loss": 0.6122,
+      "step": 29617
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8607304878683545e-06,
+      "loss": 0.6206,
+      "step": 29618
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.860119009538916e-06,
+      "loss": 0.6194,
+      "step": 29619
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.859507557326718e-06,
+      "loss": 0.5821,
+      "step": 29620
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.858896131234867e-06,
+      "loss": 0.6007,
+      "step": 29621
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.858284731266467e-06,
+      "loss": 0.6112,
+      "step": 29622
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.857673357424633e-06,
+      "loss": 0.6055,
+      "step": 29623
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.857062009712461e-06,
+      "loss": 0.5977,
+      "step": 29624
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.856450688133072e-06,
+      "loss": 0.5945,
+      "step": 29625
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.855839392689554e-06,
+      "loss": 0.5581,
+      "step": 29626
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.855228123385028e-06,
+      "loss": 0.6171,
+      "step": 29627
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.854616880222591e-06,
+      "loss": 0.6134,
+      "step": 29628
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.854005663205356e-06,
+      "loss": 0.5705,
+      "step": 29629
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.853394472336425e-06,
+      "loss": 0.6691,
+      "step": 29630
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.852783307618905e-06,
+      "loss": 0.5754,
+      "step": 29631
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.852172169055894e-06,
+      "loss": 0.6504,
+      "step": 29632
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.85156105665051e-06,
+      "loss": 0.6407,
+      "step": 29633
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.850949970405852e-06,
+      "loss": 0.5765,
+      "step": 29634
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.850338910325021e-06,
+      "loss": 0.5656,
+      "step": 29635
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.849727876411129e-06,
+      "loss": 0.5769,
+      "step": 29636
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.849116868667279e-06,
+      "loss": 0.5886,
+      "step": 29637
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.848505887096574e-06,
+      "loss": 0.6018,
+      "step": 29638
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.847894931702116e-06,
+      "loss": 0.5791,
+      "step": 29639
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.847284002487017e-06,
+      "loss": 0.5871,
+      "step": 29640
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.846673099454372e-06,
+      "loss": 0.5928,
+      "step": 29641
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8460622226072985e-06,
+      "loss": 0.6176,
+      "step": 29642
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.845451371948885e-06,
+      "loss": 0.6271,
+      "step": 29643
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.844840547482246e-06,
+      "loss": 0.5588,
+      "step": 29644
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.844229749210479e-06,
+      "loss": 0.5905,
+      "step": 29645
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.843618977136693e-06,
+      "loss": 0.6313,
+      "step": 29646
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.84300823126399e-06,
+      "loss": 0.5569,
+      "step": 29647
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.842397511595471e-06,
+      "loss": 0.6258,
+      "step": 29648
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.841786818134239e-06,
+      "loss": 0.614,
+      "step": 29649
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.841176150883401e-06,
+      "loss": 0.588,
+      "step": 29650
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.840565509846059e-06,
+      "loss": 0.5865,
+      "step": 29651
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.839954895025314e-06,
+      "loss": 0.6575,
+      "step": 29652
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.839344306424265e-06,
+      "loss": 0.6025,
+      "step": 29653
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.838733744046024e-06,
+      "loss": 0.546,
+      "step": 29654
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.838123207893688e-06,
+      "loss": 0.6085,
+      "step": 29655
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8375126979703555e-06,
+      "loss": 0.5337,
+      "step": 29656
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.836902214279136e-06,
+      "loss": 0.5823,
+      "step": 29657
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8362917568231304e-06,
+      "loss": 0.6309,
+      "step": 29658
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.835681325605438e-06,
+      "loss": 0.6101,
+      "step": 29659
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.835070920629157e-06,
+      "loss": 0.6181,
+      "step": 29660
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.834460541897397e-06,
+      "loss": 0.67,
+      "step": 29661
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.833850189413251e-06,
+      "loss": 0.6192,
+      "step": 29662
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.833239863179835e-06,
+      "loss": 0.5702,
+      "step": 29663
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.832629563200232e-06,
+      "loss": 0.6239,
+      "step": 29664
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.832019289477554e-06,
+      "loss": 0.6028,
+      "step": 29665
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.831409042014897e-06,
+      "loss": 0.6075,
+      "step": 29666
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.830798820815367e-06,
+      "loss": 0.6166,
+      "step": 29667
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.83018862588206e-06,
+      "loss": 0.6414,
+      "step": 29668
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.829578457218086e-06,
+      "loss": 0.66,
+      "step": 29669
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.828968314826529e-06,
+      "loss": 0.6076,
+      "step": 29670
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8283581987105035e-06,
+      "loss": 0.5894,
+      "step": 29671
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.827748108873104e-06,
+      "loss": 0.6769,
+      "step": 29672
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.827138045317428e-06,
+      "loss": 0.6458,
+      "step": 29673
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.826528008046581e-06,
+      "loss": 0.5725,
+      "step": 29674
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.825917997063662e-06,
+      "loss": 0.5663,
+      "step": 29675
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.825308012371768e-06,
+      "loss": 0.5856,
+      "step": 29676
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.824698053973995e-06,
+      "loss": 0.6054,
+      "step": 29677
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.824088121873452e-06,
+      "loss": 0.6052,
+      "step": 29678
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.823478216073229e-06,
+      "loss": 0.6446,
+      "step": 29679
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.822868336576435e-06,
+      "loss": 0.6068,
+      "step": 29680
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.822258483386157e-06,
+      "loss": 0.6194,
+      "step": 29681
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.821648656505504e-06,
+      "loss": 0.6786,
+      "step": 29682
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.821038855937567e-06,
+      "loss": 0.5743,
+      "step": 29683
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.820429081685452e-06,
+      "loss": 0.5533,
+      "step": 29684
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.819819333752254e-06,
+      "loss": 0.6402,
+      "step": 29685
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.81920961214107e-06,
+      "loss": 0.5955,
+      "step": 29686
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.818599916854997e-06,
+      "loss": 0.5816,
+      "step": 29687
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.817990247897138e-06,
+      "loss": 0.5814,
+      "step": 29688
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.817380605270584e-06,
+      "loss": 0.6326,
+      "step": 29689
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8167709889784454e-06,
+      "loss": 0.5813,
+      "step": 29690
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.816161399023803e-06,
+      "loss": 0.5693,
+      "step": 29691
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.815551835409767e-06,
+      "loss": 0.6605,
+      "step": 29692
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.814942298139431e-06,
+      "loss": 0.59,
+      "step": 29693
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.814332787215888e-06,
+      "loss": 0.5866,
+      "step": 29694
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.813723302642242e-06,
+      "loss": 0.5942,
+      "step": 29695
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.813113844421586e-06,
+      "loss": 0.5847,
+      "step": 29696
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.812504412557018e-06,
+      "loss": 0.649,
+      "step": 29697
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.81189500705163e-06,
+      "loss": 0.5777,
+      "step": 29698
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.811285627908529e-06,
+      "loss": 0.633,
+      "step": 29699
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.810676275130799e-06,
+      "loss": 0.6188,
+      "step": 29700
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.810066948721551e-06,
+      "loss": 0.6046,
+      "step": 29701
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.809457648683865e-06,
+      "loss": 0.5937,
+      "step": 29702
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.808848375020849e-06,
+      "loss": 0.5595,
+      "step": 29703
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8082391277355915e-06,
+      "loss": 0.5959,
+      "step": 29704
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.8076299068311945e-06,
+      "loss": 0.6405,
+      "step": 29705
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.807020712310747e-06,
+      "loss": 0.6175,
+      "step": 29706
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.806411544177357e-06,
+      "loss": 0.6191,
+      "step": 29707
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.805802402434101e-06,
+      "loss": 0.5624,
+      "step": 29708
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.805193287084089e-06,
+      "loss": 0.6674,
+      "step": 29709
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.804584198130411e-06,
+      "loss": 0.6339,
+      "step": 29710
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.803975135576159e-06,
+      "loss": 0.5681,
+      "step": 29711
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.803366099424435e-06,
+      "loss": 0.5969,
+      "step": 29712
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.802757089678328e-06,
+      "loss": 0.5676,
+      "step": 29713
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.802148106340936e-06,
+      "loss": 0.6197,
+      "step": 29714
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.801539149415347e-06,
+      "loss": 0.5958,
+      "step": 29715
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.800930218904664e-06,
+      "loss": 0.6169,
+      "step": 29716
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.800321314811979e-06,
+      "loss": 0.678,
+      "step": 29717
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.799712437140382e-06,
+      "loss": 0.648,
+      "step": 29718
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.799103585892967e-06,
+      "loss": 0.6227,
+      "step": 29719
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.798494761072833e-06,
+      "loss": 0.5632,
+      "step": 29720
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.797885962683065e-06,
+      "loss": 0.5918,
+      "step": 29721
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.797277190726768e-06,
+      "loss": 0.5991,
+      "step": 29722
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7966684452070275e-06,
+      "loss": 0.563,
+      "step": 29723
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.79605972612694e-06,
+      "loss": 0.6132,
+      "step": 29724
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7954510334895925e-06,
+      "loss": 0.5797,
+      "step": 29725
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.794842367298086e-06,
+      "loss": 0.6618,
+      "step": 29726
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.794233727555507e-06,
+      "loss": 0.6263,
+      "step": 29727
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.793625114264958e-06,
+      "loss": 0.5333,
+      "step": 29728
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.793016527429517e-06,
+      "loss": 0.6284,
+      "step": 29729
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.792407967052288e-06,
+      "loss": 0.5478,
+      "step": 29730
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.791799433136359e-06,
+      "loss": 0.6111,
+      "step": 29731
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7911909256848185e-06,
+      "loss": 0.5836,
+      "step": 29732
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.790582444700766e-06,
+      "loss": 0.632,
+      "step": 29733
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.78997399018729e-06,
+      "loss": 0.5947,
+      "step": 29734
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7893655621474825e-06,
+      "loss": 0.6051,
+      "step": 29735
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.788757160584429e-06,
+      "loss": 0.6269,
+      "step": 29736
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.788148785501232e-06,
+      "loss": 0.6537,
+      "step": 29737
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.787540436900974e-06,
+      "loss": 0.5841,
+      "step": 29738
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7869321147867565e-06,
+      "loss": 0.5749,
+      "step": 29739
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.786323819161656e-06,
+      "loss": 0.571,
+      "step": 29740
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.785715550028774e-06,
+      "loss": 0.6556,
+      "step": 29741
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.785107307391196e-06,
+      "loss": 0.5961,
+      "step": 29742
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7844990912520184e-06,
+      "loss": 0.6395,
+      "step": 29743
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7838909016143276e-06,
+      "loss": 0.5942,
+      "step": 29744
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.783282738481215e-06,
+      "loss": 0.5587,
+      "step": 29745
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.782674601855768e-06,
+      "loss": 0.6576,
+      "step": 29746
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.782066491741082e-06,
+      "loss": 0.6183,
+      "step": 29747
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.7814584081402405e-06,
+      "loss": 0.6248,
+      "step": 29748
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.780850351056341e-06,
+      "loss": 0.5874,
+      "step": 29749
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.78024232049247e-06,
+      "loss": 0.5997,
+      "step": 29750
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.779634316451715e-06,
+      "loss": 0.5712,
+      "step": 29751
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.779026338937168e-06,
+      "loss": 0.6242,
+      "step": 29752
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.778418387951913e-06,
+      "loss": 0.6429,
+      "step": 29753
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.777810463499049e-06,
+      "loss": 0.6278,
+      "step": 29754
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.777202565581656e-06,
+      "loss": 0.6174,
+      "step": 29755
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7765946942028286e-06,
+      "loss": 0.6145,
+      "step": 29756
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7759868493656494e-06,
+      "loss": 0.605,
+      "step": 29757
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.775379031073215e-06,
+      "loss": 0.5634,
+      "step": 29758
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.774771239328605e-06,
+      "loss": 0.6012,
+      "step": 29759
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.774163474134917e-06,
+      "loss": 0.6016,
+      "step": 29760
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.773555735495234e-06,
+      "loss": 0.587,
+      "step": 29761
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.772948023412647e-06,
+      "loss": 0.5913,
+      "step": 29762
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.772340337890237e-06,
+      "loss": 0.6146,
+      "step": 29763
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.771732678931099e-06,
+      "loss": 0.6112,
+      "step": 29764
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.771125046538316e-06,
+      "loss": 0.5784,
+      "step": 29765
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.770517440714986e-06,
+      "loss": 0.6155,
+      "step": 29766
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.76990986146418e-06,
+      "loss": 0.6067,
+      "step": 29767
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.769302308788998e-06,
+      "loss": 0.5891,
+      "step": 29768
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7686947826925186e-06,
+      "loss": 0.6018,
+      "step": 29769
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.768087283177838e-06,
+      "loss": 0.5955,
+      "step": 29770
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7674798102480366e-06,
+      "loss": 0.5596,
+      "step": 29771
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.766872363906203e-06,
+      "loss": 0.6036,
+      "step": 29772
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.766264944155424e-06,
+      "loss": 0.5426,
+      "step": 29773
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.765657550998783e-06,
+      "loss": 0.6094,
+      "step": 29774
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7650501844393704e-06,
+      "loss": 0.6332,
+      "step": 29775
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.764442844480268e-06,
+      "loss": 0.5929,
+      "step": 29776
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.763835531124573e-06,
+      "loss": 0.584,
+      "step": 29777
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.763228244375355e-06,
+      "loss": 0.5558,
+      "step": 29778
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.762620984235712e-06,
+      "loss": 0.638,
+      "step": 29779
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.762013750708721e-06,
+      "loss": 0.5835,
+      "step": 29780
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.761406543797478e-06,
+      "loss": 0.5774,
+      "step": 29781
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.760799363505061e-06,
+      "loss": 0.5655,
+      "step": 29782
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.760192209834557e-06,
+      "loss": 0.6555,
+      "step": 29783
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.759585082789047e-06,
+      "loss": 0.5951,
+      "step": 29784
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7589779823716235e-06,
+      "loss": 0.6172,
+      "step": 29785
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.758370908585365e-06,
+      "loss": 0.5634,
+      "step": 29786
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.757763861433361e-06,
+      "loss": 0.5961,
+      "step": 29787
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.757156840918695e-06,
+      "loss": 0.6287,
+      "step": 29788
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.756549847044452e-06,
+      "loss": 0.6254,
+      "step": 29789
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.755942879813713e-06,
+      "loss": 0.6036,
+      "step": 29790
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.75533593922956e-06,
+      "loss": 0.5934,
+      "step": 29791
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.754729025295086e-06,
+      "loss": 0.5973,
+      "step": 29792
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.754122138013369e-06,
+      "loss": 0.5792,
+      "step": 29793
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.753515277387496e-06,
+      "loss": 0.5683,
+      "step": 29794
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.752908443420542e-06,
+      "loss": 0.62,
+      "step": 29795
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.752301636115601e-06,
+      "loss": 0.654,
+      "step": 29796
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.75169485547575e-06,
+      "loss": 0.5965,
+      "step": 29797
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.751088101504078e-06,
+      "loss": 0.6322,
+      "step": 29798
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.750481374203664e-06,
+      "loss": 0.5462,
+      "step": 29799
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.749874673577592e-06,
+      "loss": 0.63,
+      "step": 29800
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.749267999628939e-06,
+      "loss": 0.5521,
+      "step": 29801
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7486613523607974e-06,
+      "loss": 0.6484,
+      "step": 29802
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.748054731776243e-06,
+      "loss": 0.5718,
+      "step": 29803
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7474481378783665e-06,
+      "loss": 0.5995,
+      "step": 29804
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7468415706702375e-06,
+      "loss": 0.6037,
+      "step": 29805
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.746235030154948e-06,
+      "loss": 0.6727,
+      "step": 29806
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.745628516335575e-06,
+      "loss": 0.6454,
+      "step": 29807
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.745022029215205e-06,
+      "loss": 0.5682,
+      "step": 29808
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.744415568796916e-06,
+      "loss": 0.5733,
+      "step": 29809
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.743809135083792e-06,
+      "loss": 0.5828,
+      "step": 29810
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7432027280789125e-06,
+      "loss": 0.586,
+      "step": 29811
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.742596347785357e-06,
+      "loss": 0.6112,
+      "step": 29812
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.741989994206212e-06,
+      "loss": 0.6317,
+      "step": 29813
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.741383667344551e-06,
+      "loss": 0.5871,
+      "step": 29814
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7407773672034685e-06,
+      "loss": 0.6236,
+      "step": 29815
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.740171093786028e-06,
+      "loss": 0.583,
+      "step": 29816
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.739564847095324e-06,
+      "loss": 0.6145,
+      "step": 29817
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.738958627134426e-06,
+      "loss": 0.5939,
+      "step": 29818
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.738352433906424e-06,
+      "loss": 0.6268,
+      "step": 29819
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7377462674143946e-06,
+      "loss": 0.6353,
+      "step": 29820
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.737140127661418e-06,
+      "loss": 0.6673,
+      "step": 29821
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.736534014650569e-06,
+      "loss": 0.5818,
+      "step": 29822
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.735927928384937e-06,
+      "loss": 0.6465,
+      "step": 29823
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7353218688675925e-06,
+      "loss": 0.6027,
+      "step": 29824
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.734715836101624e-06,
+      "loss": 0.6083,
+      "step": 29825
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.734109830090104e-06,
+      "loss": 0.6055,
+      "step": 29826
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.733503850836116e-06,
+      "loss": 0.6073,
+      "step": 29827
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.732897898342733e-06,
+      "loss": 0.6078,
+      "step": 29828
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.732291972613042e-06,
+      "loss": 0.6038,
+      "step": 29829
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7316860736501195e-06,
+      "loss": 0.6367,
+      "step": 29830
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.731080201457041e-06,
+      "loss": 0.5596,
+      "step": 29831
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.730474356036888e-06,
+      "loss": 0.5785,
+      "step": 29832
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.729868537392735e-06,
+      "loss": 0.6348,
+      "step": 29833
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.729262745527666e-06,
+      "loss": 0.6303,
+      "step": 29834
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.728656980444754e-06,
+      "loss": 0.5892,
+      "step": 29835
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.728051242147086e-06,
+      "loss": 0.5831,
+      "step": 29836
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7274455306377265e-06,
+      "loss": 0.6157,
+      "step": 29837
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.726839845919765e-06,
+      "loss": 0.6024,
+      "step": 29838
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.726234187996269e-06,
+      "loss": 0.6291,
+      "step": 29839
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.725628556870327e-06,
+      "loss": 0.557,
+      "step": 29840
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.725022952545006e-06,
+      "loss": 0.6109,
+      "step": 29841
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.724417375023397e-06,
+      "loss": 0.5652,
+      "step": 29842
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.72381182430856e-06,
+      "loss": 0.6146,
+      "step": 29843
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7232063004035844e-06,
+      "loss": 0.5817,
+      "step": 29844
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.722600803311539e-06,
+      "loss": 0.6013,
+      "step": 29845
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.721995333035508e-06,
+      "loss": 0.586,
+      "step": 29846
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7213898895785645e-06,
+      "loss": 0.6162,
+      "step": 29847
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7207844729437855e-06,
+      "loss": 0.6417,
+      "step": 29848
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7201790831342435e-06,
+      "loss": 0.5475,
+      "step": 29849
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.71957372015302e-06,
+      "loss": 0.5607,
+      "step": 29850
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.71896838400319e-06,
+      "loss": 0.5673,
+      "step": 29851
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.718363074687823e-06,
+      "loss": 0.5808,
+      "step": 29852
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.717757792210009e-06,
+      "loss": 0.578,
+      "step": 29853
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.717152536572807e-06,
+      "loss": 0.6326,
+      "step": 29854
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.716547307779304e-06,
+      "loss": 0.5983,
+      "step": 29855
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.715942105832566e-06,
+      "loss": 0.6175,
+      "step": 29856
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.715336930735679e-06,
+      "loss": 0.5857,
+      "step": 29857
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.714731782491713e-06,
+      "loss": 0.6028,
+      "step": 29858
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7141266611037416e-06,
+      "loss": 0.6011,
+      "step": 29859
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.713521566574838e-06,
+      "loss": 0.58,
+      "step": 29860
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7129164989080826e-06,
+      "loss": 0.5995,
+      "step": 29861
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.712311458106544e-06,
+      "loss": 0.6177,
+      "step": 29862
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.711706444173301e-06,
+      "loss": 0.6155,
+      "step": 29863
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.711101457111427e-06,
+      "loss": 0.5747,
+      "step": 29864
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.710496496923998e-06,
+      "loss": 0.5837,
+      "step": 29865
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.709891563614078e-06,
+      "loss": 0.5387,
+      "step": 29866
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.709286657184755e-06,
+      "loss": 0.5604,
+      "step": 29867
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.708681777639095e-06,
+      "loss": 0.5973,
+      "step": 29868
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.708076924980172e-06,
+      "loss": 0.5568,
+      "step": 29869
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.707472099211056e-06,
+      "loss": 0.5424,
+      "step": 29870
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.706867300334829e-06,
+      "loss": 0.5918,
+      "step": 29871
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.7062625283545584e-06,
+      "loss": 0.6467,
+      "step": 29872
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.705657783273315e-06,
+      "loss": 0.5857,
+      "step": 29873
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.705053065094183e-06,
+      "loss": 0.5878,
+      "step": 29874
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.704448373820219e-06,
+      "loss": 0.682,
+      "step": 29875
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.703843709454507e-06,
+      "loss": 0.615,
+      "step": 29876
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.703239072000113e-06,
+      "loss": 0.6773,
+      "step": 29877
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.702634461460116e-06,
+      "loss": 0.6421,
+      "step": 29878
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.702029877837581e-06,
+      "loss": 0.5688,
+      "step": 29879
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.701425321135591e-06,
+      "loss": 0.6244,
+      "step": 29880
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.700820791357204e-06,
+      "loss": 0.5704,
+      "step": 29881
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.700216288505502e-06,
+      "loss": 0.6048,
+      "step": 29882
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.699611812583549e-06,
+      "loss": 0.577,
+      "step": 29883
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.699007363594425e-06,
+      "loss": 0.6042,
+      "step": 29884
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.698402941541196e-06,
+      "loss": 0.6107,
+      "step": 29885
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.697798546426935e-06,
+      "loss": 0.6192,
+      "step": 29886
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.697194178254708e-06,
+      "loss": 0.5971,
+      "step": 29887
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.696589837027595e-06,
+      "loss": 0.6154,
+      "step": 29888
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.695985522748661e-06,
+      "loss": 0.5608,
+      "step": 29889
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.695381235420976e-06,
+      "loss": 0.6361,
+      "step": 29890
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.694776975047619e-06,
+      "loss": 0.5761,
+      "step": 29891
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.694172741631646e-06,
+      "loss": 0.5847,
+      "step": 29892
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.69356853517614e-06,
+      "loss": 0.6641,
+      "step": 29893
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.692964355684162e-06,
+      "loss": 0.6147,
+      "step": 29894
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.69236020315879e-06,
+      "loss": 0.5679,
+      "step": 29895
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.69175607760309e-06,
+      "loss": 0.5942,
+      "step": 29896
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.691151979020134e-06,
+      "loss": 0.6303,
+      "step": 29897
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.690547907412982e-06,
+      "loss": 0.5959,
+      "step": 29898
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.6899438627847174e-06,
+      "loss": 0.5449,
+      "step": 29899
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.689339845138399e-06,
+      "loss": 0.5982,
+      "step": 29900
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.688735854477108e-06,
+      "loss": 0.553,
+      "step": 29901
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.688131890803897e-06,
+      "loss": 0.5322,
+      "step": 29902
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.687527954121847e-06,
+      "loss": 0.6158,
+      "step": 29903
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.686924044434022e-06,
+      "loss": 0.5923,
+      "step": 29904
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.686320161743494e-06,
+      "loss": 0.5434,
+      "step": 29905
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6857163060533296e-06,
+      "loss": 0.6173,
+      "step": 29906
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.685112477366596e-06,
+      "loss": 0.5835,
+      "step": 29907
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.68450867568636e-06,
+      "loss": 0.6132,
+      "step": 29908
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6839049010156956e-06,
+      "loss": 0.5487,
+      "step": 29909
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.683301153357666e-06,
+      "loss": 0.5417,
+      "step": 29910
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.682697432715339e-06,
+      "loss": 0.6045,
+      "step": 29911
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6820937390917894e-06,
+      "loss": 0.5455,
+      "step": 29912
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6814900724900725e-06,
+      "loss": 0.5737,
+      "step": 29913
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.680886432913265e-06,
+      "loss": 0.5688,
+      "step": 29914
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.680282820364428e-06,
+      "loss": 0.6017,
+      "step": 29915
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.679679234846636e-06,
+      "loss": 0.5974,
+      "step": 29916
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.679075676362948e-06,
+      "loss": 0.5703,
+      "step": 29917
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.678472144916441e-06,
+      "loss": 0.5941,
+      "step": 29918
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.677868640510169e-06,
+      "loss": 0.6034,
+      "step": 29919
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6772651631472084e-06,
+      "loss": 0.6364,
+      "step": 29920
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6766617128306205e-06,
+      "loss": 0.6043,
+      "step": 29921
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.676058289563476e-06,
+      "loss": 0.6038,
+      "step": 29922
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.675454893348837e-06,
+      "loss": 0.5988,
+      "step": 29923
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.674851524189773e-06,
+      "loss": 0.61,
+      "step": 29924
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.674248182089342e-06,
+      "loss": 0.6639,
+      "step": 29925
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.673644867050622e-06,
+      "loss": 0.6875,
+      "step": 29926
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.673041579076671e-06,
+      "loss": 0.6126,
+      "step": 29927
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.672438318170557e-06,
+      "loss": 0.6308,
+      "step": 29928
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.671835084335339e-06,
+      "loss": 0.5973,
+      "step": 29929
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.671231877574092e-06,
+      "loss": 0.6059,
+      "step": 29930
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.670628697889876e-06,
+      "loss": 0.5833,
+      "step": 29931
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6700255452857525e-06,
+      "loss": 0.5644,
+      "step": 29932
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.669422419764794e-06,
+      "loss": 0.6259,
+      "step": 29933
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.668819321330062e-06,
+      "loss": 0.6246,
+      "step": 29934
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6682162499846195e-06,
+      "loss": 0.6113,
+      "step": 29935
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6676132057315285e-06,
+      "loss": 0.651,
+      "step": 29936
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.667010188573859e-06,
+      "loss": 0.6406,
+      "step": 29937
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.66640719851467e-06,
+      "loss": 0.5763,
+      "step": 29938
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.665804235557036e-06,
+      "loss": 0.6265,
+      "step": 29939
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.665201299704004e-06,
+      "loss": 0.6357,
+      "step": 29940
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.664598390958651e-06,
+      "loss": 0.6394,
+      "step": 29941
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.663995509324032e-06,
+      "loss": 0.5794,
+      "step": 29942
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.663392654803219e-06,
+      "loss": 0.6193,
+      "step": 29943
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.66278982739927e-06,
+      "loss": 0.5768,
+      "step": 29944
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.66218702711525e-06,
+      "loss": 0.6395,
+      "step": 29945
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.661584253954216e-06,
+      "loss": 0.5712,
+      "step": 29946
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.660981507919241e-06,
+      "loss": 0.5962,
+      "step": 29947
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.660378789013381e-06,
+      "loss": 0.6014,
+      "step": 29948
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6597760972396975e-06,
+      "loss": 0.5138,
+      "step": 29949
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.659173432601259e-06,
+      "loss": 0.584,
+      "step": 29950
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.658570795101125e-06,
+      "loss": 0.6353,
+      "step": 29951
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6579681847423565e-06,
+      "loss": 0.6471,
+      "step": 29952
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.657365601528012e-06,
+      "loss": 0.5479,
+      "step": 29953
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6567630454611626e-06,
+      "loss": 0.5682,
+      "step": 29954
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.65616051654486e-06,
+      "loss": 0.6144,
+      "step": 29955
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6555580147821785e-06,
+      "loss": 0.5612,
+      "step": 29956
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.654955540176165e-06,
+      "loss": 0.6496,
+      "step": 29957
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.65435309272989e-06,
+      "loss": 0.647,
+      "step": 29958
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.653750672446411e-06,
+      "loss": 0.5476,
+      "step": 29959
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.653148279328793e-06,
+      "loss": 0.6348,
+      "step": 29960
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.652545913380094e-06,
+      "loss": 0.5825,
+      "step": 29961
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.651943574603377e-06,
+      "loss": 0.6756,
+      "step": 29962
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.651341263001695e-06,
+      "loss": 0.6181,
+      "step": 29963
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.65073897857812e-06,
+      "loss": 0.5998,
+      "step": 29964
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6501367213357065e-06,
+      "loss": 0.5281,
+      "step": 29965
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.649534491277514e-06,
+      "loss": 0.5708,
+      "step": 29966
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.648932288406601e-06,
+      "loss": 0.6271,
+      "step": 29967
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6483301127260334e-06,
+      "loss": 0.6223,
+      "step": 29968
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.647727964238868e-06,
+      "loss": 0.598,
+      "step": 29969
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.647125842948161e-06,
+      "loss": 0.5429,
+      "step": 29970
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.646523748856979e-06,
+      "loss": 0.5339,
+      "step": 29971
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.645921681968377e-06,
+      "loss": 0.5709,
+      "step": 29972
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.645319642285416e-06,
+      "loss": 0.6017,
+      "step": 29973
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.644717629811149e-06,
+      "loss": 0.5265,
+      "step": 29974
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.644115644548645e-06,
+      "loss": 0.6299,
+      "step": 29975
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6435136865009535e-06,
+      "loss": 0.6103,
+      "step": 29976
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.642911755671145e-06,
+      "loss": 0.6027,
+      "step": 29977
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.642309852062265e-06,
+      "loss": 0.6165,
+      "step": 29978
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.64170797567738e-06,
+      "loss": 0.6372,
+      "step": 29979
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.641106126519542e-06,
+      "loss": 0.5721,
+      "step": 29980
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.640504304591817e-06,
+      "loss": 0.5645,
+      "step": 29981
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.639902509897259e-06,
+      "loss": 0.6548,
+      "step": 29982
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.639300742438927e-06,
+      "loss": 0.6367,
+      "step": 29983
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.638699002219874e-06,
+      "loss": 0.5732,
+      "step": 29984
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.638097289243165e-06,
+      "loss": 0.5808,
+      "step": 29985
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.637495603511855e-06,
+      "loss": 0.5989,
+      "step": 29986
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.636893945028995e-06,
+      "loss": 0.5912,
+      "step": 29987
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.636292313797651e-06,
+      "loss": 0.6518,
+      "step": 29988
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.635690709820878e-06,
+      "loss": 0.611,
+      "step": 29989
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.635089133101731e-06,
+      "loss": 0.6562,
+      "step": 29990
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.634487583643264e-06,
+      "loss": 0.6159,
+      "step": 29991
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.63388606144854e-06,
+      "loss": 0.5858,
+      "step": 29992
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.633284566520613e-06,
+      "loss": 0.64,
+      "step": 29993
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.632683098862538e-06,
+      "loss": 0.5692,
+      "step": 29994
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6320816584773685e-06,
+      "loss": 0.5898,
+      "step": 29995
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.631480245368169e-06,
+      "loss": 0.6406,
+      "step": 29996
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.630878859537986e-06,
+      "loss": 0.6366,
+      "step": 29997
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.630277500989884e-06,
+      "loss": 0.5706,
+      "step": 29998
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6296761697269145e-06,
+      "loss": 0.6255,
+      "step": 29999
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.629074865752132e-06,
+      "loss": 0.6279,
+      "step": 30000
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6284735890685915e-06,
+      "loss": 0.5564,
+      "step": 30001
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6278723396793525e-06,
+      "loss": 0.5777,
+      "step": 30002
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.627271117587467e-06,
+      "loss": 0.6287,
+      "step": 30003
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.626669922795991e-06,
+      "loss": 0.6431,
+      "step": 30004
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.626068755307974e-06,
+      "loss": 0.6136,
+      "step": 30005
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.625467615126481e-06,
+      "loss": 0.5581,
+      "step": 30006
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.62486650225456e-06,
+      "loss": 0.5958,
+      "step": 30007
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.624265416695263e-06,
+      "loss": 0.5444,
+      "step": 30008
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6236643584516515e-06,
+      "loss": 0.5299,
+      "step": 30009
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.623063327526775e-06,
+      "loss": 0.5892,
+      "step": 30010
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.62246232392369e-06,
+      "loss": 0.61,
+      "step": 30011
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6218613476454444e-06,
+      "loss": 0.5641,
+      "step": 30012
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6212603986951e-06,
+      "loss": 0.5987,
+      "step": 30013
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.620659477075703e-06,
+      "loss": 0.6155,
+      "step": 30014
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.620058582790318e-06,
+      "loss": 0.5936,
+      "step": 30015
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6194577158419825e-06,
+      "loss": 0.5883,
+      "step": 30016
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.618856876233764e-06,
+      "loss": 0.5614,
+      "step": 30017
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.618256063968705e-06,
+      "loss": 0.6157,
+      "step": 30018
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6176552790498665e-06,
+      "loss": 0.5509,
+      "step": 30019
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.617054521480297e-06,
+      "loss": 0.5453,
+      "step": 30020
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6164537912630515e-06,
+      "loss": 0.6289,
+      "step": 30021
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.615853088401176e-06,
+      "loss": 0.6495,
+      "step": 30022
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.615252412897732e-06,
+      "loss": 0.6179,
+      "step": 30023
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6146517647557685e-06,
+      "loss": 0.6392,
+      "step": 30024
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.614051143978332e-06,
+      "loss": 0.5775,
+      "step": 30025
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.613450550568482e-06,
+      "loss": 0.5762,
+      "step": 30026
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.612849984529269e-06,
+      "loss": 0.6075,
+      "step": 30027
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.612249445863741e-06,
+      "loss": 0.5728,
+      "step": 30028
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.61164893457495e-06,
+      "loss": 0.5741,
+      "step": 30029
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.611048450665951e-06,
+      "loss": 0.5755,
+      "step": 30030
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.610447994139792e-06,
+      "loss": 0.59,
+      "step": 30031
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6098475649995265e-06,
+      "loss": 0.5859,
+      "step": 30032
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.609247163248201e-06,
+      "loss": 0.6118,
+      "step": 30033
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6086467888888706e-06,
+      "loss": 0.5775,
+      "step": 30034
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.608046441924584e-06,
+      "loss": 0.5836,
+      "step": 30035
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.607446122358394e-06,
+      "loss": 0.5438,
+      "step": 30036
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6068458301933515e-06,
+      "loss": 0.5726,
+      "step": 30037
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6062455654325035e-06,
+      "loss": 0.5597,
+      "step": 30038
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.605645328078898e-06,
+      "loss": 0.6535,
+      "step": 30039
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.6050451181355925e-06,
+      "loss": 0.5937,
+      "step": 30040
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.604444935605629e-06,
+      "loss": 0.5894,
+      "step": 30041
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.603844780492067e-06,
+      "loss": 0.6218,
+      "step": 30042
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.603244652797945e-06,
+      "loss": 0.593,
+      "step": 30043
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.602644552526319e-06,
+      "loss": 0.6203,
+      "step": 30044
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.602044479680238e-06,
+      "loss": 0.556,
+      "step": 30045
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.601444434262745e-06,
+      "loss": 0.5938,
+      "step": 30046
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.600844416276899e-06,
+      "loss": 0.5995,
+      "step": 30047
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.600244425725743e-06,
+      "loss": 0.6235,
+      "step": 30048
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.599644462612326e-06,
+      "loss": 0.5914,
+      "step": 30049
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.599044526939694e-06,
+      "loss": 0.5861,
+      "step": 30050
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.598444618710902e-06,
+      "loss": 0.6251,
+      "step": 30051
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.597844737928991e-06,
+      "loss": 0.5825,
+      "step": 30052
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5972448845970206e-06,
+      "loss": 0.6007,
+      "step": 30053
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.596645058718023e-06,
+      "loss": 0.5891,
+      "step": 30054
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.596045260295059e-06,
+      "loss": 0.5702,
+      "step": 30055
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.595445489331168e-06,
+      "loss": 0.6219,
+      "step": 30056
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.594845745829405e-06,
+      "loss": 0.5395,
+      "step": 30057
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.594246029792813e-06,
+      "loss": 0.6407,
+      "step": 30058
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.593646341224441e-06,
+      "loss": 0.6057,
+      "step": 30059
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.593046680127331e-06,
+      "loss": 0.5879,
+      "step": 30060
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.592447046504538e-06,
+      "loss": 0.6247,
+      "step": 30061
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5918474403591e-06,
+      "loss": 0.6105,
+      "step": 30062
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.591247861694076e-06,
+      "loss": 0.6489,
+      "step": 30063
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.590648310512503e-06,
+      "loss": 0.6246,
+      "step": 30064
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5900487868174315e-06,
+      "loss": 0.5928,
+      "step": 30065
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.589449290611906e-06,
+      "loss": 0.5742,
+      "step": 30066
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.58884982189897e-06,
+      "loss": 0.6291,
+      "step": 30067
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.588250380681676e-06,
+      "loss": 0.634,
+      "step": 30068
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.587650966963067e-06,
+      "loss": 0.5492,
+      "step": 30069
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5870515807461875e-06,
+      "loss": 0.606,
+      "step": 30070
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.586452222034081e-06,
+      "loss": 0.6335,
+      "step": 30071
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5858528908298005e-06,
+      "loss": 0.6105,
+      "step": 30072
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.585253587136383e-06,
+      "loss": 0.6136,
+      "step": 30073
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.584654310956881e-06,
+      "loss": 0.6299,
+      "step": 30074
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.584055062294337e-06,
+      "loss": 0.6621,
+      "step": 30075
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.583455841151796e-06,
+      "loss": 0.591,
+      "step": 30076
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.582856647532298e-06,
+      "loss": 0.6353,
+      "step": 30077
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.582257481438894e-06,
+      "loss": 0.5959,
+      "step": 30078
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.581658342874625e-06,
+      "loss": 0.6615,
+      "step": 30079
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.581059231842543e-06,
+      "loss": 0.5569,
+      "step": 30080
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.580460148345679e-06,
+      "loss": 0.6105,
+      "step": 30081
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.579861092387088e-06,
+      "loss": 0.6261,
+      "step": 30082
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.579262063969809e-06,
+      "loss": 0.599,
+      "step": 30083
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.578663063096884e-06,
+      "loss": 0.6247,
+      "step": 30084
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.578064089771363e-06,
+      "loss": 0.6053,
+      "step": 30085
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.577465143996288e-06,
+      "loss": 0.6209,
+      "step": 30086
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5768662257747e-06,
+      "loss": 0.6243,
+      "step": 30087
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.576267335109638e-06,
+      "loss": 0.568,
+      "step": 30088
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.575668472004155e-06,
+      "loss": 0.6153,
+      "step": 30089
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.575069636461285e-06,
+      "loss": 0.6109,
+      "step": 30090
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.574470828484082e-06,
+      "loss": 0.5894,
+      "step": 30091
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.573872048075574e-06,
+      "loss": 0.6187,
+      "step": 30092
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.573273295238816e-06,
+      "loss": 0.586,
+      "step": 30093
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5726745699768406e-06,
+      "loss": 0.5945,
+      "step": 30094
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5720758722927e-06,
+      "loss": 0.6039,
+      "step": 30095
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.571477202189432e-06,
+      "loss": 0.6313,
+      "step": 30096
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.570878559670076e-06,
+      "loss": 0.5703,
+      "step": 30097
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.570279944737674e-06,
+      "loss": 0.6036,
+      "step": 30098
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5696813573952735e-06,
+      "loss": 0.5849,
+      "step": 30099
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5690827976459085e-06,
+      "loss": 0.6186,
+      "step": 30100
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.568484265492628e-06,
+      "loss": 0.6028,
+      "step": 30101
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.567885760938468e-06,
+      "loss": 0.5609,
+      "step": 30102
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.567287283986473e-06,
+      "loss": 0.5828,
+      "step": 30103
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.566688834639683e-06,
+      "loss": 0.5907,
+      "step": 30104
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.566090412901134e-06,
+      "loss": 0.6023,
+      "step": 30105
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.565492018773874e-06,
+      "loss": 0.5665,
+      "step": 30106
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.564893652260941e-06,
+      "loss": 0.7133,
+      "step": 30107
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.564295313365375e-06,
+      "loss": 0.5999,
+      "step": 30108
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.563697002090214e-06,
+      "loss": 0.5931,
+      "step": 30109
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.563098718438502e-06,
+      "loss": 0.5548,
+      "step": 30110
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.562500462413275e-06,
+      "loss": 0.5795,
+      "step": 30111
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.561902234017583e-06,
+      "loss": 0.6214,
+      "step": 30112
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5613040332544525e-06,
+      "loss": 0.5699,
+      "step": 30113
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.56070586012693e-06,
+      "loss": 0.5787,
+      "step": 30114
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5601077146380526e-06,
+      "loss": 0.5844,
+      "step": 30115
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.559509596790864e-06,
+      "loss": 0.5729,
+      "step": 30116
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.558911506588396e-06,
+      "loss": 0.537,
+      "step": 30117
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.558313444033701e-06,
+      "loss": 0.5834,
+      "step": 30118
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5577154091298e-06,
+      "loss": 0.5945,
+      "step": 30119
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.557117401879746e-06,
+      "loss": 0.5999,
+      "step": 30120
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.556519422286567e-06,
+      "loss": 0.6589,
+      "step": 30121
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.555921470353313e-06,
+      "loss": 0.7086,
+      "step": 30122
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.555323546083015e-06,
+      "loss": 0.6588,
+      "step": 30123
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5547256494787125e-06,
+      "loss": 0.6307,
+      "step": 30124
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.554127780543443e-06,
+      "loss": 0.6348,
+      "step": 30125
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.553529939280243e-06,
+      "loss": 0.603,
+      "step": 30126
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.552932125692155e-06,
+      "loss": 0.5852,
+      "step": 30127
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.552334339782212e-06,
+      "loss": 0.593,
+      "step": 30128
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.551736581553459e-06,
+      "loss": 0.5802,
+      "step": 30129
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.551138851008923e-06,
+      "loss": 0.6257,
+      "step": 30130
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.550541148151648e-06,
+      "loss": 0.5678,
+      "step": 30131
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.549943472984667e-06,
+      "loss": 0.6612,
+      "step": 30132
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.549345825511022e-06,
+      "loss": 0.6058,
+      "step": 30133
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.548748205733747e-06,
+      "loss": 0.6047,
+      "step": 30134
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.54815061365588e-06,
+      "loss": 0.5929,
+      "step": 30135
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.547553049280452e-06,
+      "loss": 0.5798,
+      "step": 30136
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.546955512610508e-06,
+      "loss": 0.5452,
+      "step": 30137
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.546358003649075e-06,
+      "loss": 0.6082,
+      "step": 30138
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.545760522399199e-06,
+      "loss": 0.5945,
+      "step": 30139
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.545163068863911e-06,
+      "loss": 0.6121,
+      "step": 30140
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.544565643046247e-06,
+      "loss": 0.6665,
+      "step": 30141
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.543968244949238e-06,
+      "loss": 0.5819,
+      "step": 30142
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5433708745759295e-06,
+      "loss": 0.6048,
+      "step": 30143
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.542773531929351e-06,
+      "loss": 0.5857,
+      "step": 30144
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5421762170125385e-06,
+      "loss": 0.6004,
+      "step": 30145
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.541578929828526e-06,
+      "loss": 0.5736,
+      "step": 30146
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.540981670380347e-06,
+      "loss": 0.5725,
+      "step": 30147
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.540384438671042e-06,
+      "loss": 0.6394,
+      "step": 30148
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.539787234703639e-06,
+      "loss": 0.626,
+      "step": 30149
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.539190058481184e-06,
+      "loss": 0.5387,
+      "step": 30150
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.538592910006696e-06,
+      "loss": 0.6049,
+      "step": 30151
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.53799578928322e-06,
+      "loss": 0.6175,
+      "step": 30152
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.537398696313784e-06,
+      "loss": 0.5591,
+      "step": 30153
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.536801631101429e-06,
+      "loss": 0.5295,
+      "step": 30154
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5362045936491795e-06,
+      "loss": 0.6535,
+      "step": 30155
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.535607583960083e-06,
+      "loss": 0.6019,
+      "step": 30156
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5350106020371576e-06,
+      "loss": 0.5739,
+      "step": 30157
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.534413647883448e-06,
+      "loss": 0.5895,
+      "step": 30158
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.533816721501979e-06,
+      "loss": 0.6343,
+      "step": 30159
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.533219822895792e-06,
+      "loss": 0.6218,
+      "step": 30160
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.532622952067916e-06,
+      "loss": 0.628,
+      "step": 30161
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.532026109021384e-06,
+      "loss": 0.5502,
+      "step": 30162
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.531429293759224e-06,
+      "loss": 0.6381,
+      "step": 30163
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.530832506284478e-06,
+      "loss": 0.5985,
+      "step": 30164
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.530235746600174e-06,
+      "loss": 0.5903,
+      "step": 30165
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.529639014709339e-06,
+      "loss": 0.6152,
+      "step": 30166
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.529042310615019e-06,
+      "loss": 0.5332,
+      "step": 30167
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.52844563432023e-06,
+      "loss": 0.6171,
+      "step": 30168
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.527848985828014e-06,
+      "loss": 0.6007,
+      "step": 30169
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.527252365141397e-06,
+      "loss": 0.6273,
+      "step": 30170
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5266557722634165e-06,
+      "loss": 0.5942,
+      "step": 30171
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.526059207197101e-06,
+      "loss": 0.5293,
+      "step": 30172
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.525462669945482e-06,
+      "loss": 0.5932,
+      "step": 30173
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5248661605115864e-06,
+      "loss": 0.6306,
+      "step": 30174
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.524269678898453e-06,
+      "loss": 0.6643,
+      "step": 30175
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5236732251091065e-06,
+      "loss": 0.5984,
+      "step": 30176
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.523076799146586e-06,
+      "loss": 0.626,
+      "step": 30177
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.52248040101391e-06,
+      "loss": 0.5823,
+      "step": 30178
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.521884030714119e-06,
+      "loss": 0.6158,
+      "step": 30179
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.521287688250235e-06,
+      "loss": 0.6318,
+      "step": 30180
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.520691373625296e-06,
+      "loss": 0.6156,
+      "step": 30181
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.520095086842331e-06,
+      "loss": 0.5638,
+      "step": 30182
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.519498827904366e-06,
+      "loss": 0.5527,
+      "step": 30183
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.518902596814433e-06,
+      "loss": 0.5845,
+      "step": 30184
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.518306393575557e-06,
+      "loss": 0.5207,
+      "step": 30185
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.517710218190776e-06,
+      "loss": 0.5434,
+      "step": 30186
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5171140706631115e-06,
+      "loss": 0.589,
+      "step": 30187
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5165179509956025e-06,
+      "loss": 0.584,
+      "step": 30188
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.515921859191264e-06,
+      "loss": 0.6685,
+      "step": 30189
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.515325795253137e-06,
+      "loss": 0.6462,
+      "step": 30190
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5147297591842434e-06,
+      "loss": 0.5579,
+      "step": 30191
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.514133750987616e-06,
+      "loss": 0.5966,
+      "step": 30192
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.513537770666278e-06,
+      "loss": 0.6053,
+      "step": 30193
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.5129418182232706e-06,
+      "loss": 0.5985,
+      "step": 30194
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.512345893661605e-06,
+      "loss": 0.5955,
+      "step": 30195
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.511749996984319e-06,
+      "loss": 0.6289,
+      "step": 30196
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.511154128194437e-06,
+      "loss": 0.5723,
+      "step": 30197
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.510558287294991e-06,
+      "loss": 0.5854,
+      "step": 30198
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.509962474289006e-06,
+      "loss": 0.5765,
+      "step": 30199
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.509366689179509e-06,
+      "loss": 0.5771,
+      "step": 30200
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.508770931969525e-06,
+      "loss": 0.5687,
+      "step": 30201
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.5081752026620875e-06,
+      "loss": 0.607,
+      "step": 30202
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.5075795012602194e-06,
+      "loss": 0.6145,
+      "step": 30203
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.5069838277669495e-06,
+      "loss": 0.5884,
+      "step": 30204
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.506388182185303e-06,
+      "loss": 0.6605,
+      "step": 30205
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.505792564518304e-06,
+      "loss": 0.585,
+      "step": 30206
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.505196974768985e-06,
+      "loss": 0.5864,
+      "step": 30207
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.504601412940366e-06,
+      "loss": 0.5843,
+      "step": 30208
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.50400587903548e-06,
+      "loss": 0.6034,
+      "step": 30209
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.503410373057349e-06,
+      "loss": 0.6771,
+      "step": 30210
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.502814895009001e-06,
+      "loss": 0.5986,
+      "step": 30211
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.502219444893457e-06,
+      "loss": 0.5719,
+      "step": 30212
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.501624022713748e-06,
+      "loss": 0.643,
+      "step": 30213
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.501028628472896e-06,
+      "loss": 0.6612,
+      "step": 30214
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.500433262173935e-06,
+      "loss": 0.5756,
+      "step": 30215
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.499837923819874e-06,
+      "loss": 0.5371,
+      "step": 30216
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.499242613413752e-06,
+      "loss": 0.5615,
+      "step": 30217
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.498647330958586e-06,
+      "loss": 0.6115,
+      "step": 30218
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4980520764574075e-06,
+      "loss": 0.6383,
+      "step": 30219
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.497456849913237e-06,
+      "loss": 0.5698,
+      "step": 30220
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.496861651329102e-06,
+      "loss": 0.5704,
+      "step": 30221
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.49626648070802e-06,
+      "loss": 0.6186,
+      "step": 30222
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.495671338053022e-06,
+      "loss": 0.5652,
+      "step": 30223
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4950762233671305e-06,
+      "loss": 0.5868,
+      "step": 30224
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.494481136653366e-06,
+      "loss": 0.6404,
+      "step": 30225
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.493886077914762e-06,
+      "loss": 0.6364,
+      "step": 30226
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.493291047154328e-06,
+      "loss": 0.5712,
+      "step": 30227
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.492696044375099e-06,
+      "loss": 0.6327,
+      "step": 30228
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.492101069580091e-06,
+      "loss": 0.6048,
+      "step": 30229
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.491506122772332e-06,
+      "loss": 0.594,
+      "step": 30230
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.490911203954842e-06,
+      "loss": 0.629,
+      "step": 30231
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4903163131306525e-06,
+      "loss": 0.5509,
+      "step": 30232
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.489721450302772e-06,
+      "loss": 0.6141,
+      "step": 30233
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.489126615474234e-06,
+      "loss": 0.6137,
+      "step": 30234
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.488531808648054e-06,
+      "loss": 0.5251,
+      "step": 30235
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.487937029827263e-06,
+      "loss": 0.595,
+      "step": 30236
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.487342279014877e-06,
+      "loss": 0.6135,
+      "step": 30237
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.48674755621392e-06,
+      "loss": 0.5979,
+      "step": 30238
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.486152861427409e-06,
+      "loss": 0.6125,
+      "step": 30239
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.485558194658373e-06,
+      "loss": 0.5504,
+      "step": 30240
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.484963555909831e-06,
+      "loss": 0.6012,
+      "step": 30241
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.484368945184806e-06,
+      "loss": 0.5679,
+      "step": 30242
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.483774362486313e-06,
+      "loss": 0.5696,
+      "step": 30243
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4831798078173815e-06,
+      "loss": 0.5909,
+      "step": 30244
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.482585281181029e-06,
+      "loss": 0.589,
+      "step": 30245
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.481990782580274e-06,
+      "loss": 0.5887,
+      "step": 30246
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.481396312018141e-06,
+      "loss": 0.5397,
+      "step": 30247
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.480801869497652e-06,
+      "loss": 0.655,
+      "step": 30248
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.480207455021824e-06,
+      "loss": 0.5859,
+      "step": 30249
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.479613068593674e-06,
+      "loss": 0.6228,
+      "step": 30250
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.479018710216233e-06,
+      "loss": 0.6255,
+      "step": 30251
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.478424379892509e-06,
+      "loss": 0.5716,
+      "step": 30252
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.477830077625536e-06,
+      "loss": 0.6111,
+      "step": 30253
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.477235803418317e-06,
+      "loss": 0.6175,
+      "step": 30254
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.476641557273885e-06,
+      "loss": 0.5999,
+      "step": 30255
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.476047339195252e-06,
+      "loss": 0.6055,
+      "step": 30256
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4754531491854426e-06,
+      "loss": 0.6021,
+      "step": 30257
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.474858987247475e-06,
+      "loss": 0.6113,
+      "step": 30258
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.474264853384366e-06,
+      "loss": 0.559,
+      "step": 30259
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.473670747599132e-06,
+      "loss": 0.6005,
+      "step": 30260
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.473076669894798e-06,
+      "loss": 0.5953,
+      "step": 30261
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.472482620274382e-06,
+      "loss": 0.5783,
+      "step": 30262
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.471888598740895e-06,
+      "loss": 0.5773,
+      "step": 30263
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.47129460529737e-06,
+      "loss": 0.5527,
+      "step": 30264
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.470700639946809e-06,
+      "loss": 0.6697,
+      "step": 30265
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.47010670269224e-06,
+      "loss": 0.5985,
+      "step": 30266
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4695127935366744e-06,
+      "loss": 0.5654,
+      "step": 30267
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.468918912483139e-06,
+      "loss": 0.5908,
+      "step": 30268
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.468325059534646e-06,
+      "loss": 0.5878,
+      "step": 30269
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.467731234694212e-06,
+      "loss": 0.5875,
+      "step": 30270
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4671374379648545e-06,
+      "loss": 0.6075,
+      "step": 30271
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4665436693495945e-06,
+      "loss": 0.6227,
+      "step": 30272
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.465949928851443e-06,
+      "loss": 0.5541,
+      "step": 30273
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4653562164734246e-06,
+      "loss": 0.6164,
+      "step": 30274
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4647625322185515e-06,
+      "loss": 0.5758,
+      "step": 30275
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.464168876089842e-06,
+      "loss": 0.6016,
+      "step": 30276
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.463575248090307e-06,
+      "loss": 0.5857,
+      "step": 30277
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.462981648222972e-06,
+      "loss": 0.5492,
+      "step": 30278
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.462388076490848e-06,
+      "loss": 0.597,
+      "step": 30279
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.461794532896952e-06,
+      "loss": 0.552,
+      "step": 30280
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.461201017444298e-06,
+      "loss": 0.5994,
+      "step": 30281
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.460607530135906e-06,
+      "loss": 0.6246,
+      "step": 30282
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4600140709747885e-06,
+      "loss": 0.5649,
+      "step": 30283
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.459420639963959e-06,
+      "loss": 0.5405,
+      "step": 30284
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4588272371064415e-06,
+      "loss": 0.6242,
+      "step": 30285
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.458233862405244e-06,
+      "loss": 0.612,
+      "step": 30286
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.457640515863383e-06,
+      "loss": 0.6074,
+      "step": 30287
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.45704719748387e-06,
+      "loss": 0.653,
+      "step": 30288
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.456453907269728e-06,
+      "loss": 0.6677,
+      "step": 30289
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.455860645223962e-06,
+      "loss": 0.5859,
+      "step": 30290
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4552674113496e-06,
+      "loss": 0.6079,
+      "step": 30291
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.454674205649641e-06,
+      "loss": 0.5879,
+      "step": 30292
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.45408102812711e-06,
+      "loss": 0.6205,
+      "step": 30293
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.453487878785013e-06,
+      "loss": 0.5827,
+      "step": 30294
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.452894757626374e-06,
+      "loss": 0.6127,
+      "step": 30295
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4523016646542e-06,
+      "loss": 0.632,
+      "step": 30296
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.451708599871506e-06,
+      "loss": 0.5538,
+      "step": 30297
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.451115563281302e-06,
+      "loss": 0.6361,
+      "step": 30298
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.450522554886608e-06,
+      "loss": 0.6074,
+      "step": 30299
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.449929574690435e-06,
+      "loss": 0.5844,
+      "step": 30300
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.449336622695791e-06,
+      "loss": 0.6631,
+      "step": 30301
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.448743698905698e-06,
+      "loss": 0.6208,
+      "step": 30302
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.448150803323162e-06,
+      "loss": 0.5989,
+      "step": 30303
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4475579359511985e-06,
+      "loss": 0.5714,
+      "step": 30304
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4469650967928164e-06,
+      "loss": 0.6419,
+      "step": 30305
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.446372285851034e-06,
+      "loss": 0.6374,
+      "step": 30306
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4457795031288605e-06,
+      "loss": 0.5743,
+      "step": 30307
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.445186748629309e-06,
+      "loss": 0.6518,
+      "step": 30308
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.444594022355385e-06,
+      "loss": 0.5863,
+      "step": 30309
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.44400132431011e-06,
+      "loss": 0.6096,
+      "step": 30310
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.443408654496488e-06,
+      "loss": 0.6907,
+      "step": 30311
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4428160129175375e-06,
+      "loss": 0.6591,
+      "step": 30312
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4422233995762664e-06,
+      "loss": 0.6257,
+      "step": 30313
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.441630814475686e-06,
+      "loss": 0.6597,
+      "step": 30314
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.441038257618804e-06,
+      "loss": 0.5949,
+      "step": 30315
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.440445729008638e-06,
+      "loss": 0.5796,
+      "step": 30316
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.439853228648194e-06,
+      "loss": 0.6032,
+      "step": 30317
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.439260756540485e-06,
+      "loss": 0.563,
+      "step": 30318
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.438668312688519e-06,
+      "loss": 0.5652,
+      "step": 30319
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.43807589709531e-06,
+      "loss": 0.6181,
+      "step": 30320
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.437483509763866e-06,
+      "loss": 0.5755,
+      "step": 30321
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.436891150697194e-06,
+      "loss": 0.5946,
+      "step": 30322
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.43629881989831e-06,
+      "loss": 0.5789,
+      "step": 30323
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4357065173702216e-06,
+      "loss": 0.5604,
+      "step": 30324
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.435114243115939e-06,
+      "loss": 0.5711,
+      "step": 30325
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.434521997138467e-06,
+      "loss": 0.5696,
+      "step": 30326
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.433929779440821e-06,
+      "loss": 0.5826,
+      "step": 30327
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4333375900260045e-06,
+      "loss": 0.6075,
+      "step": 30328
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.432745428897038e-06,
+      "loss": 0.5811,
+      "step": 30329
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.432153296056914e-06,
+      "loss": 0.6009,
+      "step": 30330
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4315611915086556e-06,
+      "loss": 0.6363,
+      "step": 30331
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4309691152552605e-06,
+      "loss": 0.5923,
+      "step": 30332
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.430377067299746e-06,
+      "loss": 0.6323,
+      "step": 30333
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.429785047645116e-06,
+      "loss": 0.6075,
+      "step": 30334
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4291930562943805e-06,
+      "loss": 0.5947,
+      "step": 30335
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.428601093250543e-06,
+      "loss": 0.6493,
+      "step": 30336
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.428009158516618e-06,
+      "loss": 0.6477,
+      "step": 30337
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.42741725209561e-06,
+      "loss": 0.5862,
+      "step": 30338
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.426825373990524e-06,
+      "loss": 0.623,
+      "step": 30339
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.426233524204374e-06,
+      "loss": 0.6158,
+      "step": 30340
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.425641702740164e-06,
+      "loss": 0.6261,
+      "step": 30341
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.425049909600902e-06,
+      "loss": 0.6065,
+      "step": 30342
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.424458144789588e-06,
+      "loss": 0.6343,
+      "step": 30343
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.423866408309241e-06,
+      "loss": 0.5669,
+      "step": 30344
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4232747001628605e-06,
+      "loss": 0.6448,
+      "step": 30345
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.422683020353454e-06,
+      "loss": 0.5903,
+      "step": 30346
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.422091368884026e-06,
+      "loss": 0.5637,
+      "step": 30347
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.421499745757588e-06,
+      "loss": 0.6382,
+      "step": 30348
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.420908150977141e-06,
+      "loss": 0.5513,
+      "step": 30349
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.420316584545697e-06,
+      "loss": 0.5977,
+      "step": 30350
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4197250464662576e-06,
+      "loss": 0.5946,
+      "step": 30351
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.419133536741831e-06,
+      "loss": 0.6426,
+      "step": 30352
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.418542055375417e-06,
+      "loss": 0.597,
+      "step": 30353
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.417950602370028e-06,
+      "loss": 0.5717,
+      "step": 30354
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.417359177728664e-06,
+      "loss": 0.5916,
+      "step": 30355
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4167677814543406e-06,
+      "loss": 0.6138,
+      "step": 30356
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.416176413550047e-06,
+      "loss": 0.6022,
+      "step": 30357
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.415585074018801e-06,
+      "loss": 0.6452,
+      "step": 30358
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.414993762863602e-06,
+      "loss": 0.6313,
+      "step": 30359
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.414402480087452e-06,
+      "loss": 0.5852,
+      "step": 30360
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.413811225693363e-06,
+      "loss": 0.6226,
+      "step": 30361
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.413219999684334e-06,
+      "loss": 0.6235,
+      "step": 30362
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.412628802063372e-06,
+      "loss": 0.618,
+      "step": 30363
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.412037632833473e-06,
+      "loss": 0.6538,
+      "step": 30364
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.411446491997654e-06,
+      "loss": 0.6219,
+      "step": 30365
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4108553795589075e-06,
+      "loss": 0.556,
+      "step": 30366
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.410264295520248e-06,
+      "loss": 0.596,
+      "step": 30367
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.409673239884665e-06,
+      "loss": 0.578,
+      "step": 30368
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.409082212655175e-06,
+      "loss": 0.6004,
+      "step": 30369
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.40849121383477e-06,
+      "loss": 0.5583,
+      "step": 30370
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.407900243426464e-06,
+      "loss": 0.6367,
+      "step": 30371
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.407309301433254e-06,
+      "loss": 0.5962,
+      "step": 30372
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.406718387858143e-06,
+      "loss": 0.6122,
+      "step": 30373
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.406127502704131e-06,
+      "loss": 0.5731,
+      "step": 30374
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.405536645974226e-06,
+      "loss": 0.5861,
+      "step": 30375
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.404945817671429e-06,
+      "loss": 0.6514,
+      "step": 30376
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.404355017798736e-06,
+      "loss": 0.6316,
+      "step": 30377
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.403764246359158e-06,
+      "loss": 0.5667,
+      "step": 30378
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4031735033556934e-06,
+      "loss": 0.5747,
+      "step": 30379
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.402582788791342e-06,
+      "loss": 0.6224,
+      "step": 30380
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.401992102669105e-06,
+      "loss": 0.5974,
+      "step": 30381
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.401401444991987e-06,
+      "loss": 0.6311,
+      "step": 30382
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.400810815762989e-06,
+      "loss": 0.6311,
+      "step": 30383
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.400220214985112e-06,
+      "loss": 0.5661,
+      "step": 30384
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.399629642661351e-06,
+      "loss": 0.6078,
+      "step": 30385
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.399039098794716e-06,
+      "loss": 0.6297,
+      "step": 30386
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.398448583388201e-06,
+      "loss": 0.6672,
+      "step": 30387
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.397858096444815e-06,
+      "loss": 0.5548,
+      "step": 30388
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.397267637967546e-06,
+      "loss": 0.6201,
+      "step": 30389
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.396677207959404e-06,
+      "loss": 0.6192,
+      "step": 30390
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.396086806423383e-06,
+      "loss": 0.6021,
+      "step": 30391
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.395496433362491e-06,
+      "loss": 0.623,
+      "step": 30392
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.394906088779718e-06,
+      "loss": 0.5852,
+      "step": 30393
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.394315772678076e-06,
+      "loss": 0.5995,
+      "step": 30394
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.393725485060549e-06,
+      "loss": 0.6027,
+      "step": 30395
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.393135225930151e-06,
+      "loss": 0.5468,
+      "step": 30396
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.392544995289872e-06,
+      "loss": 0.6009,
+      "step": 30397
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.391954793142712e-06,
+      "loss": 0.6389,
+      "step": 30398
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.391364619491676e-06,
+      "loss": 0.5713,
+      "step": 30399
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.390774474339758e-06,
+      "loss": 0.5803,
+      "step": 30400
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.390184357689958e-06,
+      "loss": 0.5546,
+      "step": 30401
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.389594269545272e-06,
+      "loss": 0.5997,
+      "step": 30402
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.389004209908703e-06,
+      "loss": 0.5601,
+      "step": 30403
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.388414178783243e-06,
+      "loss": 0.6166,
+      "step": 30404
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.387824176171901e-06,
+      "loss": 0.612,
+      "step": 30405
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.387234202077661e-06,
+      "loss": 0.648,
+      "step": 30406
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3866442565035315e-06,
+      "loss": 0.6191,
+      "step": 30407
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3860543394525035e-06,
+      "loss": 0.5902,
+      "step": 30408
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38546445092758e-06,
+      "loss": 0.5799,
+      "step": 30409
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.384874590931757e-06,
+      "loss": 0.5891,
+      "step": 30410
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3842847594680306e-06,
+      "loss": 0.561,
+      "step": 30411
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.383694956539393e-06,
+      "loss": 0.6291,
+      "step": 30412
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.383105182148852e-06,
+      "loss": 0.6363,
+      "step": 30413
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.382515436299393e-06,
+      "loss": 0.6317,
+      "step": 30414
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.381925718994023e-06,
+      "loss": 0.6844,
+      "step": 30415
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.381336030235732e-06,
+      "loss": 0.5725,
+      "step": 30416
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38074637002752e-06,
+      "loss": 0.6037,
+      "step": 30417
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.38015673837238e-06,
+      "loss": 0.5905,
+      "step": 30418
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.379567135273307e-06,
+      "loss": 0.7096,
+      "step": 30419
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.378977560733302e-06,
+      "loss": 0.5876,
+      "step": 30420
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.37838801475536e-06,
+      "loss": 0.6111,
+      "step": 30421
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377798497342472e-06,
+      "loss": 0.5996,
+      "step": 30422
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.377209008497635e-06,
+      "loss": 0.574,
+      "step": 30423
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.376619548223848e-06,
+      "loss": 0.5781,
+      "step": 30424
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.376030116524101e-06,
+      "loss": 0.6468,
+      "step": 30425
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3754407134014e-06,
+      "loss": 0.5411,
+      "step": 30426
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.374851338858724e-06,
+      "loss": 0.6166,
+      "step": 30427
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.374261992899079e-06,
+      "loss": 0.6521,
+      "step": 30428
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373672675525455e-06,
+      "loss": 0.5757,
+      "step": 30429
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.373083386740848e-06,
+      "loss": 0.5413,
+      "step": 30430
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3724941265482515e-06,
+      "loss": 0.6169,
+      "step": 30431
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371904894950666e-06,
+      "loss": 0.6361,
+      "step": 30432
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.371315691951073e-06,
+      "loss": 0.6709,
+      "step": 30433
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.370726517552477e-06,
+      "loss": 0.5238,
+      "step": 30434
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.370137371757866e-06,
+      "loss": 0.6293,
+      "step": 30435
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.369548254570237e-06,
+      "loss": 0.6478,
+      "step": 30436
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3689591659925846e-06,
+      "loss": 0.6111,
+      "step": 30437
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.368370106027898e-06,
+      "loss": 0.5867,
+      "step": 30438
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.367781074679173e-06,
+      "loss": 0.6147,
+      "step": 30439
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.367192071949398e-06,
+      "loss": 0.6178,
+      "step": 30440
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366603097841574e-06,
+      "loss": 0.64,
+      "step": 30441
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.366014152358685e-06,
+      "loss": 0.6053,
+      "step": 30442
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3654252355037355e-06,
+      "loss": 0.6303,
+      "step": 30443
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.364836347279704e-06,
+      "loss": 0.6418,
+      "step": 30444
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.364247487689592e-06,
+      "loss": 0.6424,
+      "step": 30445
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.363658656736386e-06,
+      "loss": 0.5967,
+      "step": 30446
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.363069854423084e-06,
+      "loss": 0.6475,
+      "step": 30447
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.362481080752676e-06,
+      "loss": 0.6045,
+      "step": 30448
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.361892335728151e-06,
+      "loss": 0.6316,
+      "step": 30449
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3613036193525e-06,
+      "loss": 0.5797,
+      "step": 30450
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.36071493162872e-06,
+      "loss": 0.6031,
+      "step": 30451
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.360126272559796e-06,
+      "loss": 0.6157,
+      "step": 30452
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.359537642148729e-06,
+      "loss": 0.6313,
+      "step": 30453
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358949040398495e-06,
+      "loss": 0.6336,
+      "step": 30454
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.358360467312097e-06,
+      "loss": 0.554,
+      "step": 30455
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.357771922892522e-06,
+      "loss": 0.5868,
+      "step": 30456
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.357183407142756e-06,
+      "loss": 0.5672,
+      "step": 30457
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3565949200657985e-06,
+      "loss": 0.5896,
+      "step": 30458
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3560064616646345e-06,
+      "loss": 0.6362,
+      "step": 30459
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.355418031942254e-06,
+      "loss": 0.6014,
+      "step": 30460
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.354829630901645e-06,
+      "loss": 0.5239,
+      "step": 30461
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.354241258545803e-06,
+      "loss": 0.6166,
+      "step": 30462
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35365291487771e-06,
+      "loss": 0.6606,
+      "step": 30463
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.353064599900369e-06,
+      "loss": 0.5648,
+      "step": 30464
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.352476313616753e-06,
+      "loss": 0.568,
+      "step": 30465
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.35188805602986e-06,
+      "loss": 0.5857,
+      "step": 30466
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.351299827142677e-06,
+      "loss": 0.621,
+      "step": 30467
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.350711626958197e-06,
+      "loss": 0.6192,
+      "step": 30468
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.350123455479401e-06,
+      "loss": 0.5877,
+      "step": 30469
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.349535312709291e-06,
+      "loss": 0.6135,
+      "step": 30470
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.348947198650839e-06,
+      "loss": 0.6575,
+      "step": 30471
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.348359113307046e-06,
+      "loss": 0.5731,
+      "step": 30472
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.347771056680891e-06,
+      "loss": 0.5558,
+      "step": 30473
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.347183028775369e-06,
+      "loss": 0.6538,
+      "step": 30474
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3465950295934665e-06,
+      "loss": 0.5688,
+      "step": 30475
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.346007059138171e-06,
+      "loss": 0.6151,
+      "step": 30476
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.345419117412468e-06,
+      "loss": 0.5238,
+      "step": 30477
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.344831204419344e-06,
+      "loss": 0.5878,
+      "step": 30478
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.344243320161793e-06,
+      "loss": 0.6065,
+      "step": 30479
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.343655464642798e-06,
+      "loss": 0.6258,
+      "step": 30480
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.343067637865344e-06,
+      "loss": 0.6118,
+      "step": 30481
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.342479839832419e-06,
+      "loss": 0.6421,
+      "step": 30482
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.341892070547012e-06,
+      "loss": 0.5739,
+      "step": 30483
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.341304330012107e-06,
+      "loss": 0.5218,
+      "step": 30484
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.340716618230695e-06,
+      "loss": 0.6182,
+      "step": 30485
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3401289352057585e-06,
+      "loss": 0.578,
+      "step": 30486
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.339541280940285e-06,
+      "loss": 0.6021,
+      "step": 30487
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.338953655437256e-06,
+      "loss": 0.5641,
+      "step": 30488
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.3383660586996655e-06,
+      "loss": 0.5964,
+      "step": 30489
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.337778490730491e-06,
+      "loss": 0.5803,
+      "step": 30490
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.337190951532729e-06,
+      "loss": 0.6048,
+      "step": 30491
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.336603441109353e-06,
+      "loss": 0.6321,
+      "step": 30492
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3360159594633554e-06,
+      "loss": 0.5983,
+      "step": 30493
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.335428506597716e-06,
+      "loss": 0.6565,
+      "step": 30494
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.334841082515427e-06,
+      "loss": 0.6251,
+      "step": 30495
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.334253687219471e-06,
+      "loss": 0.5972,
+      "step": 30496
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.333666320712829e-06,
+      "loss": 0.6138,
+      "step": 30497
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3330789829984895e-06,
+      "loss": 0.6162,
+      "step": 30498
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.332491674079431e-06,
+      "loss": 0.582,
+      "step": 30499
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.331904393958646e-06,
+      "loss": 0.5934,
+      "step": 30500
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.331317142639111e-06,
+      "loss": 0.588,
+      "step": 30501
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330729920123822e-06,
+      "loss": 0.6177,
+      "step": 30502
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.330142726415746e-06,
+      "loss": 0.5853,
+      "step": 30503
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.329555561517879e-06,
+      "loss": 0.6491,
+      "step": 30504
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328968425433198e-06,
+      "loss": 0.6015,
+      "step": 30505
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.328381318164693e-06,
+      "loss": 0.5897,
+      "step": 30506
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.327794239715338e-06,
+      "loss": 0.5927,
+      "step": 30507
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.327207190088131e-06,
+      "loss": 0.5871,
+      "step": 30508
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.326620169286038e-06,
+      "loss": 0.5921,
+      "step": 30509
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.326033177312052e-06,
+      "loss": 0.6113,
+      "step": 30510
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.325446214169151e-06,
+      "loss": 0.6,
+      "step": 30511
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324859279860322e-06,
+      "loss": 0.5555,
+      "step": 30512
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.324272374388546e-06,
+      "loss": 0.5616,
+      "step": 30513
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.323685497756804e-06,
+      "loss": 0.5805,
+      "step": 30514
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.323098649968074e-06,
+      "loss": 0.5972,
+      "step": 30515
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3225118310253466e-06,
+      "loss": 0.6071,
+      "step": 30516
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321925040931599e-06,
+      "loss": 0.5783,
+      "step": 30517
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.321338279689814e-06,
+      "loss": 0.6215,
+      "step": 30518
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.320751547302972e-06,
+      "loss": 0.5846,
+      "step": 30519
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.32016484377405e-06,
+      "loss": 0.6006,
+      "step": 30520
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3195781691060375e-06,
+      "loss": 0.6292,
+      "step": 30521
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.31899152330191e-06,
+      "loss": 0.6572,
+      "step": 30522
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.318404906364653e-06,
+      "loss": 0.6248,
+      "step": 30523
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317818318297244e-06,
+      "loss": 0.6946,
+      "step": 30524
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.317231759102665e-06,
+      "loss": 0.5841,
+      "step": 30525
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.316645228783892e-06,
+      "loss": 0.6217,
+      "step": 30526
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.316058727343912e-06,
+      "loss": 0.5376,
+      "step": 30527
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3154722547857e-06,
+      "loss": 0.5427,
+      "step": 30528
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.314885811112244e-06,
+      "loss": 0.5584,
+      "step": 30529
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.314299396326511e-06,
+      "loss": 0.6418,
+      "step": 30530
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.313713010431492e-06,
+      "loss": 0.6005,
+      "step": 30531
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.31312665343016e-06,
+      "loss": 0.6319,
+      "step": 30532
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3125403253255e-06,
+      "loss": 0.5636,
+      "step": 30533
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3119540261204865e-06,
+      "loss": 0.6674,
+      "step": 30534
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3113677558181025e-06,
+      "loss": 0.6027,
+      "step": 30535
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.310781514421319e-06,
+      "loss": 0.5569,
+      "step": 30536
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.310195301933127e-06,
+      "loss": 0.5784,
+      "step": 30537
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.309609118356498e-06,
+      "loss": 0.5702,
+      "step": 30538
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.309022963694407e-06,
+      "loss": 0.5818,
+      "step": 30539
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.308436837949846e-06,
+      "loss": 0.5052,
+      "step": 30540
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.307850741125775e-06,
+      "loss": 0.5912,
+      "step": 30541
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3072646732251866e-06,
+      "loss": 0.6051,
+      "step": 30542
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.30667863425105e-06,
+      "loss": 0.6247,
+      "step": 30543
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.30609262420635e-06,
+      "loss": 0.5798,
+      "step": 30544
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.305506643094061e-06,
+      "loss": 0.6391,
+      "step": 30545
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304920690917159e-06,
+      "loss": 0.5861,
+      "step": 30546
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.304334767678621e-06,
+      "loss": 0.6186,
+      "step": 30547
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.303748873381429e-06,
+      "loss": 0.5992,
+      "step": 30548
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.303163008028554e-06,
+      "loss": 0.6114,
+      "step": 30549
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.302577171622979e-06,
+      "loss": 0.5122,
+      "step": 30550
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.301991364167678e-06,
+      "loss": 0.6222,
+      "step": 30551
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.3014055856656276e-06,
+      "loss": 0.6093,
+      "step": 30552
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.300819836119802e-06,
+      "loss": 0.642,
+      "step": 30553
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.300234115533183e-06,
+      "loss": 0.59,
+      "step": 30554
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.299648423908743e-06,
+      "loss": 0.5829,
+      "step": 30555
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.29906276124946e-06,
+      "loss": 0.6123,
+      "step": 30556
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.298477127558307e-06,
+      "loss": 0.6576,
+      "step": 30557
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.297891522838259e-06,
+      "loss": 0.5813,
+      "step": 30558
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.297305947092297e-06,
+      "loss": 0.6282,
+      "step": 30559
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.296720400323391e-06,
+      "loss": 0.5972,
+      "step": 30560
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.296134882534521e-06,
+      "loss": 0.6002,
+      "step": 30561
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2955493937286615e-06,
+      "loss": 0.6331,
+      "step": 30562
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.294963933908786e-06,
+      "loss": 0.6002,
+      "step": 30563
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.294378503077865e-06,
+      "loss": 0.621,
+      "step": 30564
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2937931012388825e-06,
+      "loss": 0.5453,
+      "step": 30565
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2932077283948045e-06,
+      "loss": 0.5634,
+      "step": 30566
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.292622384548616e-06,
+      "loss": 0.5753,
+      "step": 30567
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.292037069703277e-06,
+      "loss": 0.5911,
+      "step": 30568
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.291451783861775e-06,
+      "loss": 0.6244,
+      "step": 30569
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.290866527027073e-06,
+      "loss": 0.5568,
+      "step": 30570
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.290281299202155e-06,
+      "loss": 0.5752,
+      "step": 30571
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.289696100389989e-06,
+      "loss": 0.586,
+      "step": 30572
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.28911093059355e-06,
+      "loss": 0.5273,
+      "step": 30573
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.288525789815806e-06,
+      "loss": 0.6235,
+      "step": 30574
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.28794067805974e-06,
+      "loss": 0.6125,
+      "step": 30575
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.28735559532832e-06,
+      "loss": 0.6289,
+      "step": 30576
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.286770541624516e-06,
+      "loss": 0.634,
+      "step": 30577
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.28618551695131e-06,
+      "loss": 0.6038,
+      "step": 30578
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.285600521311662e-06,
+      "loss": 0.6069,
+      "step": 30579
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.285015554708556e-06,
+      "loss": 0.5302,
+      "step": 30580
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.284430617144955e-06,
+      "loss": 0.6103,
+      "step": 30581
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2838457086238395e-06,
+      "loss": 0.596,
+      "step": 30582
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2832608291481785e-06,
+      "loss": 0.5968,
+      "step": 30583
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.282675978720944e-06,
+      "loss": 0.5632,
+      "step": 30584
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.282091157345102e-06,
+      "loss": 0.6385,
+      "step": 30585
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2815063650236345e-06,
+      "loss": 0.6257,
+      "step": 30586
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.280921601759504e-06,
+      "loss": 0.596,
+      "step": 30587
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.280336867555689e-06,
+      "loss": 0.586,
+      "step": 30588
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.279752162415157e-06,
+      "loss": 0.6718,
+      "step": 30589
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.27916748634088e-06,
+      "loss": 0.6206,
+      "step": 30590
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.278582839335825e-06,
+      "loss": 0.5611,
+      "step": 30591
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.277998221402971e-06,
+      "loss": 0.5959,
+      "step": 30592
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.277413632545281e-06,
+      "loss": 0.5967,
+      "step": 30593
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.27682907276573e-06,
+      "loss": 0.5986,
+      "step": 30594
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.276244542067283e-06,
+      "loss": 0.5747,
+      "step": 30595
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.275660040452917e-06,
+      "loss": 0.6198,
+      "step": 30596
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.275075567925599e-06,
+      "loss": 0.6531,
+      "step": 30597
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.274491124488295e-06,
+      "loss": 0.6424,
+      "step": 30598
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.273906710143981e-06,
+      "loss": 0.5382,
+      "step": 30599
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.273322324895626e-06,
+      "loss": 0.5615,
+      "step": 30600
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.272737968746196e-06,
+      "loss": 0.5908,
+      "step": 30601
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.272153641698658e-06,
+      "loss": 0.5467,
+      "step": 30602
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.271569343755989e-06,
+      "loss": 0.5916,
+      "step": 30603
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.27098507492115e-06,
+      "loss": 0.5772,
+      "step": 30604
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.27040083519712e-06,
+      "loss": 0.5947,
+      "step": 30605
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.269816624586855e-06,
+      "loss": 0.6204,
+      "step": 30606
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.269232443093334e-06,
+      "loss": 0.6146,
+      "step": 30607
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.268648290719518e-06,
+      "loss": 0.5893,
+      "step": 30608
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.268064167468381e-06,
+      "loss": 0.5972,
+      "step": 30609
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.267480073342889e-06,
+      "loss": 0.6139,
+      "step": 30610
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.26689600834601e-06,
+      "loss": 0.5735,
+      "step": 30611
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.266311972480706e-06,
+      "loss": 0.5709,
+      "step": 30612
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.265727965749956e-06,
+      "loss": 0.5915,
+      "step": 30613
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2651439881567205e-06,
+      "loss": 0.6091,
+      "step": 30614
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.264560039703964e-06,
+      "loss": 0.646,
+      "step": 30615
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.263976120394663e-06,
+      "loss": 0.5663,
+      "step": 30616
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2633922302317775e-06,
+      "loss": 0.5798,
+      "step": 30617
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.262808369218276e-06,
+      "loss": 0.5693,
+      "step": 30618
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.262224537357123e-06,
+      "loss": 0.5391,
+      "step": 30619
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.26164073465129e-06,
+      "loss": 0.6026,
+      "step": 30620
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.261056961103741e-06,
+      "loss": 0.5903,
+      "step": 30621
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.260473216717443e-06,
+      "loss": 0.6039,
+      "step": 30622
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.259889501495358e-06,
+      "loss": 0.5173,
+      "step": 30623
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2593058154404585e-06,
+      "loss": 0.5772,
+      "step": 30624
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.258722158555704e-06,
+      "loss": 0.6395,
+      "step": 30625
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.258138530844067e-06,
+      "loss": 0.5479,
+      "step": 30626
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.25755493230851e-06,
+      "loss": 0.6094,
+      "step": 30627
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.256971362951998e-06,
+      "loss": 0.6324,
+      "step": 30628
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.256387822777492e-06,
+      "loss": 0.5955,
+      "step": 30629
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.255804311787967e-06,
+      "loss": 0.6761,
+      "step": 30630
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.255220829986382e-06,
+      "loss": 0.6242,
+      "step": 30631
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.254637377375702e-06,
+      "loss": 0.6057,
+      "step": 30632
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.254053953958888e-06,
+      "loss": 0.5532,
+      "step": 30633
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.253470559738912e-06,
+      "loss": 0.5326,
+      "step": 30634
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.2528871947187366e-06,
+      "loss": 0.6139,
+      "step": 30635
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.25230385890132e-06,
+      "loss": 0.6521,
+      "step": 30636
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.251720552289634e-06,
+      "loss": 0.5887,
+      "step": 30637
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.251137274886641e-06,
+      "loss": 0.5374,
+      "step": 30638
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.250554026695302e-06,
+      "loss": 0.6031,
+      "step": 30639
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2499708077185775e-06,
+      "loss": 0.6011,
+      "step": 30640
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.24938761795944e-06,
+      "loss": 0.5831,
+      "step": 30641
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2488044574208435e-06,
+      "loss": 0.6524,
+      "step": 30642
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.248221326105764e-06,
+      "loss": 0.6224,
+      "step": 30643
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2476382240171486e-06,
+      "loss": 0.6159,
+      "step": 30644
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.247055151157972e-06,
+      "loss": 0.6145,
+      "step": 30645
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.246472107531188e-06,
+      "loss": 0.5573,
+      "step": 30646
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.24588909313977e-06,
+      "loss": 0.6225,
+      "step": 30647
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.245306107986674e-06,
+      "loss": 0.6761,
+      "step": 30648
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.244723152074864e-06,
+      "loss": 0.6282,
+      "step": 30649
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.244140225407295e-06,
+      "loss": 0.5255,
+      "step": 30650
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.24355732798694e-06,
+      "loss": 0.6451,
+      "step": 30651
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.242974459816756e-06,
+      "loss": 0.5915,
+      "step": 30652
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.242391620899702e-06,
+      "loss": 0.6148,
+      "step": 30653
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.241808811238746e-06,
+      "loss": 0.5835,
+      "step": 30654
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.241226030836845e-06,
+      "loss": 0.5525,
+      "step": 30655
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.240643279696961e-06,
+      "loss": 0.6684,
+      "step": 30656
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2400605578220525e-06,
+      "loss": 0.6424,
+      "step": 30657
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.239477865215086e-06,
+      "loss": 0.6402,
+      "step": 30658
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2388952018790195e-06,
+      "loss": 0.5621,
+      "step": 30659
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2383125678168135e-06,
+      "loss": 0.619,
+      "step": 30660
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.237729963031426e-06,
+      "loss": 0.5919,
+      "step": 30661
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2371473875258225e-06,
+      "loss": 0.6059,
+      "step": 30662
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.236564841302958e-06,
+      "loss": 0.6662,
+      "step": 30663
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.235982324365803e-06,
+      "loss": 0.5983,
+      "step": 30664
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.235399836717302e-06,
+      "loss": 0.6331,
+      "step": 30665
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.234817378360428e-06,
+      "loss": 0.6411,
+      "step": 30666
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.234234949298129e-06,
+      "loss": 0.6089,
+      "step": 30667
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.233652549533376e-06,
+      "loss": 0.5872,
+      "step": 30668
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2330701790691235e-06,
+      "loss": 0.6461,
+      "step": 30669
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.232487837908329e-06,
+      "loss": 0.5744,
+      "step": 30670
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.23190552605395e-06,
+      "loss": 0.6208,
+      "step": 30671
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.231323243508953e-06,
+      "loss": 0.6437,
+      "step": 30672
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.23074099027629e-06,
+      "loss": 0.6166,
+      "step": 30673
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2301587663589195e-06,
+      "loss": 0.5584,
+      "step": 30674
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.229576571759805e-06,
+      "loss": 0.5388,
+      "step": 30675
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.228994406481901e-06,
+      "loss": 0.5773,
+      "step": 30676
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.228412270528168e-06,
+      "loss": 0.558,
+      "step": 30677
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.227830163901559e-06,
+      "loss": 0.6008,
+      "step": 30678
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.227248086605038e-06,
+      "loss": 0.5592,
+      "step": 30679
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.226666038641557e-06,
+      "loss": 0.6124,
+      "step": 30680
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.226084020014084e-06,
+      "loss": 0.6326,
+      "step": 30681
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.225502030725561e-06,
+      "loss": 0.6208,
+      "step": 30682
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.224920070778959e-06,
+      "loss": 0.5686,
+      "step": 30683
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.224338140177223e-06,
+      "loss": 0.5329,
+      "step": 30684
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.223756238923322e-06,
+      "loss": 0.5657,
+      "step": 30685
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.223174367020206e-06,
+      "loss": 0.5436,
+      "step": 30686
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.222592524470835e-06,
+      "loss": 0.5955,
+      "step": 30687
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.222010711278157e-06,
+      "loss": 0.594,
+      "step": 30688
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.221428927445141e-06,
+      "loss": 0.6033,
+      "step": 30689
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.220847172974735e-06,
+      "loss": 0.6318,
+      "step": 30690
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.220265447869893e-06,
+      "loss": 0.6406,
+      "step": 30691
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.219683752133581e-06,
+      "loss": 0.6418,
+      "step": 30692
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2191020857687474e-06,
+      "loss": 0.6101,
+      "step": 30693
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.218520448778348e-06,
+      "loss": 0.584,
+      "step": 30694
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.217938841165336e-06,
+      "loss": 0.5487,
+      "step": 30695
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.217357262932675e-06,
+      "loss": 0.6202,
+      "step": 30696
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.216775714083316e-06,
+      "loss": 0.6375,
+      "step": 30697
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.216194194620211e-06,
+      "loss": 0.5734,
+      "step": 30698
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2156127045463145e-06,
+      "loss": 0.6564,
+      "step": 30699
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.215031243864588e-06,
+      "loss": 0.5894,
+      "step": 30700
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.214449812577978e-06,
+      "loss": 0.5759,
+      "step": 30701
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.21386841068945e-06,
+      "loss": 0.6214,
+      "step": 30702
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.213287038201943e-06,
+      "loss": 0.6043,
+      "step": 30703
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.212705695118424e-06,
+      "loss": 0.6073,
+      "step": 30704
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.212124381441839e-06,
+      "loss": 0.6083,
+      "step": 30705
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.211543097175148e-06,
+      "loss": 0.5856,
+      "step": 30706
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2109618423212985e-06,
+      "loss": 0.5621,
+      "step": 30707
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.210380616883254e-06,
+      "loss": 0.5573,
+      "step": 30708
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.209799420863955e-06,
+      "loss": 0.6087,
+      "step": 30709
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.209218254266363e-06,
+      "loss": 0.5917,
+      "step": 30710
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.208637117093429e-06,
+      "loss": 0.6154,
+      "step": 30711
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.208056009348104e-06,
+      "loss": 0.5924,
+      "step": 30712
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.207474931033345e-06,
+      "loss": 0.5794,
+      "step": 30713
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2068938821521025e-06,
+      "loss": 0.5671,
+      "step": 30714
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.206312862707329e-06,
+      "loss": 0.5434,
+      "step": 30715
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.2057318727019724e-06,
+      "loss": 0.641,
+      "step": 30716
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.205150912138994e-06,
+      "loss": 0.6006,
+      "step": 30717
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.204569981021336e-06,
+      "loss": 0.5911,
+      "step": 30718
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.203989079351962e-06,
+      "loss": 0.6007,
+      "step": 30719
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.203408207133811e-06,
+      "loss": 0.5402,
+      "step": 30720
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.202827364369844e-06,
+      "loss": 0.6501,
+      "step": 30721
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.202246551063005e-06,
+      "loss": 0.5638,
+      "step": 30722
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.201665767216252e-06,
+      "loss": 0.5791,
+      "step": 30723
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.201085012832534e-06,
+      "loss": 0.6027,
+      "step": 30724
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.200504287914801e-06,
+      "loss": 0.6106,
+      "step": 30725
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.199923592466002e-06,
+      "loss": 0.5688,
+      "step": 30726
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.199342926489093e-06,
+      "loss": 0.6198,
+      "step": 30727
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.198762289987017e-06,
+      "loss": 0.5948,
+      "step": 30728
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.198181682962736e-06,
+      "loss": 0.6152,
+      "step": 30729
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.197601105419187e-06,
+      "loss": 0.6365,
+      "step": 30730
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.197020557359328e-06,
+      "loss": 0.5769,
+      "step": 30731
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.196440038786108e-06,
+      "loss": 0.6013,
+      "step": 30732
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.195859549702472e-06,
+      "loss": 0.5967,
+      "step": 30733
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.195279090111377e-06,
+      "loss": 0.627,
+      "step": 30734
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.194698660015769e-06,
+      "loss": 0.588,
+      "step": 30735
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1941182594185965e-06,
+      "loss": 0.5562,
+      "step": 30736
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.193537888322808e-06,
+      "loss": 0.6401,
+      "step": 30737
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.192957546731356e-06,
+      "loss": 0.6872,
+      "step": 30738
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.192377234647183e-06,
+      "loss": 0.6125,
+      "step": 30739
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.191796952073252e-06,
+      "loss": 0.6224,
+      "step": 30740
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.191216699012492e-06,
+      "loss": 0.6202,
+      "step": 30741
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.190636475467865e-06,
+      "loss": 0.6099,
+      "step": 30742
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.190056281442313e-06,
+      "loss": 0.6067,
+      "step": 30743
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.18947611693879e-06,
+      "loss": 0.6402,
+      "step": 30744
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.188895981960236e-06,
+      "loss": 0.618,
+      "step": 30745
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.188315876509611e-06,
+      "loss": 0.6443,
+      "step": 30746
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.187735800589848e-06,
+      "loss": 0.5862,
+      "step": 30747
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.187155754203905e-06,
+      "loss": 0.5479,
+      "step": 30748
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.186575737354727e-06,
+      "loss": 0.6264,
+      "step": 30749
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1859957500452556e-06,
+      "loss": 0.6111,
+      "step": 30750
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1854157922784454e-06,
+      "loss": 0.5509,
+      "step": 30751
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1848358640572414e-06,
+      "loss": 0.6242,
+      "step": 30752
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1842559653845895e-06,
+      "loss": 0.6219,
+      "step": 30753
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.183676096263432e-06,
+      "loss": 0.6347,
+      "step": 30754
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.183096256696723e-06,
+      "loss": 0.5753,
+      "step": 30755
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.182516446687407e-06,
+      "loss": 0.646,
+      "step": 30756
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1819366662384274e-06,
+      "loss": 0.6063,
+      "step": 30757
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.181356915352729e-06,
+      "loss": 0.6257,
+      "step": 30758
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.180777194033263e-06,
+      "loss": 0.5824,
+      "step": 30759
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.180197502282969e-06,
+      "loss": 0.6102,
+      "step": 30760
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.179617840104799e-06,
+      "loss": 0.6283,
+      "step": 30761
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.179038207501695e-06,
+      "loss": 0.5118,
+      "step": 30762
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.178458604476604e-06,
+      "loss": 0.5735,
+      "step": 30763
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.177879031032466e-06,
+      "loss": 0.5344,
+      "step": 30764
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1772994871722325e-06,
+      "loss": 0.6256,
+      "step": 30765
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.176719972898842e-06,
+      "loss": 0.5901,
+      "step": 30766
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.17614048821525e-06,
+      "loss": 0.624,
+      "step": 30767
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.175561033124387e-06,
+      "loss": 0.5533,
+      "step": 30768
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.174981607629207e-06,
+      "loss": 0.5786,
+      "step": 30769
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.174402211732652e-06,
+      "loss": 0.6028,
+      "step": 30770
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1738228454376614e-06,
+      "loss": 0.5655,
+      "step": 30771
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.173243508747188e-06,
+      "loss": 0.6244,
+      "step": 30772
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1726642016641695e-06,
+      "loss": 0.5996,
+      "step": 30773
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.172084924191552e-06,
+      "loss": 0.6272,
+      "step": 30774
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.171505676332273e-06,
+      "loss": 0.5877,
+      "step": 30775
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.170926458089285e-06,
+      "loss": 0.6309,
+      "step": 30776
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1703472694655225e-06,
+      "loss": 0.5735,
+      "step": 30777
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1697681104639394e-06,
+      "loss": 0.6102,
+      "step": 30778
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.169188981087467e-06,
+      "loss": 0.6364,
+      "step": 30779
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.168609881339055e-06,
+      "loss": 0.5505,
+      "step": 30780
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.16803081122164e-06,
+      "loss": 0.5568,
+      "step": 30781
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.167451770738174e-06,
+      "loss": 0.5793,
+      "step": 30782
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.166872759891588e-06,
+      "loss": 0.6484,
+      "step": 30783
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.1662937786848375e-06,
+      "loss": 0.6954,
+      "step": 30784
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.16571482712085e-06,
+      "loss": 0.6099,
+      "step": 30785
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.165135905202578e-06,
+      "loss": 0.6053,
+      "step": 30786
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.164557012932956e-06,
+      "loss": 0.6144,
+      "step": 30787
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.163978150314931e-06,
+      "loss": 0.6252,
+      "step": 30788
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.163399317351443e-06,
+      "loss": 0.6307,
+      "step": 30789
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.162820514045432e-06,
+      "loss": 0.6037,
+      "step": 30790
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.16224174039984e-06,
+      "loss": 0.5837,
+      "step": 30791
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.161662996417604e-06,
+      "loss": 0.5234,
+      "step": 30792
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.161084282101671e-06,
+      "loss": 0.6311,
+      "step": 30793
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.160505597454979e-06,
+      "loss": 0.6033,
+      "step": 30794
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.159926942480469e-06,
+      "loss": 0.5892,
+      "step": 30795
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.159348317181077e-06,
+      "loss": 0.5454,
+      "step": 30796
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.158769721559749e-06,
+      "loss": 0.5348,
+      "step": 30797
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.15819115561942e-06,
+      "loss": 0.5697,
+      "step": 30798
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.157612619363035e-06,
+      "loss": 0.6022,
+      "step": 30799
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.157034112793533e-06,
+      "loss": 0.6249,
+      "step": 30800
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.156455635913851e-06,
+      "loss": 0.5846,
+      "step": 30801
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1558771887269255e-06,
+      "loss": 0.5838,
+      "step": 30802
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1552987712357044e-06,
+      "loss": 0.609,
+      "step": 30803
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.154720383443116e-06,
+      "loss": 0.6147,
+      "step": 30804
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.154142025352115e-06,
+      "loss": 0.588,
+      "step": 30805
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.153563696965621e-06,
+      "loss": 0.6123,
+      "step": 30806
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1529853982865865e-06,
+      "loss": 0.6014,
+      "step": 30807
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1524071293179425e-06,
+      "loss": 0.5629,
+      "step": 30808
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.151828890062633e-06,
+      "loss": 0.5934,
+      "step": 30809
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.151250680523595e-06,
+      "loss": 0.5836,
+      "step": 30810
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.150672500703764e-06,
+      "loss": 0.6553,
+      "step": 30811
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1500943506060795e-06,
+      "loss": 0.5574,
+      "step": 30812
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.149516230233476e-06,
+      "loss": 0.6257,
+      "step": 30813
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.148938139588896e-06,
+      "loss": 0.5847,
+      "step": 30814
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.148360078675272e-06,
+      "loss": 0.557,
+      "step": 30815
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1477820474955525e-06,
+      "loss": 0.6735,
+      "step": 30816
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.147204046052658e-06,
+      "loss": 0.5986,
+      "step": 30817
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.146626074349537e-06,
+      "loss": 0.5767,
+      "step": 30818
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.14604813238912e-06,
+      "loss": 0.588,
+      "step": 30819
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.14547022017435e-06,
+      "loss": 0.6011,
+      "step": 30820
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.144892337708161e-06,
+      "loss": 0.6393,
+      "step": 30821
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.144314484993488e-06,
+      "loss": 0.6045,
+      "step": 30822
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1437366620332656e-06,
+      "loss": 0.6038,
+      "step": 30823
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.143158868830434e-06,
+      "loss": 0.6007,
+      "step": 30824
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.142581105387926e-06,
+      "loss": 0.5616,
+      "step": 30825
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.142003371708682e-06,
+      "loss": 0.6088,
+      "step": 30826
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.141425667795634e-06,
+      "loss": 0.615,
+      "step": 30827
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1408479936517175e-06,
+      "loss": 0.6197,
+      "step": 30828
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1402703492798655e-06,
+      "loss": 0.6736,
+      "step": 30829
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.13969273468302e-06,
+      "loss": 0.6054,
+      "step": 30830
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1391151498641105e-06,
+      "loss": 0.6444,
+      "step": 30831
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.138537594826074e-06,
+      "loss": 0.5588,
+      "step": 30832
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1379600695718446e-06,
+      "loss": 0.5675,
+      "step": 30833
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.137382574104354e-06,
+      "loss": 0.5671,
+      "step": 30834
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.136805108426542e-06,
+      "loss": 0.5913,
+      "step": 30835
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.136227672541338e-06,
+      "loss": 0.5831,
+      "step": 30836
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.135650266451681e-06,
+      "loss": 0.6038,
+      "step": 30837
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.135072890160502e-06,
+      "loss": 0.6465,
+      "step": 30838
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.134495543670737e-06,
+      "loss": 0.6134,
+      "step": 30839
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.133918226985313e-06,
+      "loss": 0.6454,
+      "step": 30840
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.133340940107172e-06,
+      "loss": 0.6303,
+      "step": 30841
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.132763683039241e-06,
+      "loss": 0.5771,
+      "step": 30842
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.132186455784462e-06,
+      "loss": 0.6352,
+      "step": 30843
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.131609258345757e-06,
+      "loss": 0.6545,
+      "step": 30844
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.131032090726066e-06,
+      "loss": 0.4907,
+      "step": 30845
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.130454952928316e-06,
+      "loss": 0.5744,
+      "step": 30846
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.129877844955448e-06,
+      "loss": 0.6089,
+      "step": 30847
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.129300766810389e-06,
+      "loss": 0.622,
+      "step": 30848
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1287237184960714e-06,
+      "loss": 0.5973,
+      "step": 30849
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1281467000154295e-06,
+      "loss": 0.5692,
+      "step": 30850
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.12756971137139e-06,
+      "loss": 0.5755,
+      "step": 30851
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.126992752566892e-06,
+      "loss": 0.6089,
+      "step": 30852
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.12641582360486e-06,
+      "loss": 0.5486,
+      "step": 30853
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.125838924488237e-06,
+      "loss": 0.5867,
+      "step": 30854
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.12526205521994e-06,
+      "loss": 0.6518,
+      "step": 30855
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.124685215802909e-06,
+      "loss": 0.6239,
+      "step": 30856
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1241084062400706e-06,
+      "loss": 0.5992,
+      "step": 30857
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1235316265343615e-06,
+      "loss": 0.5667,
+      "step": 30858
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.122954876688709e-06,
+      "loss": 0.6158,
+      "step": 30859
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.122378156706045e-06,
+      "loss": 0.7265,
+      "step": 30860
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1218014665892945e-06,
+      "loss": 0.6104,
+      "step": 30861
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.121224806341396e-06,
+      "loss": 0.6042,
+      "step": 30862
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.120648175965273e-06,
+      "loss": 0.6192,
+      "step": 30863
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.120071575463861e-06,
+      "loss": 0.6145,
+      "step": 30864
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1194950048400875e-06,
+      "loss": 0.6517,
+      "step": 30865
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1189184640968825e-06,
+      "loss": 0.6298,
+      "step": 30866
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.118341953237172e-06,
+      "loss": 0.5794,
+      "step": 30867
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1177654722638905e-06,
+      "loss": 0.5931,
+      "step": 30868
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.117189021179966e-06,
+      "loss": 0.6192,
+      "step": 30869
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1166125999883275e-06,
+      "loss": 0.6492,
+      "step": 30870
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.116036208691904e-06,
+      "loss": 0.5818,
+      "step": 30871
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.115459847293618e-06,
+      "loss": 0.6187,
+      "step": 30872
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.11488351579641e-06,
+      "loss": 0.5807,
+      "step": 30873
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.114307214203197e-06,
+      "loss": 0.5983,
+      "step": 30874
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.113730942516917e-06,
+      "loss": 0.5953,
+      "step": 30875
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.113154700740493e-06,
+      "loss": 0.6133,
+      "step": 30876
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.112578488876856e-06,
+      "loss": 0.5489,
+      "step": 30877
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.112002306928927e-06,
+      "loss": 0.605,
+      "step": 30878
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.111426154899643e-06,
+      "loss": 0.5745,
+      "step": 30879
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.110850032791922e-06,
+      "loss": 0.5835,
+      "step": 30880
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.110273940608705e-06,
+      "loss": 0.6907,
+      "step": 30881
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.109697878352904e-06,
+      "loss": 0.6485,
+      "step": 30882
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.109121846027457e-06,
+      "loss": 0.5942,
+      "step": 30883
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.108545843635283e-06,
+      "loss": 0.5651,
+      "step": 30884
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.107969871179316e-06,
+      "loss": 0.6235,
+      "step": 30885
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.107393928662481e-06,
+      "loss": 0.5487,
+      "step": 30886
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.106818016087702e-06,
+      "loss": 0.6257,
+      "step": 30887
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1062421334579035e-06,
+      "loss": 0.575,
+      "step": 30888
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1056662807760186e-06,
+      "loss": 0.6177,
+      "step": 30889
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.105090458044969e-06,
+      "loss": 0.5771,
+      "step": 30890
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.104514665267678e-06,
+      "loss": 0.5853,
+      "step": 30891
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1039389024470814e-06,
+      "loss": 0.5706,
+      "step": 30892
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.103363169586091e-06,
+      "loss": 0.636,
+      "step": 30893
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.1027874666876425e-06,
+      "loss": 0.618,
+      "step": 30894
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.102211793754655e-06,
+      "loss": 0.6168,
+      "step": 30895
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.10163615079006e-06,
+      "loss": 0.6031,
+      "step": 30896
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.101060537796779e-06,
+      "loss": 0.6242,
+      "step": 30897
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.100484954777737e-06,
+      "loss": 0.5838,
+      "step": 30898
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.099909401735856e-06,
+      "loss": 0.6126,
+      "step": 30899
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.099333878674066e-06,
+      "loss": 0.6034,
+      "step": 30900
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.098758385595284e-06,
+      "loss": 0.6469,
+      "step": 30901
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.0981829225024435e-06,
+      "loss": 0.6126,
+      "step": 30902
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.097607489398464e-06,
+      "loss": 0.5621,
+      "step": 30903
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.09703208628627e-06,
+      "loss": 0.5945,
+      "step": 30904
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.09645671316878e-06,
+      "loss": 0.5613,
+      "step": 30905
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.095881370048926e-06,
+      "loss": 0.6237,
+      "step": 30906
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.095306056929629e-06,
+      "loss": 0.5919,
+      "step": 30907
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.094730773813811e-06,
+      "loss": 0.6271,
+      "step": 30908
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.094155520704392e-06,
+      "loss": 0.609,
+      "step": 30909
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.0935802976043005e-06,
+      "loss": 0.5773,
+      "step": 30910
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.093005104516458e-06,
+      "loss": 0.5923,
+      "step": 30911
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.092429941443784e-06,
+      "loss": 0.6596,
+      "step": 30912
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.091854808389209e-06,
+      "loss": 0.603,
+      "step": 30913
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.091279705355644e-06,
+      "loss": 0.5688,
+      "step": 30914
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.090704632346021e-06,
+      "loss": 0.6212,
+      "step": 30915
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.090129589363254e-06,
+      "loss": 0.5833,
+      "step": 30916
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.089554576410273e-06,
+      "loss": 0.5798,
+      "step": 30917
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.0889795934899934e-06,
+      "loss": 0.5422,
+      "step": 30918
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.088404640605346e-06,
+      "loss": 0.5906,
+      "step": 30919
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.087829717759238e-06,
+      "loss": 0.6421,
+      "step": 30920
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.087254824954604e-06,
+      "loss": 0.6195,
+      "step": 30921
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.086679962194355e-06,
+      "loss": 0.6238,
+      "step": 30922
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.086105129481421e-06,
+      "loss": 0.6457,
+      "step": 30923
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.085530326818719e-06,
+      "loss": 0.5893,
+      "step": 30924
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.08495555420917e-06,
+      "loss": 0.597,
+      "step": 30925
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.084380811655689e-06,
+      "loss": 0.6101,
+      "step": 30926
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.0838060991612064e-06,
+      "loss": 0.6173,
+      "step": 30927
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.083231416728637e-06,
+      "loss": 0.5596,
+      "step": 30928
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.082656764360898e-06,
+      "loss": 0.6788,
+      "step": 30929
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.082082142060921e-06,
+      "loss": 0.5883,
+      "step": 30930
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.08150754983161e-06,
+      "loss": 0.6464,
+      "step": 30931
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.080932987675896e-06,
+      "loss": 0.5089,
+      "step": 30932
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.080358455596691e-06,
+      "loss": 0.6575,
+      "step": 30933
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.079783953596923e-06,
+      "loss": 0.607,
+      "step": 30934
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.079209481679506e-06,
+      "loss": 0.6433,
+      "step": 30935
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.07863503984736e-06,
+      "loss": 0.5533,
+      "step": 30936
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.078060628103399e-06,
+      "loss": 0.5472,
+      "step": 30937
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.077486246450549e-06,
+      "loss": 0.6147,
+      "step": 30938
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.076911894891724e-06,
+      "loss": 0.5818,
+      "step": 30939
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0763375734298515e-06,
+      "loss": 0.5787,
+      "step": 30940
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.075763282067835e-06,
+      "loss": 0.6311,
+      "step": 30941
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.075189020808604e-06,
+      "loss": 0.6242,
+      "step": 30942
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.074614789655069e-06,
+      "loss": 0.6812,
+      "step": 30943
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.074040588610156e-06,
+      "loss": 0.5626,
+      "step": 30944
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0734664176767765e-06,
+      "loss": 0.6133,
+      "step": 30945
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.07289227685785e-06,
+      "loss": 0.5635,
+      "step": 30946
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.072318166156292e-06,
+      "loss": 0.6966,
+      "step": 30947
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.071744085575024e-06,
+      "loss": 0.6545,
+      "step": 30948
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.071170035116959e-06,
+      "loss": 0.5706,
+      "step": 30949
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.070596014785013e-06,
+      "loss": 0.5855,
+      "step": 30950
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0700220245821135e-06,
+      "loss": 0.6302,
+      "step": 30951
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.06944806451116e-06,
+      "loss": 0.6188,
+      "step": 30952
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.068874134575082e-06,
+      "loss": 0.631,
+      "step": 30953
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.068300234776787e-06,
+      "loss": 0.5956,
+      "step": 30954
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0677263651192e-06,
+      "loss": 0.6056,
+      "step": 30955
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.06715252560523e-06,
+      "loss": 0.6316,
+      "step": 30956
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0665787162378016e-06,
+      "loss": 0.6379,
+      "step": 30957
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.066004937019817e-06,
+      "loss": 0.6078,
+      "step": 30958
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0654311879542044e-06,
+      "loss": 0.5572,
+      "step": 30959
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.064857469043868e-06,
+      "loss": 0.5746,
+      "step": 30960
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0642837802917345e-06,
+      "loss": 0.6396,
+      "step": 30961
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.063710121700713e-06,
+      "loss": 0.5359,
+      "step": 30962
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.063136493273718e-06,
+      "loss": 0.6443,
+      "step": 30963
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.062562895013662e-06,
+      "loss": 0.5955,
+      "step": 30964
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0619893269234665e-06,
+      "loss": 0.6079,
+      "step": 30965
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.061415789006042e-06,
+      "loss": 0.6314,
+      "step": 30966
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.060842281264301e-06,
+      "loss": 0.589,
+      "step": 30967
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0602688037011615e-06,
+      "loss": 0.6312,
+      "step": 30968
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.059695356319536e-06,
+      "loss": 0.5939,
+      "step": 30969
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.059121939122339e-06,
+      "loss": 0.6382,
+      "step": 30970
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.058548552112479e-06,
+      "loss": 0.5565,
+      "step": 30971
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0579751952928784e-06,
+      "loss": 0.6594,
+      "step": 30972
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.057401868666445e-06,
+      "loss": 0.6038,
+      "step": 30973
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.056828572236093e-06,
+      "loss": 0.5925,
+      "step": 30974
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.056255306004733e-06,
+      "loss": 0.6245,
+      "step": 30975
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.055682069975283e-06,
+      "loss": 0.5782,
+      "step": 30976
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.055108864150649e-06,
+      "loss": 0.5769,
+      "step": 30977
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0545356885337576e-06,
+      "loss": 0.6715,
+      "step": 30978
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.053962543127503e-06,
+      "loss": 0.6255,
+      "step": 30979
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.05338942793481e-06,
+      "loss": 0.5981,
+      "step": 30980
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.052816342958582e-06,
+      "loss": 0.5484,
+      "step": 30981
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.052243288201741e-06,
+      "loss": 0.6124,
+      "step": 30982
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0516702636671925e-06,
+      "loss": 0.6058,
+      "step": 30983
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.05109726935785e-06,
+      "loss": 0.5939,
+      "step": 30984
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0505243052766215e-06,
+      "loss": 0.6262,
+      "step": 30985
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0499513714264236e-06,
+      "loss": 0.6588,
+      "step": 30986
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.049378467810166e-06,
+      "loss": 0.6034,
+      "step": 30987
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.048805594430757e-06,
+      "loss": 0.6095,
+      "step": 30988
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.048232751291111e-06,
+      "loss": 0.6444,
+      "step": 30989
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.047659938394139e-06,
+      "loss": 0.568,
+      "step": 30990
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0470871557427495e-06,
+      "loss": 0.585,
+      "step": 30991
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.04651440333985e-06,
+      "loss": 0.5679,
+      "step": 30992
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.045941681188358e-06,
+      "loss": 0.6415,
+      "step": 30993
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.045368989291176e-06,
+      "loss": 0.6134,
+      "step": 30994
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.044796327651226e-06,
+      "loss": 0.6019,
+      "step": 30995
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.044223696271402e-06,
+      "loss": 0.6846,
+      "step": 30996
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.043651095154626e-06,
+      "loss": 0.5782,
+      "step": 30997
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0430785243037984e-06,
+      "loss": 0.5333,
+      "step": 30998
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.042505983721838e-06,
+      "loss": 0.6072,
+      "step": 30999
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.04193347341165e-06,
+      "loss": 0.6562,
+      "step": 31000
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.041360993376141e-06,
+      "loss": 0.5896,
+      "step": 31001
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0407885436182205e-06,
+      "loss": 0.6441,
+      "step": 31002
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.040216124140801e-06,
+      "loss": 0.6147,
+      "step": 31003
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.039643734946789e-06,
+      "loss": 0.5634,
+      "step": 31004
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.039071376039092e-06,
+      "loss": 0.5825,
+      "step": 31005
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.038499047420616e-06,
+      "loss": 0.6097,
+      "step": 31006
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.037926749094277e-06,
+      "loss": 0.6074,
+      "step": 31007
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.037354481062976e-06,
+      "loss": 0.5454,
+      "step": 31008
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.036782243329621e-06,
+      "loss": 0.632,
+      "step": 31009
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.036210035897125e-06,
+      "loss": 0.6117,
+      "step": 31010
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.035637858768393e-06,
+      "loss": 0.6243,
+      "step": 31011
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.035065711946331e-06,
+      "loss": 0.5365,
+      "step": 31012
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.034493595433844e-06,
+      "loss": 0.6748,
+      "step": 31013
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.033921509233846e-06,
+      "loss": 0.5301,
+      "step": 31014
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.033349453349236e-06,
+      "loss": 0.5354,
+      "step": 31015
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.032777427782931e-06,
+      "loss": 0.6199,
+      "step": 31016
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.032205432537826e-06,
+      "loss": 0.6192,
+      "step": 31017
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.031633467616836e-06,
+      "loss": 0.6433,
+      "step": 31018
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.03106153302286e-06,
+      "loss": 0.6264,
+      "step": 31019
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.030489628758813e-06,
+      "loss": 0.6063,
+      "step": 31020
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.029917754827592e-06,
+      "loss": 0.6017,
+      "step": 31021
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.029345911232114e-06,
+      "loss": 0.56,
+      "step": 31022
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.028774097975272e-06,
+      "loss": 0.6059,
+      "step": 31023
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.028202315059982e-06,
+      "loss": 0.6265,
+      "step": 31024
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.027630562489143e-06,
+      "loss": 0.6633,
+      "step": 31025
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.02705884026566e-06,
+      "loss": 0.5938,
+      "step": 31026
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.026487148392443e-06,
+      "loss": 0.6078,
+      "step": 31027
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.025915486872395e-06,
+      "loss": 0.6169,
+      "step": 31028
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.025343855708419e-06,
+      "loss": 0.6078,
+      "step": 31029
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.024772254903418e-06,
+      "loss": 0.6367,
+      "step": 31030
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.024200684460302e-06,
+      "loss": 0.6539,
+      "step": 31031
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.023629144381972e-06,
+      "loss": 0.624,
+      "step": 31032
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.023057634671334e-06,
+      "loss": 0.5757,
+      "step": 31033
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0224861553312855e-06,
+      "loss": 0.6241,
+      "step": 31034
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.02191470636474e-06,
+      "loss": 0.5825,
+      "step": 31035
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.021343287774593e-06,
+      "loss": 0.6297,
+      "step": 31036
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.020771899563755e-06,
+      "loss": 0.6563,
+      "step": 31037
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0202005417351266e-06,
+      "loss": 0.5911,
+      "step": 31038
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.01962921429161e-06,
+      "loss": 0.6114,
+      "step": 31039
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.019057917236105e-06,
+      "loss": 0.5953,
+      "step": 31040
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.018486650571522e-06,
+      "loss": 0.5975,
+      "step": 31041
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0179154143007606e-06,
+      "loss": 0.5533,
+      "step": 31042
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.017344208426722e-06,
+      "loss": 0.5935,
+      "step": 31043
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.016773032952307e-06,
+      "loss": 0.5805,
+      "step": 31044
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.016201887880424e-06,
+      "loss": 0.5934,
+      "step": 31045
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.015630773213971e-06,
+      "loss": 0.5696,
+      "step": 31046
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.015059688955848e-06,
+      "loss": 0.649,
+      "step": 31047
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.014488635108962e-06,
+      "loss": 0.6076,
+      "step": 31048
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.013917611676212e-06,
+      "loss": 0.6007,
+      "step": 31049
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.013346618660501e-06,
+      "loss": 0.6491,
+      "step": 31050
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.012775656064725e-06,
+      "loss": 0.6627,
+      "step": 31051
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0122047238917925e-06,
+      "loss": 0.6251,
+      "step": 31052
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.011633822144599e-06,
+      "loss": 0.6164,
+      "step": 31053
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.011062950826054e-06,
+      "loss": 0.6208,
+      "step": 31054
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0104921099390445e-06,
+      "loss": 0.6215,
+      "step": 31055
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.009921299486483e-06,
+      "loss": 0.5927,
+      "step": 31056
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.009350519471262e-06,
+      "loss": 0.6209,
+      "step": 31057
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.008779769896288e-06,
+      "loss": 0.6458,
+      "step": 31058
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.008209050764457e-06,
+      "loss": 0.6208,
+      "step": 31059
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.007638362078676e-06,
+      "loss": 0.6098,
+      "step": 31060
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.007067703841833e-06,
+      "loss": 0.6071,
+      "step": 31061
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.006497076056838e-06,
+      "loss": 0.6088,
+      "step": 31062
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.005926478726586e-06,
+      "loss": 0.6262,
+      "step": 31063
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.005355911853974e-06,
+      "loss": 0.6155,
+      "step": 31064
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0047853754419074e-06,
+      "loss": 0.5925,
+      "step": 31065
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0042148694932825e-06,
+      "loss": 0.5796,
+      "step": 31066
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.003644394010997e-06,
+      "loss": 0.4902,
+      "step": 31067
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.003073948997948e-06,
+      "loss": 0.6171,
+      "step": 31068
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.002503534457039e-06,
+      "loss": 0.591,
+      "step": 31069
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.0019331503911665e-06,
+      "loss": 0.6431,
+      "step": 31070
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.001362796803228e-06,
+      "loss": 0.5336,
+      "step": 31071
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.000792473696119e-06,
+      "loss": 0.5704,
+      "step": 31072
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.000222181072743e-06,
+      "loss": 0.5807,
+      "step": 31073
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.999651918935993e-06,
+      "loss": 0.6107,
+      "step": 31074
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.999081687288772e-06,
+      "loss": 0.6154,
+      "step": 31075
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.998511486133974e-06,
+      "loss": 0.5766,
+      "step": 31076
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.997941315474496e-06,
+      "loss": 0.6372,
+      "step": 31077
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.997371175313234e-06,
+      "loss": 0.6198,
+      "step": 31078
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.99680106565309e-06,
+      "loss": 0.5855,
+      "step": 31079
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.996230986496954e-06,
+      "loss": 0.5465,
+      "step": 31080
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.995660937847733e-06,
+      "loss": 0.606,
+      "step": 31081
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.995090919708311e-06,
+      "loss": 0.5796,
+      "step": 31082
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.994520932081594e-06,
+      "loss": 0.5529,
+      "step": 31083
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.993950974970477e-06,
+      "loss": 0.6297,
+      "step": 31084
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.993381048377848e-06,
+      "loss": 0.6178,
+      "step": 31085
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.992811152306612e-06,
+      "loss": 0.5761,
+      "step": 31086
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.992241286759664e-06,
+      "loss": 0.5824,
+      "step": 31087
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.991671451739896e-06,
+      "loss": 0.5624,
+      "step": 31088
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.991101647250202e-06,
+      "loss": 0.6358,
+      "step": 31089
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.990531873293484e-06,
+      "loss": 0.6556,
+      "step": 31090
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9899621298726295e-06,
+      "loss": 0.6252,
+      "step": 31091
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.989392416990545e-06,
+      "loss": 0.5697,
+      "step": 31092
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.988822734650111e-06,
+      "loss": 0.5832,
+      "step": 31093
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.988253082854232e-06,
+      "loss": 0.6123,
+      "step": 31094
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.987683461605797e-06,
+      "loss": 0.5246,
+      "step": 31095
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.987113870907706e-06,
+      "loss": 0.5895,
+      "step": 31096
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.986544310762851e-06,
+      "loss": 0.6187,
+      "step": 31097
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.985974781174125e-06,
+      "loss": 0.6017,
+      "step": 31098
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.985405282144419e-06,
+      "loss": 0.5549,
+      "step": 31099
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.984835813676633e-06,
+      "loss": 0.5563,
+      "step": 31100
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.984266375773655e-06,
+      "loss": 0.5621,
+      "step": 31101
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.983696968438385e-06,
+      "loss": 0.6101,
+      "step": 31102
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.983127591673713e-06,
+      "loss": 0.6027,
+      "step": 31103
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.982558245482531e-06,
+      "loss": 0.585,
+      "step": 31104
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.981988929867734e-06,
+      "loss": 0.6271,
+      "step": 31105
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.981419644832208e-06,
+      "loss": 0.6068,
+      "step": 31106
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.980850390378858e-06,
+      "loss": 0.6141,
+      "step": 31107
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.980281166510568e-06,
+      "loss": 0.6709,
+      "step": 31108
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.979711973230232e-06,
+      "loss": 0.632,
+      "step": 31109
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9791428105407406e-06,
+      "loss": 0.6288,
+      "step": 31110
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.97857367844499e-06,
+      "loss": 0.5877,
+      "step": 31111
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.978004576945867e-06,
+      "loss": 0.5465,
+      "step": 31112
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.97743550604627e-06,
+      "loss": 0.6101,
+      "step": 31113
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.976866465749086e-06,
+      "loss": 0.6168,
+      "step": 31114
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.976297456057209e-06,
+      "loss": 0.6256,
+      "step": 31115
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.975728476973524e-06,
+      "loss": 0.5829,
+      "step": 31116
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.975159528500931e-06,
+      "loss": 0.6785,
+      "step": 31117
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.974590610642314e-06,
+      "loss": 0.5879,
+      "step": 31118
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.974021723400573e-06,
+      "loss": 0.6696,
+      "step": 31119
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.973452866778585e-06,
+      "loss": 0.6079,
+      "step": 31120
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.972884040779253e-06,
+      "loss": 0.5641,
+      "step": 31121
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9723152454054605e-06,
+      "loss": 0.636,
+      "step": 31122
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.971746480660097e-06,
+      "loss": 0.5804,
+      "step": 31123
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.971177746546057e-06,
+      "loss": 0.6206,
+      "step": 31124
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.97060904306623e-06,
+      "loss": 0.6142,
+      "step": 31125
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.970040370223505e-06,
+      "loss": 0.5792,
+      "step": 31126
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9694717280207665e-06,
+      "loss": 0.6118,
+      "step": 31127
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.968903116460911e-06,
+      "loss": 0.5287,
+      "step": 31128
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.968334535546823e-06,
+      "loss": 0.5989,
+      "step": 31129
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9677659852814e-06,
+      "loss": 0.6423,
+      "step": 31130
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.967197465667517e-06,
+      "loss": 0.6482,
+      "step": 31131
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.966628976708074e-06,
+      "loss": 0.551,
+      "step": 31132
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.966060518405954e-06,
+      "loss": 0.6122,
+      "step": 31133
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.96549209076405e-06,
+      "loss": 0.65,
+      "step": 31134
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.964923693785249e-06,
+      "loss": 0.6083,
+      "step": 31135
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.964355327472437e-06,
+      "loss": 0.6426,
+      "step": 31136
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9637869918285e-06,
+      "loss": 0.5884,
+      "step": 31137
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.963218686856332e-06,
+      "loss": 0.5881,
+      "step": 31138
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.962650412558816e-06,
+      "loss": 0.5863,
+      "step": 31139
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.962082168938843e-06,
+      "loss": 0.5581,
+      "step": 31140
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9615139559993e-06,
+      "loss": 0.5829,
+      "step": 31141
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.960945773743074e-06,
+      "loss": 0.5434,
+      "step": 31142
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.960377622173049e-06,
+      "loss": 0.6484,
+      "step": 31143
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.959809501292111e-06,
+      "loss": 0.5934,
+      "step": 31144
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.959241411103154e-06,
+      "loss": 0.5756,
+      "step": 31145
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.95867335160906e-06,
+      "loss": 0.5903,
+      "step": 31146
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.958105322812716e-06,
+      "loss": 0.5495,
+      "step": 31147
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.957537324717006e-06,
+      "loss": 0.5303,
+      "step": 31148
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.956969357324821e-06,
+      "loss": 0.5664,
+      "step": 31149
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.956401420639041e-06,
+      "loss": 0.5897,
+      "step": 31150
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.955833514662559e-06,
+      "loss": 0.5823,
+      "step": 31151
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.955265639398257e-06,
+      "loss": 0.541,
+      "step": 31152
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.954697794849021e-06,
+      "loss": 0.6164,
+      "step": 31153
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.954129981017731e-06,
+      "loss": 0.6082,
+      "step": 31154
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.953562197907282e-06,
+      "loss": 0.5337,
+      "step": 31155
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.95299444552055e-06,
+      "loss": 0.6405,
+      "step": 31156
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.952426723860432e-06,
+      "loss": 0.6407,
+      "step": 31157
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.951859032929798e-06,
+      "loss": 0.5973,
+      "step": 31158
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.951291372731542e-06,
+      "loss": 0.6147,
+      "step": 31159
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.950723743268543e-06,
+      "loss": 0.6168,
+      "step": 31160
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.950156144543692e-06,
+      "loss": 0.5882,
+      "step": 31161
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.949588576559869e-06,
+      "loss": 0.5803,
+      "step": 31162
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.949021039319959e-06,
+      "loss": 0.5929,
+      "step": 31163
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.948453532826845e-06,
+      "loss": 0.6765,
+      "step": 31164
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.947886057083406e-06,
+      "loss": 0.5971,
+      "step": 31165
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9473186120925354e-06,
+      "loss": 0.5551,
+      "step": 31166
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.946751197857108e-06,
+      "loss": 0.6073,
+      "step": 31167
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.946183814380017e-06,
+      "loss": 0.5408,
+      "step": 31168
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9456164616641326e-06,
+      "loss": 0.5185,
+      "step": 31169
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.945049139712347e-06,
+      "loss": 0.5816,
+      "step": 31170
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.944481848527536e-06,
+      "loss": 0.6096,
+      "step": 31171
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.943914588112591e-06,
+      "loss": 0.5718,
+      "step": 31172
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.943347358470389e-06,
+      "loss": 0.6116,
+      "step": 31173
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.942780159603813e-06,
+      "loss": 0.5723,
+      "step": 31174
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.94221299151574e-06,
+      "loss": 0.6024,
+      "step": 31175
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.941645854209063e-06,
+      "loss": 0.5811,
+      "step": 31176
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.941078747686652e-06,
+      "loss": 0.5721,
+      "step": 31177
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9405116719513996e-06,
+      "loss": 0.6276,
+      "step": 31178
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.93994462700618e-06,
+      "loss": 0.5533,
+      "step": 31179
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9393776128538774e-06,
+      "loss": 0.5899,
+      "step": 31180
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.938810629497369e-06,
+      "loss": 0.5575,
+      "step": 31181
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.938243676939543e-06,
+      "loss": 0.5589,
+      "step": 31182
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.937676755183275e-06,
+      "loss": 0.6232,
+      "step": 31183
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.937109864231447e-06,
+      "loss": 0.5678,
+      "step": 31184
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.936543004086939e-06,
+      "loss": 0.5602,
+      "step": 31185
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.935976174752629e-06,
+      "loss": 0.6051,
+      "step": 31186
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.935409376231402e-06,
+      "loss": 0.5949,
+      "step": 31187
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.934842608526135e-06,
+      "loss": 0.6282,
+      "step": 31188
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.934275871639715e-06,
+      "loss": 0.6101,
+      "step": 31189
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9337091655750086e-06,
+      "loss": 0.6246,
+      "step": 31190
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.933142490334906e-06,
+      "loss": 0.5902,
+      "step": 31191
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.93257584592228e-06,
+      "loss": 0.5812,
+      "step": 31192
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.932009232340015e-06,
+      "loss": 0.605,
+      "step": 31193
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.931442649590986e-06,
+      "loss": 0.6709,
+      "step": 31194
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.930876097678081e-06,
+      "loss": 0.5966,
+      "step": 31195
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.930309576604165e-06,
+      "loss": 0.6714,
+      "step": 31196
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9297430863721275e-06,
+      "loss": 0.6306,
+      "step": 31197
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9291766269848386e-06,
+      "loss": 0.6209,
+      "step": 31198
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.928610198445185e-06,
+      "loss": 0.605,
+      "step": 31199
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.928043800756042e-06,
+      "loss": 0.6226,
+      "step": 31200
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.927477433920286e-06,
+      "loss": 0.5673,
+      "step": 31201
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.926911097940792e-06,
+      "loss": 0.614,
+      "step": 31202
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9263447928204444e-06,
+      "loss": 0.6632,
+      "step": 31203
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.925778518562119e-06,
+      "loss": 0.6119,
+      "step": 31204
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.925212275168687e-06,
+      "loss": 0.6075,
+      "step": 31205
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.924646062643038e-06,
+      "loss": 0.5832,
+      "step": 31206
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.924079880988035e-06,
+      "loss": 0.6112,
+      "step": 31207
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.923513730206565e-06,
+      "loss": 0.5826,
+      "step": 31208
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.922947610301497e-06,
+      "loss": 0.5863,
+      "step": 31209
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.922381521275717e-06,
+      "loss": 0.6062,
+      "step": 31210
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.921815463132095e-06,
+      "loss": 0.6107,
+      "step": 31211
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.921249435873508e-06,
+      "loss": 0.5471,
+      "step": 31212
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.920683439502831e-06,
+      "loss": 0.5838,
+      "step": 31213
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.920117474022945e-06,
+      "loss": 0.5596,
+      "step": 31214
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9195515394367185e-06,
+      "loss": 0.6862,
+      "step": 31215
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9189856357470405e-06,
+      "loss": 0.6362,
+      "step": 31216
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.918419762956768e-06,
+      "loss": 0.6111,
+      "step": 31217
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.91785392106879e-06,
+      "loss": 0.5994,
+      "step": 31218
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.917288110085974e-06,
+      "loss": 0.5993,
+      "step": 31219
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.9167223300112025e-06,
+      "loss": 0.5996,
+      "step": 31220
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.916156580847346e-06,
+      "loss": 0.5837,
+      "step": 31221
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.91559086259728e-06,
+      "loss": 0.667,
+      "step": 31222
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.915025175263877e-06,
+      "loss": 0.6168,
+      "step": 31223
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.914459518850011e-06,
+      "loss": 0.6036,
+      "step": 31224
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.913893893358562e-06,
+      "loss": 0.5879,
+      "step": 31225
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.913328298792397e-06,
+      "loss": 0.588,
+      "step": 31226
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.912762735154401e-06,
+      "loss": 0.5387,
+      "step": 31227
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.912197202447432e-06,
+      "loss": 0.5934,
+      "step": 31228
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9116317006743765e-06,
+      "loss": 0.6498,
+      "step": 31229
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.911066229838098e-06,
+      "loss": 0.6338,
+      "step": 31230
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.910500789941482e-06,
+      "loss": 0.5657,
+      "step": 31231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9099353809873885e-06,
+      "loss": 0.5488,
+      "step": 31232
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9093700029787055e-06,
+      "loss": 0.616,
+      "step": 31233
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.908804655918289e-06,
+      "loss": 0.6098,
+      "step": 31234
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.908239339809025e-06,
+      "loss": 0.6286,
+      "step": 31235
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9076740546537764e-06,
+      "loss": 0.592,
+      "step": 31236
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.907108800455423e-06,
+      "loss": 0.58,
+      "step": 31237
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.906543577216836e-06,
+      "loss": 0.6358,
+      "step": 31238
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.905978384940884e-06,
+      "loss": 0.6232,
+      "step": 31239
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9054132236304374e-06,
+      "loss": 0.581,
+      "step": 31240
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.904848093288375e-06,
+      "loss": 0.6316,
+      "step": 31241
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.904282993917564e-06,
+      "loss": 0.6492,
+      "step": 31242
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.903717925520877e-06,
+      "loss": 0.6047,
+      "step": 31243
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.903152888101185e-06,
+      "loss": 0.5502,
+      "step": 31244
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.902587881661354e-06,
+      "loss": 0.5689,
+      "step": 31245
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.902022906204264e-06,
+      "loss": 0.5346,
+      "step": 31246
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.901457961732778e-06,
+      "loss": 0.5822,
+      "step": 31247
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9008930482497735e-06,
+      "loss": 0.6614,
+      "step": 31248
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.9003281657581175e-06,
+      "loss": 0.6224,
+      "step": 31249
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8997633142606795e-06,
+      "loss": 0.5985,
+      "step": 31250
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8991984937603275e-06,
+      "loss": 0.6211,
+      "step": 31251
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.89863370425994e-06,
+      "loss": 0.5805,
+      "step": 31252
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.898068945762376e-06,
+      "loss": 0.6218,
+      "step": 31253
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.897504218270518e-06,
+      "loss": 0.5552,
+      "step": 31254
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.896939521787222e-06,
+      "loss": 0.596,
+      "step": 31255
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.896374856315367e-06,
+      "loss": 0.6046,
+      "step": 31256
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.895810221857815e-06,
+      "loss": 0.6301,
+      "step": 31257
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.895245618417443e-06,
+      "loss": 0.5319,
+      "step": 31258
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.894681045997115e-06,
+      "loss": 0.5766,
+      "step": 31259
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.894116504599702e-06,
+      "loss": 0.5941,
+      "step": 31260
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.893551994228067e-06,
+      "loss": 0.6515,
+      "step": 31261
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.892987514885086e-06,
+      "loss": 0.6476,
+      "step": 31262
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.892423066573624e-06,
+      "loss": 0.5531,
+      "step": 31263
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.891858649296545e-06,
+      "loss": 0.6578,
+      "step": 31264
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.891294263056729e-06,
+      "loss": 0.6632,
+      "step": 31265
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.890729907857028e-06,
+      "loss": 0.658,
+      "step": 31266
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.890165583700322e-06,
+      "loss": 0.643,
+      "step": 31267
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.88960129058947e-06,
+      "loss": 0.5697,
+      "step": 31268
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.889037028527347e-06,
+      "loss": 0.6266,
+      "step": 31269
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.888472797516813e-06,
+      "loss": 0.6143,
+      "step": 31270
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.887908597560745e-06,
+      "loss": 0.6057,
+      "step": 31271
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.887344428661998e-06,
+      "loss": 0.5689,
+      "step": 31272
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8867802908234465e-06,
+      "loss": 0.5903,
+      "step": 31273
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.886216184047951e-06,
+      "loss": 0.6562,
+      "step": 31274
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.885652108338386e-06,
+      "loss": 0.6554,
+      "step": 31275
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.885088063697612e-06,
+      "loss": 0.5592,
+      "step": 31276
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.884524050128498e-06,
+      "loss": 0.6412,
+      "step": 31277
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.883960067633903e-06,
+      "loss": 0.597,
+      "step": 31278
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.883396116216702e-06,
+      "loss": 0.5461,
+      "step": 31279
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8828321958797576e-06,
+      "loss": 0.6342,
+      "step": 31280
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.882268306625934e-06,
+      "loss": 0.6122,
+      "step": 31281
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.881704448458092e-06,
+      "loss": 0.6359,
+      "step": 31282
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.881140621379107e-06,
+      "loss": 0.6396,
+      "step": 31283
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.880576825391837e-06,
+      "loss": 0.595,
+      "step": 31284
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.880013060499146e-06,
+      "loss": 0.5774,
+      "step": 31285
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.879449326703903e-06,
+      "loss": 0.6696,
+      "step": 31286
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.878885624008972e-06,
+      "loss": 0.6217,
+      "step": 31287
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.878321952417215e-06,
+      "loss": 0.5691,
+      "step": 31288
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8777583119314944e-06,
+      "loss": 0.5638,
+      "step": 31289
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.877194702554678e-06,
+      "loss": 0.5843,
+      "step": 31290
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.876631124289627e-06,
+      "loss": 0.5927,
+      "step": 31291
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.876067577139213e-06,
+      "loss": 0.6101,
+      "step": 31292
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.875504061106286e-06,
+      "loss": 0.6306,
+      "step": 31293
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.874940576193721e-06,
+      "loss": 0.6073,
+      "step": 31294
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.874377122404372e-06,
+      "loss": 0.5799,
+      "step": 31295
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.873813699741111e-06,
+      "loss": 0.5723,
+      "step": 31296
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.873250308206796e-06,
+      "loss": 0.6317,
+      "step": 31297
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.872686947804292e-06,
+      "loss": 0.5563,
+      "step": 31298
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.872123618536455e-06,
+      "loss": 0.6222,
+      "step": 31299
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.871560320406156e-06,
+      "loss": 0.569,
+      "step": 31300
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.870997053416254e-06,
+      "loss": 0.5999,
+      "step": 31301
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.870433817569609e-06,
+      "loss": 0.5939,
+      "step": 31302
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.869870612869087e-06,
+      "loss": 0.5726,
+      "step": 31303
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8693074393175475e-06,
+      "loss": 0.5774,
+      "step": 31304
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.868744296917852e-06,
+      "loss": 0.581,
+      "step": 31305
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8681811856728604e-06,
+      "loss": 0.596,
+      "step": 31306
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.867618105585439e-06,
+      "loss": 0.6085,
+      "step": 31307
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8670550566584455e-06,
+      "loss": 0.6285,
+      "step": 31308
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8664920388947415e-06,
+      "loss": 0.6165,
+      "step": 31309
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.865929052297184e-06,
+      "loss": 0.6303,
+      "step": 31310
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.865366096868641e-06,
+      "loss": 0.5846,
+      "step": 31311
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.864803172611966e-06,
+      "loss": 0.6287,
+      "step": 31312
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.864240279530027e-06,
+      "loss": 0.6231,
+      "step": 31313
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.86367741762568e-06,
+      "loss": 0.6204,
+      "step": 31314
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.863114586901784e-06,
+      "loss": 0.5528,
+      "step": 31315
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8625517873611975e-06,
+      "loss": 0.5883,
+      "step": 31316
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.861989019006785e-06,
+      "loss": 0.6705,
+      "step": 31317
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.861426281841406e-06,
+      "loss": 0.5636,
+      "step": 31318
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.860863575867917e-06,
+      "loss": 0.6066,
+      "step": 31319
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.860300901089175e-06,
+      "loss": 0.5609,
+      "step": 31320
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.859738257508044e-06,
+      "loss": 0.5659,
+      "step": 31321
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.859175645127382e-06,
+      "loss": 0.614,
+      "step": 31322
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.858613063950044e-06,
+      "loss": 0.6097,
+      "step": 31323
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.858050513978895e-06,
+      "loss": 0.5867,
+      "step": 31324
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.85748799521679e-06,
+      "loss": 0.5345,
+      "step": 31325
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8569255076665876e-06,
+      "loss": 0.5684,
+      "step": 31326
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.856363051331142e-06,
+      "loss": 0.5487,
+      "step": 31327
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.855800626213318e-06,
+      "loss": 0.5803,
+      "step": 31328
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.855238232315968e-06,
+      "loss": 0.6115,
+      "step": 31329
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.854675869641959e-06,
+      "loss": 0.5494,
+      "step": 31330
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.854113538194134e-06,
+      "loss": 0.5148,
+      "step": 31331
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.853551237975362e-06,
+      "loss": 0.57,
+      "step": 31332
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.852988968988492e-06,
+      "loss": 0.6108,
+      "step": 31333
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.852426731236389e-06,
+      "loss": 0.6205,
+      "step": 31334
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.851864524721907e-06,
+      "loss": 0.6497,
+      "step": 31335
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.851302349447901e-06,
+      "loss": 0.5643,
+      "step": 31336
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.850740205417225e-06,
+      "loss": 0.5783,
+      "step": 31337
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8501780926327435e-06,
+      "loss": 0.582,
+      "step": 31338
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.849616011097307e-06,
+      "loss": 0.6065,
+      "step": 31339
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.84905396081377e-06,
+      "loss": 0.6452,
+      "step": 31340
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8484919417849945e-06,
+      "loss": 0.6013,
+      "step": 31341
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.847929954013833e-06,
+      "loss": 0.6542,
+      "step": 31342
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.847367997503141e-06,
+      "loss": 0.5579,
+      "step": 31343
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.84680607225577e-06,
+      "loss": 0.6442,
+      "step": 31344
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.846244178274584e-06,
+      "loss": 0.6027,
+      "step": 31345
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.845682315562432e-06,
+      "loss": 0.6159,
+      "step": 31346
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.845120484122173e-06,
+      "loss": 0.6172,
+      "step": 31347
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.844558683956654e-06,
+      "loss": 0.5785,
+      "step": 31348
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.843996915068738e-06,
+      "loss": 0.6051,
+      "step": 31349
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.843435177461274e-06,
+      "loss": 0.5739,
+      "step": 31350
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.842873471137123e-06,
+      "loss": 0.5815,
+      "step": 31351
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8423117960991355e-06,
+      "loss": 0.6253,
+      "step": 31352
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.841750152350163e-06,
+      "loss": 0.5369,
+      "step": 31353
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.84118853989306e-06,
+      "loss": 0.5721,
+      "step": 31354
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.840626958730684e-06,
+      "loss": 0.5665,
+      "step": 31355
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.840065408865886e-06,
+      "loss": 0.56,
+      "step": 31356
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8395038903015215e-06,
+      "loss": 0.586,
+      "step": 31357
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.838942403040437e-06,
+      "loss": 0.587,
+      "step": 31358
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8383809470854936e-06,
+      "loss": 0.5432,
+      "step": 31359
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.837819522439542e-06,
+      "loss": 0.5973,
+      "step": 31360
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.83725812910543e-06,
+      "loss": 0.5336,
+      "step": 31361
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.836696767086019e-06,
+      "loss": 0.6022,
+      "step": 31362
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.8361354363841555e-06,
+      "loss": 0.6377,
+      "step": 31363
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.835574137002693e-06,
+      "loss": 0.6288,
+      "step": 31364
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.835012868944481e-06,
+      "loss": 0.5996,
+      "step": 31365
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.834451632212377e-06,
+      "loss": 0.6019,
+      "step": 31366
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.833890426809227e-06,
+      "loss": 0.5709,
+      "step": 31367
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.833329252737892e-06,
+      "loss": 0.5873,
+      "step": 31368
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.832768110001209e-06,
+      "loss": 0.551,
+      "step": 31369
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.832206998602043e-06,
+      "loss": 0.5839,
+      "step": 31370
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.831645918543235e-06,
+      "loss": 0.7329,
+      "step": 31371
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.831084869827645e-06,
+      "loss": 0.609,
+      "step": 31372
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8305238524581175e-06,
+      "loss": 0.6433,
+      "step": 31373
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8299628664375064e-06,
+      "loss": 0.5597,
+      "step": 31374
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.829401911768658e-06,
+      "loss": 0.5837,
+      "step": 31375
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.828840988454429e-06,
+      "loss": 0.5789,
+      "step": 31376
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.828280096497665e-06,
+      "loss": 0.6295,
+      "step": 31377
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.827719235901214e-06,
+      "loss": 0.6025,
+      "step": 31378
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.827158406667933e-06,
+      "loss": 0.6208,
+      "step": 31379
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.826597608800669e-06,
+      "loss": 0.6225,
+      "step": 31380
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8260368423022695e-06,
+      "loss": 0.6882,
+      "step": 31381
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8254761071755805e-06,
+      "loss": 0.5662,
+      "step": 31382
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.824915403423461e-06,
+      "loss": 0.5952,
+      "step": 31383
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.824354731048754e-06,
+      "loss": 0.5829,
+      "step": 31384
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.823794090054309e-06,
+      "loss": 0.614,
+      "step": 31385
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.823233480442972e-06,
+      "loss": 0.6993,
+      "step": 31386
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.822672902217598e-06,
+      "loss": 0.5678,
+      "step": 31387
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.822112355381029e-06,
+      "loss": 0.6111,
+      "step": 31388
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.82155183993612e-06,
+      "loss": 0.5804,
+      "step": 31389
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.820991355885715e-06,
+      "loss": 0.6515,
+      "step": 31390
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.820430903232661e-06,
+      "loss": 0.6518,
+      "step": 31391
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.819870481979806e-06,
+      "loss": 0.5722,
+      "step": 31392
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8193100921300015e-06,
+      "loss": 0.5738,
+      "step": 31393
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.81874973368609e-06,
+      "loss": 0.5942,
+      "step": 31394
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.818189406650927e-06,
+      "loss": 0.5998,
+      "step": 31395
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.817629111027349e-06,
+      "loss": 0.5936,
+      "step": 31396
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.81706884681821e-06,
+      "loss": 0.5852,
+      "step": 31397
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.816508614026355e-06,
+      "loss": 0.5524,
+      "step": 31398
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.815948412654628e-06,
+      "loss": 0.6146,
+      "step": 31399
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.815388242705881e-06,
+      "loss": 0.5678,
+      "step": 31400
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.814828104182958e-06,
+      "loss": 0.6153,
+      "step": 31401
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.814267997088704e-06,
+      "loss": 0.5793,
+      "step": 31402
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.813707921425963e-06,
+      "loss": 0.673,
+      "step": 31403
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8131478771975873e-06,
+      "loss": 0.6137,
+      "step": 31404
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8125878644064163e-06,
+      "loss": 0.602,
+      "step": 31405
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8120278830553046e-06,
+      "loss": 0.6703,
+      "step": 31406
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.811467933147085e-06,
+      "loss": 0.5871,
+      "step": 31407
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.810908014684612e-06,
+      "loss": 0.5945,
+      "step": 31408
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8103481276707243e-06,
+      "loss": 0.5507,
+      "step": 31409
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8097882721082747e-06,
+      "loss": 0.5577,
+      "step": 31410
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.809228448000104e-06,
+      "loss": 0.6074,
+      "step": 31411
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.808668655349056e-06,
+      "loss": 0.6055,
+      "step": 31412
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8081088941579724e-06,
+      "loss": 0.5528,
+      "step": 31413
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8075491644297036e-06,
+      "loss": 0.6203,
+      "step": 31414
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8069894661670923e-06,
+      "loss": 0.6079,
+      "step": 31415
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8064297993729773e-06,
+      "loss": 0.6318,
+      "step": 31416
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.805870164050209e-06,
+      "loss": 0.5301,
+      "step": 31417
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8053105602016307e-06,
+      "loss": 0.5821,
+      "step": 31418
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8047509878300814e-06,
+      "loss": 0.6495,
+      "step": 31419
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8041914469384045e-06,
+      "loss": 0.6173,
+      "step": 31420
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8036319375294485e-06,
+      "loss": 0.6215,
+      "step": 31421
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.803072459606053e-06,
+      "loss": 0.672,
+      "step": 31422
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.802513013171061e-06,
+      "loss": 0.5695,
+      "step": 31423
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8019535982273125e-06,
+      "loss": 0.5556,
+      "step": 31424
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.801394214777656e-06,
+      "loss": 0.564,
+      "step": 31425
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.800834862824928e-06,
+      "loss": 0.6093,
+      "step": 31426
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.8002755423719795e-06,
+      "loss": 0.6204,
+      "step": 31427
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7997162534216405e-06,
+      "loss": 0.6787,
+      "step": 31428
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7991569959767628e-06,
+      "loss": 0.5724,
+      "step": 31429
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.79859777004018e-06,
+      "loss": 0.6696,
+      "step": 31430
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.798038575614742e-06,
+      "loss": 0.631,
+      "step": 31431
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7974794127032833e-06,
+      "loss": 0.5715,
+      "step": 31432
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7969202813086548e-06,
+      "loss": 0.5328,
+      "step": 31433
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.796361181433684e-06,
+      "loss": 0.6654,
+      "step": 31434
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7958021130812216e-06,
+      "loss": 0.5815,
+      "step": 31435
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.795243076254107e-06,
+      "loss": 0.5988,
+      "step": 31436
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7946840709551756e-06,
+      "loss": 0.6041,
+      "step": 31437
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7941250971872744e-06,
+      "loss": 0.6253,
+      "step": 31438
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7935661549532422e-06,
+      "loss": 0.6136,
+      "step": 31439
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7930072442559174e-06,
+      "loss": 0.5959,
+      "step": 31440
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.792448365098137e-06,
+      "loss": 0.6167,
+      "step": 31441
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7918895174827487e-06,
+      "loss": 0.5298,
+      "step": 31442
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7913307014125833e-06,
+      "loss": 0.6012,
+      "step": 31443
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.790771916890492e-06,
+      "loss": 0.6223,
+      "step": 31444
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7902131639193007e-06,
+      "loss": 0.5618,
+      "step": 31445
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7896544425018577e-06,
+      "loss": 0.558,
+      "step": 31446
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7890957526409965e-06,
+      "loss": 0.6121,
+      "step": 31447
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7885370943395614e-06,
+      "loss": 0.5651,
+      "step": 31448
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7879784676003896e-06,
+      "loss": 0.5778,
+      "step": 31449
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7874198724263176e-06,
+      "loss": 0.6079,
+      "step": 31450
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7868613088201824e-06,
+      "loss": 0.5646,
+      "step": 31451
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.786302776784826e-06,
+      "loss": 0.6108,
+      "step": 31452
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.785744276323082e-06,
+      "loss": 0.6447,
+      "step": 31453
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.785185807437796e-06,
+      "loss": 0.6752,
+      "step": 31454
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7846273701318006e-06,
+      "loss": 0.6177,
+      "step": 31455
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.784068964407933e-06,
+      "loss": 0.628,
+      "step": 31456
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7835105902690316e-06,
+      "loss": 0.619,
+      "step": 31457
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7829522477179304e-06,
+      "loss": 0.5384,
+      "step": 31458
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7823939367574736e-06,
+      "loss": 0.6353,
+      "step": 31459
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7818356573904924e-06,
+      "loss": 0.6246,
+      "step": 31460
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7812774096198267e-06,
+      "loss": 0.6409,
+      "step": 31461
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.780719193448308e-06,
+      "loss": 0.6085,
+      "step": 31462
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.78016100887878e-06,
+      "loss": 0.5954,
+      "step": 31463
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.779602855914072e-06,
+      "loss": 0.605,
+      "step": 31464
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.77904473455703e-06,
+      "loss": 0.6667,
+      "step": 31465
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7784866448104773e-06,
+      "loss": 0.5896,
+      "step": 31466
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7779285866772588e-06,
+      "loss": 0.6218,
+      "step": 31467
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.777370560160204e-06,
+      "loss": 0.5707,
+      "step": 31468
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7768125652621556e-06,
+      "loss": 0.5461,
+      "step": 31469
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7762546019859414e-06,
+      "loss": 0.6118,
+      "step": 31470
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.775696670334408e-06,
+      "loss": 0.6063,
+      "step": 31471
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.775138770310375e-06,
+      "loss": 0.5208,
+      "step": 31472
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7745809019166877e-06,
+      "loss": 0.6525,
+      "step": 31473
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.774023065156176e-06,
+      "loss": 0.6642,
+      "step": 31474
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.773465260031679e-06,
+      "loss": 0.6044,
+      "step": 31475
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.772907486546029e-06,
+      "loss": 0.6424,
+      "step": 31476
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7723497447020598e-06,
+      "loss": 0.5876,
+      "step": 31477
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.771792034502606e-06,
+      "loss": 0.5974,
+      "step": 31478
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.771234355950496e-06,
+      "loss": 0.6253,
+      "step": 31479
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.770676709048573e-06,
+      "loss": 0.6885,
+      "step": 31480
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.770119093799661e-06,
+      "loss": 0.603,
+      "step": 31481
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7695615102066072e-06,
+      "loss": 0.586,
+      "step": 31482
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7690039582722283e-06,
+      "loss": 0.5856,
+      "step": 31483
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.768446437999368e-06,
+      "loss": 0.5287,
+      "step": 31484
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7678889493908545e-06,
+      "loss": 0.6083,
+      "step": 31485
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.767331492449524e-06,
+      "loss": 0.5771,
+      "step": 31486
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7667740671782083e-06,
+      "loss": 0.615,
+      "step": 31487
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.766216673579739e-06,
+      "loss": 0.6207,
+      "step": 31488
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.765659311656945e-06,
+      "loss": 0.5962,
+      "step": 31489
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7651019814126656e-06,
+      "loss": 0.6267,
+      "step": 31490
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7645446828497255e-06,
+      "loss": 0.6269,
+      "step": 31491
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7639874159709666e-06,
+      "loss": 0.6531,
+      "step": 31492
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7634301807792072e-06,
+      "loss": 0.5825,
+      "step": 31493
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7628729772772886e-06,
+      "loss": 0.6568,
+      "step": 31494
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7623158054680365e-06,
+      "loss": 0.636,
+      "step": 31495
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.761758665354287e-06,
+      "loss": 0.6407,
+      "step": 31496
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7612015569388693e-06,
+      "loss": 0.6191,
+      "step": 31497
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7606444802246133e-06,
+      "loss": 0.6264,
+      "step": 31498
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7600874352143502e-06,
+      "loss": 0.6387,
+      "step": 31499
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7595304219109063e-06,
+      "loss": 0.6121,
+      "step": 31500
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7589734403171195e-06,
+      "loss": 0.6729,
+      "step": 31501
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7584164904358135e-06,
+      "loss": 0.6068,
+      "step": 31502
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7578595722698286e-06,
+      "loss": 0.5896,
+      "step": 31503
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7573026858219797e-06,
+      "loss": 0.5856,
+      "step": 31504
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7567458310951077e-06,
+      "loss": 0.5808,
+      "step": 31505
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.756189008092035e-06,
+      "loss": 0.6067,
+      "step": 31506
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.755632216815598e-06,
+      "loss": 0.5632,
+      "step": 31507
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7550754572686188e-06,
+      "loss": 0.5735,
+      "step": 31508
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7545187294539367e-06,
+      "loss": 0.5921,
+      "step": 31509
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7539620333743676e-06,
+      "loss": 0.6101,
+      "step": 31510
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.75340536903275e-06,
+      "loss": 0.6051,
+      "step": 31511
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.752848736431907e-06,
+      "loss": 0.5891,
+      "step": 31512
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.75229213557467e-06,
+      "loss": 0.5994,
+      "step": 31513
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.751735566463868e-06,
+      "loss": 0.5874,
+      "step": 31514
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7511790291023277e-06,
+      "loss": 0.575,
+      "step": 31515
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.750622523492876e-06,
+      "loss": 0.5752,
+      "step": 31516
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.750066049638339e-06,
+      "loss": 0.6004,
+      "step": 31517
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7495096075415493e-06,
+      "loss": 0.6309,
+      "step": 31518
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7489531972053318e-06,
+      "loss": 0.577,
+      "step": 31519
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.748396818632514e-06,
+      "loss": 0.5803,
+      "step": 31520
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7478404718259197e-06,
+      "loss": 0.5718,
+      "step": 31521
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.747284156788382e-06,
+      "loss": 0.6211,
+      "step": 31522
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.746727873522722e-06,
+      "loss": 0.6195,
+      "step": 31523
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7461716220317714e-06,
+      "loss": 0.6362,
+      "step": 31524
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7456154023183556e-06,
+      "loss": 0.5654,
+      "step": 31525
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7450592143852984e-06,
+      "loss": 0.6454,
+      "step": 31526
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7445030582354237e-06,
+      "loss": 0.5597,
+      "step": 31527
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.743946933871565e-06,
+      "loss": 0.6309,
+      "step": 31528
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7433908412965413e-06,
+      "loss": 0.5748,
+      "step": 31529
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.742834780513187e-06,
+      "loss": 0.6394,
+      "step": 31530
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7422787515243153e-06,
+      "loss": 0.5769,
+      "step": 31531
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7417227543327606e-06,
+      "loss": 0.6078,
+      "step": 31532
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.741166788941343e-06,
+      "loss": 0.6054,
+      "step": 31533
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.740610855352892e-06,
+      "loss": 0.6032,
+      "step": 31534
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7400549535702314e-06,
+      "loss": 0.6007,
+      "step": 31535
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.739499083596185e-06,
+      "loss": 0.6513,
+      "step": 31536
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.738943245433577e-06,
+      "loss": 0.5801,
+      "step": 31537
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.738387439085228e-06,
+      "loss": 0.6337,
+      "step": 31538
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7378316645539702e-06,
+      "loss": 0.5957,
+      "step": 31539
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7372759218426203e-06,
+      "loss": 0.5984,
+      "step": 31540
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7367202109540123e-06,
+      "loss": 0.5907,
+      "step": 31541
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7361645318909568e-06,
+      "loss": 0.6108,
+      "step": 31542
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7356088846562865e-06,
+      "loss": 0.6074,
+      "step": 31543
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7350532692528184e-06,
+      "loss": 0.6137,
+      "step": 31544
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7344976856833836e-06,
+      "loss": 0.6398,
+      "step": 31545
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7339421339507977e-06,
+      "loss": 0.6358,
+      "step": 31546
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7333866140578935e-06,
+      "loss": 0.5802,
+      "step": 31547
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7328311260074803e-06,
+      "loss": 0.6786,
+      "step": 31548
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.732275669802391e-06,
+      "loss": 0.6147,
+      "step": 31549
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7317202454454405e-06,
+      "loss": 0.6189,
+      "step": 31550
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7311648529394594e-06,
+      "loss": 0.574,
+      "step": 31551
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.730609492287265e-06,
+      "loss": 0.6562,
+      "step": 31552
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.730054163491681e-06,
+      "loss": 0.6407,
+      "step": 31553
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.729498866555523e-06,
+      "loss": 0.5741,
+      "step": 31554
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7289436014816226e-06,
+      "loss": 0.6292,
+      "step": 31555
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.728388368272795e-06,
+      "loss": 0.6338,
+      "step": 31556
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7278331669318637e-06,
+      "loss": 0.6152,
+      "step": 31557
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.727277997461649e-06,
+      "loss": 0.5834,
+      "step": 31558
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7267228598649687e-06,
+      "loss": 0.5862,
+      "step": 31559
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7261677541446496e-06,
+      "loss": 0.6384,
+      "step": 31560
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7256126803035052e-06,
+      "loss": 0.5801,
+      "step": 31561
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.725057638344365e-06,
+      "loss": 0.654,
+      "step": 31562
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7245026282700446e-06,
+      "loss": 0.6399,
+      "step": 31563
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7239476500833636e-06,
+      "loss": 0.6135,
+      "step": 31564
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7233927037871398e-06,
+      "loss": 0.5156,
+      "step": 31565
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7228377893841992e-06,
+      "loss": 0.5756,
+      "step": 31566
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7222829068773546e-06,
+      "loss": 0.6193,
+      "step": 31567
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7217280562694357e-06,
+      "loss": 0.5533,
+      "step": 31568
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.721173237563248e-06,
+      "loss": 0.6235,
+      "step": 31569
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7206184507616226e-06,
+      "loss": 0.5747,
+      "step": 31570
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7200636958673697e-06,
+      "loss": 0.6414,
+      "step": 31571
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7195089728833157e-06,
+      "loss": 0.593,
+      "step": 31572
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7189542818122757e-06,
+      "loss": 0.601,
+      "step": 31573
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7183996226570684e-06,
+      "loss": 0.6506,
+      "step": 31574
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.71784499542051e-06,
+      "loss": 0.6174,
+      "step": 31575
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7172904001054233e-06,
+      "loss": 0.6015,
+      "step": 31576
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7167358367146247e-06,
+      "loss": 0.6282,
+      "step": 31577
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7161813052509275e-06,
+      "loss": 0.6135,
+      "step": 31578
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7156268057171605e-06,
+      "loss": 0.626,
+      "step": 31579
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.715072338116127e-06,
+      "loss": 0.5797,
+      "step": 31580
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.714517902450655e-06,
+      "loss": 0.6035,
+      "step": 31581
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.713963498723555e-06,
+      "loss": 0.6295,
+      "step": 31582
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7134091269376517e-06,
+      "loss": 0.6745,
+      "step": 31583
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7128547870957567e-06,
+      "loss": 0.5954,
+      "step": 31584
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.712300479200689e-06,
+      "loss": 0.5427,
+      "step": 31585
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.711746203255259e-06,
+      "loss": 0.5664,
+      "step": 31586
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7111919592622925e-06,
+      "loss": 0.5854,
+      "step": 31587
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.710637747224598e-06,
+      "loss": 0.6467,
+      "step": 31588
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7100835671449977e-06,
+      "loss": 0.5424,
+      "step": 31589
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7095294190263054e-06,
+      "loss": 0.5773,
+      "step": 31590
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7089753028713372e-06,
+      "loss": 0.5902,
+      "step": 31591
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7084212186829026e-06,
+      "loss": 0.6385,
+      "step": 31592
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.707867166463828e-06,
+      "loss": 0.6236,
+      "step": 31593
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.707313146216921e-06,
+      "loss": 0.5399,
+      "step": 31594
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.706759157945e-06,
+      "loss": 0.5989,
+      "step": 31595
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7062052016508786e-06,
+      "loss": 0.5654,
+      "step": 31596
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7056512773373687e-06,
+      "loss": 0.5866,
+      "step": 31597
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.70509738500729e-06,
+      "loss": 0.5401,
+      "step": 31598
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.704543524663453e-06,
+      "loss": 0.5851,
+      "step": 31599
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7039896963086774e-06,
+      "loss": 0.7052,
+      "step": 31600
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7034358999457744e-06,
+      "loss": 0.542,
+      "step": 31601
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.702882135577557e-06,
+      "loss": 0.5634,
+      "step": 31602
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.702328403206835e-06,
+      "loss": 0.5548,
+      "step": 31603
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7017747028364325e-06,
+      "loss": 0.5717,
+      "step": 31604
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.701221034469152e-06,
+      "loss": 0.6305,
+      "step": 31605
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.70066739810782e-06,
+      "loss": 0.5756,
+      "step": 31606
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7001137937552344e-06,
+      "loss": 0.614,
+      "step": 31607
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.69956022141422e-06,
+      "loss": 0.5387,
+      "step": 31608
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6990066810875803e-06,
+      "loss": 0.6144,
+      "step": 31609
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.698453172778138e-06,
+      "loss": 0.5994,
+      "step": 31610
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6978996964886994e-06,
+      "loss": 0.5885,
+      "step": 31611
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.697346252222079e-06,
+      "loss": 0.6262,
+      "step": 31612
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.696792839981084e-06,
+      "loss": 0.6404,
+      "step": 31613
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6962394597685346e-06,
+      "loss": 0.6459,
+      "step": 31614
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.695686111587238e-06,
+      "loss": 0.6083,
+      "step": 31615
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6951327954400028e-06,
+      "loss": 0.6332,
+      "step": 31616
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.694579511329652e-06,
+      "loss": 0.6008,
+      "step": 31617
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6940262592589826e-06,
+      "loss": 0.6024,
+      "step": 31618
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.693473039230816e-06,
+      "loss": 0.649,
+      "step": 31619
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6929198512479557e-06,
+      "loss": 0.6277,
+      "step": 31620
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.692366695313221e-06,
+      "loss": 0.6028,
+      "step": 31621
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.691813571429418e-06,
+      "loss": 0.6722,
+      "step": 31622
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.691260479599358e-06,
+      "loss": 0.5596,
+      "step": 31623
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6907074198258465e-06,
+      "loss": 0.6032,
+      "step": 31624
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6901543921117033e-06,
+      "loss": 0.5959,
+      "step": 31625
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.689601396459729e-06,
+      "loss": 0.67,
+      "step": 31626
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6890484328727418e-06,
+      "loss": 0.5983,
+      "step": 31627
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6884955013535486e-06,
+      "loss": 0.6034,
+      "step": 31628
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.687942601904957e-06,
+      "loss": 0.5498,
+      "step": 31629
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.687389734529775e-06,
+      "loss": 0.601,
+      "step": 31630
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.686836899230818e-06,
+      "loss": 0.5958,
+      "step": 31631
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6862840960108904e-06,
+      "loss": 0.6014,
+      "step": 31632
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6857313248728033e-06,
+      "loss": 0.6022,
+      "step": 31633
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6851785858193613e-06,
+      "loss": 0.6171,
+      "step": 31634
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6846258788533797e-06,
+      "loss": 0.6302,
+      "step": 31635
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6840732039776627e-06,
+      "loss": 0.6169,
+      "step": 31636
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6835205611950165e-06,
+      "loss": 0.5692,
+      "step": 31637
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.682967950508256e-06,
+      "loss": 0.6258,
+      "step": 31638
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.682415371920185e-06,
+      "loss": 0.6022,
+      "step": 31639
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.681862825433612e-06,
+      "loss": 0.6262,
+      "step": 31640
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6813103110513404e-06,
+      "loss": 0.6063,
+      "step": 31641
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6807578287761857e-06,
+      "loss": 0.5853,
+      "step": 31642
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6802053786109472e-06,
+      "loss": 0.611,
+      "step": 31643
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.679652960558443e-06,
+      "loss": 0.6307,
+      "step": 31644
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6791005746214658e-06,
+      "loss": 0.6044,
+      "step": 31645
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6785482208028344e-06,
+      "loss": 0.5399,
+      "step": 31646
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6779958991053466e-06,
+      "loss": 0.5651,
+      "step": 31647
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6774436095318167e-06,
+      "loss": 0.6,
+      "step": 31648
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.676891352085047e-06,
+      "loss": 0.5268,
+      "step": 31649
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.676339126767845e-06,
+      "loss": 0.6494,
+      "step": 31650
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.675786933583012e-06,
+      "loss": 0.5791,
+      "step": 31651
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6752347725333605e-06,
+      "loss": 0.5842,
+      "step": 31652
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.674682643621694e-06,
+      "loss": 0.5627,
+      "step": 31653
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.674130546850814e-06,
+      "loss": 0.5764,
+      "step": 31654
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.673578482223533e-06,
+      "loss": 0.6015,
+      "step": 31655
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6730264497426526e-06,
+      "loss": 0.5772,
+      "step": 31656
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6724744494109765e-06,
+      "loss": 0.5456,
+      "step": 31657
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.671922481231309e-06,
+      "loss": 0.6183,
+      "step": 31658
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6713705452064595e-06,
+      "loss": 0.639,
+      "step": 31659
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.670818641339229e-06,
+      "loss": 0.579,
+      "step": 31660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6702667696324235e-06,
+      "loss": 0.6628,
+      "step": 31661
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6697149300888423e-06,
+      "loss": 0.5581,
+      "step": 31662
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6691631227112966e-06,
+      "loss": 0.6409,
+      "step": 31663
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6686113475025854e-06,
+      "loss": 0.5755,
+      "step": 31664
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.668059604465516e-06,
+      "loss": 0.7164,
+      "step": 31665
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.6675078936028907e-06,
+      "loss": 0.5678,
+      "step": 31666
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.666956214917512e-06,
+      "loss": 0.6184,
+      "step": 31667
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6664045684121796e-06,
+      "loss": 0.5865,
+      "step": 31668
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6658529540897047e-06,
+      "loss": 0.5811,
+      "step": 31669
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.665301371952886e-06,
+      "loss": 0.6301,
+      "step": 31670
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6647498220045265e-06,
+      "loss": 0.5848,
+      "step": 31671
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.664198304247425e-06,
+      "loss": 0.6032,
+      "step": 31672
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6636468186843897e-06,
+      "loss": 0.5581,
+      "step": 31673
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6630953653182224e-06,
+      "loss": 0.5888,
+      "step": 31674
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6625439441517187e-06,
+      "loss": 0.6107,
+      "step": 31675
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.661992555187689e-06,
+      "loss": 0.5202,
+      "step": 31676
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.661441198428931e-06,
+      "loss": 0.5832,
+      "step": 31677
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.660889873878246e-06,
+      "loss": 0.5823,
+      "step": 31678
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.660338581538434e-06,
+      "loss": 0.5567,
+      "step": 31679
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6597873214123015e-06,
+      "loss": 0.6075,
+      "step": 31680
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.659236093502643e-06,
+      "loss": 0.6743,
+      "step": 31681
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.658684897812269e-06,
+      "loss": 0.6477,
+      "step": 31682
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.658133734343968e-06,
+      "loss": 0.5765,
+      "step": 31683
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.65758260310055e-06,
+      "loss": 0.5838,
+      "step": 31684
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6570315040848094e-06,
+      "loss": 0.6067,
+      "step": 31685
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6564804372995512e-06,
+      "loss": 0.6069,
+      "step": 31686
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6559294027475758e-06,
+      "loss": 0.607,
+      "step": 31687
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6553784004316795e-06,
+      "loss": 0.6205,
+      "step": 31688
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6548274303546606e-06,
+      "loss": 0.6404,
+      "step": 31689
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6542764925193254e-06,
+      "loss": 0.6286,
+      "step": 31690
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6537255869284694e-06,
+      "loss": 0.5729,
+      "step": 31691
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6531747135848895e-06,
+      "loss": 0.6648,
+      "step": 31692
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.652623872491391e-06,
+      "loss": 0.5744,
+      "step": 31693
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6520730636507686e-06,
+      "loss": 0.6375,
+      "step": 31694
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.651522287065823e-06,
+      "loss": 0.5942,
+      "step": 31695
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6509715427393477e-06,
+      "loss": 0.6099,
+      "step": 31696
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6504208306741495e-06,
+      "loss": 0.5421,
+      "step": 31697
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.649870150873023e-06,
+      "loss": 0.5951,
+      "step": 31698
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.649319503338765e-06,
+      "loss": 0.6146,
+      "step": 31699
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.648768888074171e-06,
+      "loss": 0.5761,
+      "step": 31700
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6482183050820464e-06,
+      "loss": 0.6056,
+      "step": 31701
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6476677543651804e-06,
+      "loss": 0.5774,
+      "step": 31702
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6471172359263827e-06,
+      "loss": 0.618,
+      "step": 31703
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6465667497684367e-06,
+      "loss": 0.6597,
+      "step": 31704
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.646016295894148e-06,
+      "loss": 0.5776,
+      "step": 31705
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6454658743063076e-06,
+      "loss": 0.5902,
+      "step": 31706
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6449154850077205e-06,
+      "loss": 0.5883,
+      "step": 31707
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.644365128001178e-06,
+      "loss": 0.6092,
+      "step": 31708
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.643814803289479e-06,
+      "loss": 0.6602,
+      "step": 31709
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6432645108754138e-06,
+      "loss": 0.6155,
+      "step": 31710
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6427142507617872e-06,
+      "loss": 0.6281,
+      "step": 31711
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.642164022951391e-06,
+      "loss": 0.5777,
+      "step": 31712
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6416138274470194e-06,
+      "loss": 0.593,
+      "step": 31713
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.641063664251472e-06,
+      "loss": 0.6056,
+      "step": 31714
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6405135333675424e-06,
+      "loss": 0.6131,
+      "step": 31715
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.639963434798026e-06,
+      "loss": 0.6116,
+      "step": 31716
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6394133685457154e-06,
+      "loss": 0.6517,
+      "step": 31717
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6388633346134117e-06,
+      "loss": 0.5832,
+      "step": 31718
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6383133330039022e-06,
+      "loss": 0.5806,
+      "step": 31719
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6377633637199927e-06,
+      "loss": 0.5994,
+      "step": 31720
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6372134267644645e-06,
+      "loss": 0.5793,
+      "step": 31721
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6366635221401216e-06,
+      "loss": 0.5893,
+      "step": 31722
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6361136498497507e-06,
+      "loss": 0.5965,
+      "step": 31723
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6355638098961544e-06,
+      "loss": 0.6041,
+      "step": 31724
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6350140022821222e-06,
+      "loss": 0.573,
+      "step": 31725
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.634464227010447e-06,
+      "loss": 0.6338,
+      "step": 31726
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.633914484083921e-06,
+      "loss": 0.667,
+      "step": 31727
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6333647735053422e-06,
+      "loss": 0.588,
+      "step": 31728
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6328150952775032e-06,
+      "loss": 0.5845,
+      "step": 31729
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6322654494031905e-06,
+      "loss": 0.6211,
+      "step": 31730
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.631715835885207e-06,
+      "loss": 0.652,
+      "step": 31731
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.631166254726338e-06,
+      "loss": 0.6251,
+      "step": 31732
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6306167059293806e-06,
+      "loss": 0.6464,
+      "step": 31733
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6300671894971207e-06,
+      "loss": 0.5666,
+      "step": 31734
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.629517705432358e-06,
+      "loss": 0.6437,
+      "step": 31735
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6289682537378824e-06,
+      "loss": 0.6438,
+      "step": 31736
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6284188344164838e-06,
+      "loss": 0.6562,
+      "step": 31737
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.627869447470953e-06,
+      "loss": 0.5587,
+      "step": 31738
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6273200929040866e-06,
+      "loss": 0.595,
+      "step": 31739
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6267707707186696e-06,
+      "loss": 0.6346,
+      "step": 31740
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6262214809175032e-06,
+      "loss": 0.6035,
+      "step": 31741
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.625672223503366e-06,
+      "loss": 0.6272,
+      "step": 31742
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.625122998479058e-06,
+      "loss": 0.5986,
+      "step": 31743
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6245738058473647e-06,
+      "loss": 0.5995,
+      "step": 31744
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6240246456110808e-06,
+      "loss": 0.5548,
+      "step": 31745
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.623475517772992e-06,
+      "loss": 0.6032,
+      "step": 31746
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6229264223358994e-06,
+      "loss": 0.5593,
+      "step": 31747
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.622377359302578e-06,
+      "loss": 0.5331,
+      "step": 31748
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6218283286758273e-06,
+      "loss": 0.5749,
+      "step": 31749
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.621279330458436e-06,
+      "loss": 0.5591,
+      "step": 31750
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.620730364653189e-06,
+      "loss": 0.6614,
+      "step": 31751
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6201814312628824e-06,
+      "loss": 0.5853,
+      "step": 31752
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6196325302903024e-06,
+      "loss": 0.6542,
+      "step": 31753
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6190836617382376e-06,
+      "loss": 0.6237,
+      "step": 31754
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6185348256094743e-06,
+      "loss": 0.558,
+      "step": 31755
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6179860219068084e-06,
+      "loss": 0.5902,
+      "step": 31756
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6174372506330213e-06,
+      "loss": 0.565,
+      "step": 31757
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6168885117909102e-06,
+      "loss": 0.6212,
+      "step": 31758
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.616339805383252e-06,
+      "loss": 0.6265,
+      "step": 31759
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6157911314128446e-06,
+      "loss": 0.5981,
+      "step": 31760
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.615242489882469e-06,
+      "loss": 0.5771,
+      "step": 31761
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6146938807949185e-06,
+      "loss": 0.568,
+      "step": 31762
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6141453041529782e-06,
+      "loss": 0.6081,
+      "step": 31763
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.613596759959437e-06,
+      "loss": 0.5748,
+      "step": 31764
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.613048248217077e-06,
+      "loss": 0.5727,
+      "step": 31765
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6124997689286934e-06,
+      "loss": 0.631,
+      "step": 31766
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6119513220970646e-06,
+      "loss": 0.6142,
+      "step": 31767
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6114029077249902e-06,
+      "loss": 0.5717,
+      "step": 31768
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6108545258152405e-06,
+      "loss": 0.6237,
+      "step": 31769
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6103061763706137e-06,
+      "loss": 0.5882,
+      "step": 31770
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.609757859393893e-06,
+      "loss": 0.6041,
+      "step": 31771
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6092095748878607e-06,
+      "loss": 0.6159,
+      "step": 31772
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6086613228553093e-06,
+      "loss": 0.5822,
+      "step": 31773
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6081131032990213e-06,
+      "loss": 0.5828,
+      "step": 31774
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.607564916221783e-06,
+      "loss": 0.6608,
+      "step": 31775
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6070167616263763e-06,
+      "loss": 0.6543,
+      "step": 31776
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6064686395155923e-06,
+      "loss": 0.5392,
+      "step": 31777
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6059205498922103e-06,
+      "loss": 0.6039,
+      "step": 31778
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.605372492759026e-06,
+      "loss": 0.5764,
+      "step": 31779
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6048244681188095e-06,
+      "loss": 0.6112,
+      "step": 31780
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6042764759743567e-06,
+      "loss": 0.5908,
+      "step": 31781
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6037285163284442e-06,
+      "loss": 0.6271,
+      "step": 31782
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6031805891838645e-06,
+      "loss": 0.5698,
+      "step": 31783
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.602632694543393e-06,
+      "loss": 0.6277,
+      "step": 31784
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.602084832409827e-06,
+      "loss": 0.5496,
+      "step": 31785
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.601537002785933e-06,
+      "loss": 0.5621,
+      "step": 31786
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6009892056745076e-06,
+      "loss": 0.5652,
+      "step": 31787
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.6004414410783295e-06,
+      "loss": 0.6004,
+      "step": 31788
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.599893709000181e-06,
+      "loss": 0.5696,
+      "step": 31789
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.599346009442848e-06,
+      "loss": 0.6042,
+      "step": 31790
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5987983424091134e-06,
+      "loss": 0.5748,
+      "step": 31791
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5982507079017593e-06,
+      "loss": 0.6485,
+      "step": 31792
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5977031059235645e-06,
+      "loss": 0.5647,
+      "step": 31793
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5971555364773193e-06,
+      "loss": 0.5517,
+      "step": 31794
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5966079995658022e-06,
+      "loss": 0.5637,
+      "step": 31795
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5960604951917942e-06,
+      "loss": 0.5891,
+      "step": 31796
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5955130233580748e-06,
+      "loss": 0.5834,
+      "step": 31797
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5949655840674335e-06,
+      "loss": 0.6157,
+      "step": 31798
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5944181773226438e-06,
+      "loss": 0.59,
+      "step": 31799
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5938708031264947e-06,
+      "loss": 0.5979,
+      "step": 31800
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5933234614817647e-06,
+      "loss": 0.605,
+      "step": 31801
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.592776152391234e-06,
+      "loss": 0.5912,
+      "step": 31802
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5922288758576818e-06,
+      "loss": 0.6296,
+      "step": 31803
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.591681631883893e-06,
+      "loss": 0.5525,
+      "step": 31804
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.591134420472645e-06,
+      "loss": 0.5803,
+      "step": 31805
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.590587241626725e-06,
+      "loss": 0.6361,
+      "step": 31806
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5900400953489024e-06,
+      "loss": 0.6037,
+      "step": 31807
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5894929816419666e-06,
+      "loss": 0.6296,
+      "step": 31808
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5889459005086947e-06,
+      "loss": 0.6585,
+      "step": 31809
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.588398851951862e-06,
+      "loss": 0.6615,
+      "step": 31810
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5878518359742563e-06,
+      "loss": 0.6548,
+      "step": 31811
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5873048525786537e-06,
+      "loss": 0.6287,
+      "step": 31812
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5867579017678334e-06,
+      "loss": 0.5354,
+      "step": 31813
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5862109835445704e-06,
+      "loss": 0.5836,
+      "step": 31814
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.585664097911651e-06,
+      "loss": 0.646,
+      "step": 31815
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5851172448718473e-06,
+      "loss": 0.5904,
+      "step": 31816
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.584570424427949e-06,
+      "loss": 0.5422,
+      "step": 31817
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5840236365827206e-06,
+      "loss": 0.5645,
+      "step": 31818
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.58347688133895e-06,
+      "loss": 0.5858,
+      "step": 31819
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.582930158699409e-06,
+      "loss": 0.5778,
+      "step": 31820
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5823834686668834e-06,
+      "loss": 0.5963,
+      "step": 31821
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.581836811244143e-06,
+      "loss": 0.601,
+      "step": 31822
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5812901864339753e-06,
+      "loss": 0.6177,
+      "step": 31823
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5807435942391464e-06,
+      "loss": 0.6146,
+      "step": 31824
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5801970346624417e-06,
+      "loss": 0.6448,
+      "step": 31825
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5796505077066334e-06,
+      "loss": 0.6148,
+      "step": 31826
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.579104013374505e-06,
+      "loss": 0.5799,
+      "step": 31827
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.578557551668829e-06,
+      "loss": 0.5906,
+      "step": 31828
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.578011122592382e-06,
+      "loss": 0.578,
+      "step": 31829
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5774647261479413e-06,
+      "loss": 0.5974,
+      "step": 31830
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5769183623382807e-06,
+      "loss": 0.6156,
+      "step": 31831
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.576372031166182e-06,
+      "loss": 0.6068,
+      "step": 31832
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5758257326344182e-06,
+      "loss": 0.581,
+      "step": 31833
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5752794667457657e-06,
+      "loss": 0.5613,
+      "step": 31834
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.574733233502996e-06,
+      "loss": 0.6404,
+      "step": 31835
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5741870329088913e-06,
+      "loss": 0.5641,
+      "step": 31836
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5736408649662215e-06,
+      "loss": 0.5482,
+      "step": 31837
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5730947296777674e-06,
+      "loss": 0.6288,
+      "step": 31838
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5725486270463018e-06,
+      "loss": 0.5934,
+      "step": 31839
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5720025570745987e-06,
+      "loss": 0.587,
+      "step": 31840
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5714565197654306e-06,
+      "loss": 0.5662,
+      "step": 31841
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.570910515121576e-06,
+      "loss": 0.575,
+      "step": 31842
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5703645431458065e-06,
+      "loss": 0.5448,
+      "step": 31843
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5698186038409033e-06,
+      "loss": 0.6382,
+      "step": 31844
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.569272697209629e-06,
+      "loss": 0.5713,
+      "step": 31845
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5687268232547666e-06,
+      "loss": 0.5723,
+      "step": 31846
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5681809819790823e-06,
+      "loss": 0.645,
+      "step": 31847
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.567635173385359e-06,
+      "loss": 0.6323,
+      "step": 31848
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5670893974763642e-06,
+      "loss": 0.5942,
+      "step": 31849
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5665436542548725e-06,
+      "loss": 0.6015,
+      "step": 31850
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5659979437236568e-06,
+      "loss": 0.6285,
+      "step": 31851
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.565452265885486e-06,
+      "loss": 0.6945,
+      "step": 31852
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5649066207431404e-06,
+      "loss": 0.601,
+      "step": 31853
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.564361008299385e-06,
+      "loss": 0.6585,
+      "step": 31854
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5638154285570035e-06,
+      "loss": 0.5695,
+      "step": 31855
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.563269881518754e-06,
+      "loss": 0.6401,
+      "step": 31856
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5627243671874177e-06,
+      "loss": 0.6139,
+      "step": 31857
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5621788855657624e-06,
+      "loss": 0.5254,
+      "step": 31858
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.561633436656564e-06,
+      "loss": 0.5926,
+      "step": 31859
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.561088020462592e-06,
+      "loss": 0.6258,
+      "step": 31860
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.560542636986617e-06,
+      "loss": 0.6577,
+      "step": 31861
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5599972862314068e-06,
+      "loss": 0.603,
+      "step": 31862
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5594519681997408e-06,
+      "loss": 0.5828,
+      "step": 31863
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.558906682894382e-06,
+      "loss": 0.5791,
+      "step": 31864
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.558361430318108e-06,
+      "loss": 0.6349,
+      "step": 31865
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5578162104736856e-06,
+      "loss": 0.6737,
+      "step": 31866
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5572710233638865e-06,
+      "loss": 0.6104,
+      "step": 31867
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.556725868991476e-06,
+      "loss": 0.6365,
+      "step": 31868
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5561807473592323e-06,
+      "loss": 0.5833,
+      "step": 31869
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5556356584699227e-06,
+      "loss": 0.5786,
+      "step": 31870
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.555090602326313e-06,
+      "loss": 0.5863,
+      "step": 31871
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5545455789311776e-06,
+      "loss": 0.6407,
+      "step": 31872
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5540005882872795e-06,
+      "loss": 0.6407,
+      "step": 31873
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5534556303973954e-06,
+      "loss": 0.5681,
+      "step": 31874
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5529107052642875e-06,
+      "loss": 0.6268,
+      "step": 31875
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5523658128907324e-06,
+      "loss": 0.5937,
+      "step": 31876
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5518209532794943e-06,
+      "loss": 0.6379,
+      "step": 31877
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.551276126433344e-06,
+      "loss": 0.5494,
+      "step": 31878
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.550731332355043e-06,
+      "loss": 0.589,
+      "step": 31879
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5501865710473695e-06,
+      "loss": 0.6204,
+      "step": 31880
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.549641842513083e-06,
+      "loss": 0.6199,
+      "step": 31881
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.549097146754962e-06,
+      "loss": 0.6076,
+      "step": 31882
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5485524837757614e-06,
+      "loss": 0.5956,
+      "step": 31883
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.548007853578258e-06,
+      "loss": 0.6203,
+      "step": 31884
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.547463256165212e-06,
+      "loss": 0.5905,
+      "step": 31885
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5469186915393995e-06,
+      "loss": 0.6148,
+      "step": 31886
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.546374159703583e-06,
+      "loss": 0.5623,
+      "step": 31887
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.545829660660529e-06,
+      "loss": 0.5716,
+      "step": 31888
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5452851944130053e-06,
+      "loss": 0.588,
+      "step": 31889
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.544740760963774e-06,
+      "loss": 0.5729,
+      "step": 31890
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.544196360315608e-06,
+      "loss": 0.6233,
+      "step": 31891
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5436519924712688e-06,
+      "loss": 0.56,
+      "step": 31892
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.543107657433531e-06,
+      "loss": 0.6248,
+      "step": 31893
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5425633552051475e-06,
+      "loss": 0.6161,
+      "step": 31894
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.542019085788895e-06,
+      "loss": 0.5864,
+      "step": 31895
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5414748491875293e-06,
+      "loss": 0.6212,
+      "step": 31896
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5409306454038263e-06,
+      "loss": 0.5895,
+      "step": 31897
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.540386474440546e-06,
+      "loss": 0.631,
+      "step": 31898
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5398423363004543e-06,
+      "loss": 0.6066,
+      "step": 31899
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5392982309863113e-06,
+      "loss": 0.6027,
+      "step": 31900
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.53875415850089e-06,
+      "loss": 0.6487,
+      "step": 31901
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5382101188469477e-06,
+      "loss": 0.5662,
+      "step": 31902
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.537666112027255e-06,
+      "loss": 0.6202,
+      "step": 31903
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5371221380445743e-06,
+      "loss": 0.631,
+      "step": 31904
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5365781969016677e-06,
+      "loss": 0.6459,
+      "step": 31905
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5360342886012977e-06,
+      "loss": 0.6228,
+      "step": 31906
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.535490413146234e-06,
+      "loss": 0.6389,
+      "step": 31907
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.534946570539236e-06,
+      "loss": 0.5979,
+      "step": 31908
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5344027607830687e-06,
+      "loss": 0.585,
+      "step": 31909
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.533858983880494e-06,
+      "loss": 0.6377,
+      "step": 31910
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.533315239834272e-06,
+      "loss": 0.6097,
+      "step": 31911
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.532771528647173e-06,
+      "loss": 0.5978,
+      "step": 31912
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.532227850321952e-06,
+      "loss": 0.6603,
+      "step": 31913
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5316842048613786e-06,
+      "loss": 0.6323,
+      "step": 31914
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.531140592268213e-06,
+      "loss": 0.618,
+      "step": 31915
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.530597012545216e-06,
+      "loss": 0.5708,
+      "step": 31916
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5300534656951468e-06,
+      "loss": 0.6124,
+      "step": 31917
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5295099517207743e-06,
+      "loss": 0.5629,
+      "step": 31918
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.528966470624853e-06,
+      "loss": 0.5661,
+      "step": 31919
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5284230224101556e-06,
+      "loss": 0.6015,
+      "step": 31920
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5278796070794284e-06,
+      "loss": 0.566,
+      "step": 31921
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5273362246354447e-06,
+      "loss": 0.6201,
+      "step": 31922
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5267928750809577e-06,
+      "loss": 0.6488,
+      "step": 31923
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.526249558418735e-06,
+      "loss": 0.6509,
+      "step": 31924
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5257062746515336e-06,
+      "loss": 0.5655,
+      "step": 31925
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5251630237821154e-06,
+      "loss": 0.6115,
+      "step": 31926
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5246198058132374e-06,
+      "loss": 0.6273,
+      "step": 31927
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.524076620747665e-06,
+      "loss": 0.5932,
+      "step": 31928
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.523533468588156e-06,
+      "loss": 0.5543,
+      "step": 31929
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.522990349337466e-06,
+      "loss": 0.537,
+      "step": 31930
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5224472629983673e-06,
+      "loss": 0.5916,
+      "step": 31931
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5219042095736035e-06,
+      "loss": 0.65,
+      "step": 31932
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.521361189065945e-06,
+      "loss": 0.6075,
+      "step": 31933
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5208182014781443e-06,
+      "loss": 0.5527,
+      "step": 31934
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5202752468129676e-06,
+      "loss": 0.5989,
+      "step": 31935
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5197323250731695e-06,
+      "loss": 0.623,
+      "step": 31936
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.51918943626151e-06,
+      "loss": 0.5919,
+      "step": 31937
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5186465803807422e-06,
+      "loss": 0.6573,
+      "step": 31938
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.518103757433634e-06,
+      "loss": 0.5927,
+      "step": 31939
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.517560967422935e-06,
+      "loss": 0.5952,
+      "step": 31940
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5170182103514104e-06,
+      "loss": 0.6559,
+      "step": 31941
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5164754862218154e-06,
+      "loss": 0.552,
+      "step": 31942
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.515932795036907e-06,
+      "loss": 0.656,
+      "step": 31943
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5153901367994403e-06,
+      "loss": 0.5602,
+      "step": 31944
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.514847511512177e-06,
+      "loss": 0.6185,
+      "step": 31945
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5143049191778745e-06,
+      "loss": 0.6244,
+      "step": 31946
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5137623597992875e-06,
+      "loss": 0.604,
+      "step": 31947
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.51321983337917e-06,
+      "loss": 0.5829,
+      "step": 31948
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5126773399202864e-06,
+      "loss": 0.5863,
+      "step": 31949
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5121348794253886e-06,
+      "loss": 0.6177,
+      "step": 31950
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.51159245189723e-06,
+      "loss": 0.5706,
+      "step": 31951
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5110500573385774e-06,
+      "loss": 0.5411,
+      "step": 31952
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.510507695752172e-06,
+      "loss": 0.5646,
+      "step": 31953
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.509965367140783e-06,
+      "loss": 0.6133,
+      "step": 31954
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5094230715071552e-06,
+      "loss": 0.6013,
+      "step": 31955
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.508880808854054e-06,
+      "loss": 0.6641,
+      "step": 31956
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.5083385791842274e-06,
+      "loss": 0.6625,
+      "step": 31957
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.50779638250044e-06,
+      "loss": 0.6103,
+      "step": 31958
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.507254218805434e-06,
+      "loss": 0.6104,
+      "step": 31959
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5067120881019735e-06,
+      "loss": 0.5533,
+      "step": 31960
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5061699903928083e-06,
+      "loss": 0.5829,
+      "step": 31961
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5056279256806968e-06,
+      "loss": 0.6084,
+      "step": 31962
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.505085893968393e-06,
+      "loss": 0.6148,
+      "step": 31963
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5045438952586497e-06,
+      "loss": 0.6264,
+      "step": 31964
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5040019295542183e-06,
+      "loss": 0.5451,
+      "step": 31965
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5034599968578573e-06,
+      "loss": 0.6011,
+      "step": 31966
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.502918097172321e-06,
+      "loss": 0.673,
+      "step": 31967
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.502376230500355e-06,
+      "loss": 0.5978,
+      "step": 31968
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5018343968447234e-06,
+      "loss": 0.6641,
+      "step": 31969
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5012925962081733e-06,
+      "loss": 0.5813,
+      "step": 31970
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5007508285934587e-06,
+      "loss": 0.5743,
+      "step": 31971
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.5002090940033294e-06,
+      "loss": 0.587,
+      "step": 31972
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4996673924405457e-06,
+      "loss": 0.5705,
+      "step": 31973
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4991257239078536e-06,
+      "loss": 0.6346,
+      "step": 31974
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.498584088408009e-06,
+      "loss": 0.5752,
+      "step": 31975
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.498042485943759e-06,
+      "loss": 0.6353,
+      "step": 31976
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.497500916517863e-06,
+      "loss": 0.5683,
+      "step": 31977
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.496959380133065e-06,
+      "loss": 0.6241,
+      "step": 31978
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4964178767921284e-06,
+      "loss": 0.5996,
+      "step": 31979
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4958764064977903e-06,
+      "loss": 0.592,
+      "step": 31980
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4953349692528117e-06,
+      "loss": 0.6345,
+      "step": 31981
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4947935650599384e-06,
+      "loss": 0.5745,
+      "step": 31982
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4942521939219274e-06,
+      "loss": 0.6684,
+      "step": 31983
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4937108558415266e-06,
+      "loss": 0.6469,
+      "step": 31984
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4931695508214857e-06,
+      "loss": 0.5867,
+      "step": 31985
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4926282788645527e-06,
+      "loss": 0.5637,
+      "step": 31986
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4920870399734853e-06,
+      "loss": 0.6209,
+      "step": 31987
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.491545834151029e-06,
+      "loss": 0.6111,
+      "step": 31988
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4910046613999317e-06,
+      "loss": 0.5583,
+      "step": 31989
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.490463521722951e-06,
+      "loss": 0.6159,
+      "step": 31990
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.489922415122826e-06,
+      "loss": 0.5591,
+      "step": 31991
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4893813416023158e-06,
+      "loss": 0.584,
+      "step": 31992
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4888403011641613e-06,
+      "loss": 0.5849,
+      "step": 31993
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.488299293811119e-06,
+      "loss": 0.6118,
+      "step": 31994
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.487758319545933e-06,
+      "loss": 0.6742,
+      "step": 31995
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4872173783713604e-06,
+      "loss": 0.5818,
+      "step": 31996
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4866764702901367e-06,
+      "loss": 0.5716,
+      "step": 31997
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4861355953050212e-06,
+      "loss": 0.5843,
+      "step": 31998
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.485594753418755e-06,
+      "loss": 0.6099,
+      "step": 31999
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4850539446340924e-06,
+      "loss": 0.5819,
+      "step": 32000
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.484513168953779e-06,
+      "loss": 0.5537,
+      "step": 32001
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.483972426380563e-06,
+      "loss": 0.5965,
+      "step": 32002
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.483431716917187e-06,
+      "loss": 0.6217,
+      "step": 32003
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4828910405664074e-06,
+      "loss": 0.5722,
+      "step": 32004
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4823503973309657e-06,
+      "loss": 0.5903,
+      "step": 32005
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.481809787213608e-06,
+      "loss": 0.6181,
+      "step": 32006
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4812692102170877e-06,
+      "loss": 0.6374,
+      "step": 32007
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4807286663441465e-06,
+      "loss": 0.6127,
+      "step": 32008
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4801881555975327e-06,
+      "loss": 0.5576,
+      "step": 32009
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.479647677979988e-06,
+      "loss": 0.5597,
+      "step": 32010
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4791072334942676e-06,
+      "loss": 0.6528,
+      "step": 32011
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4785668221431123e-06,
+      "loss": 0.6322,
+      "step": 32012
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.47802644392927e-06,
+      "loss": 0.5558,
+      "step": 32013
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4774860988554803e-06,
+      "loss": 0.5506,
+      "step": 32014
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4769457869244983e-06,
+      "loss": 0.5938,
+      "step": 32015
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4764055081390624e-06,
+      "loss": 0.5798,
+      "step": 32016
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4758652625019263e-06,
+      "loss": 0.6453,
+      "step": 32017
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.475325050015823e-06,
+      "loss": 0.6709,
+      "step": 32018
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4747848706835074e-06,
+      "loss": 0.5821,
+      "step": 32019
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.474244724507717e-06,
+      "loss": 0.5985,
+      "step": 32020
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4737046114912043e-06,
+      "loss": 0.642,
+      "step": 32021
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4731645316367103e-06,
+      "loss": 0.6117,
+      "step": 32022
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.472624484946978e-06,
+      "loss": 0.6126,
+      "step": 32023
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.472084471424749e-06,
+      "loss": 0.6366,
+      "step": 32024
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4715444910727735e-06,
+      "loss": 0.6362,
+      "step": 32025
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4710045438937934e-06,
+      "loss": 0.5962,
+      "step": 32026
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4704646298905475e-06,
+      "loss": 0.6376,
+      "step": 32027
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.469924749065786e-06,
+      "loss": 0.5662,
+      "step": 32028
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.46938490142225e-06,
+      "loss": 0.6228,
+      "step": 32029
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.468845086962682e-06,
+      "loss": 0.6197,
+      "step": 32030
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4683053056898206e-06,
+      "loss": 0.6032,
+      "step": 32031
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.467765557606417e-06,
+      "loss": 0.617,
+      "step": 32032
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.467225842715206e-06,
+      "loss": 0.6576,
+      "step": 32033
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.466686161018941e-06,
+      "loss": 0.6317,
+      "step": 32034
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4661465125203487e-06,
+      "loss": 0.6221,
+      "step": 32035
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4656068972221846e-06,
+      "loss": 0.6166,
+      "step": 32036
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.465067315127181e-06,
+      "loss": 0.6305,
+      "step": 32037
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4645277662380884e-06,
+      "loss": 0.6131,
+      "step": 32038
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.463988250557644e-06,
+      "loss": 0.5389,
+      "step": 32039
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4634487680885887e-06,
+      "loss": 0.6165,
+      "step": 32040
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4629093188336617e-06,
+      "loss": 0.6429,
+      "step": 32041
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4623699027956104e-06,
+      "loss": 0.5669,
+      "step": 32042
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.461830519977172e-06,
+      "loss": 0.6067,
+      "step": 32043
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.461291170381088e-06,
+      "loss": 0.5935,
+      "step": 32044
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.460751854010095e-06,
+      "loss": 0.6392,
+      "step": 32045
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.46021257086694e-06,
+      "loss": 0.6145,
+      "step": 32046
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.459673320954361e-06,
+      "loss": 0.5653,
+      "step": 32047
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4591341042750927e-06,
+      "loss": 0.5565,
+      "step": 32048
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4585949208318838e-06,
+      "loss": 0.6234,
+      "step": 32049
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.45805577062747e-06,
+      "loss": 0.5955,
+      "step": 32050
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4575166536645898e-06,
+      "loss": 0.5836,
+      "step": 32051
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4569775699459807e-06,
+      "loss": 0.6226,
+      "step": 32052
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4564385194743877e-06,
+      "loss": 0.5894,
+      "step": 32053
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4558995022525445e-06,
+      "loss": 0.6174,
+      "step": 32054
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4553605182831984e-06,
+      "loss": 0.6435,
+      "step": 32055
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4548215675690756e-06,
+      "loss": 0.6455,
+      "step": 32056
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4542826501129244e-06,
+      "loss": 0.6656,
+      "step": 32057
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4537437659174776e-06,
+      "loss": 0.6498,
+      "step": 32058
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4532049149854784e-06,
+      "loss": 0.6107,
+      "step": 32059
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4526660973196602e-06,
+      "loss": 0.5888,
+      "step": 32060
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.452127312922768e-06,
+      "loss": 0.5575,
+      "step": 32061
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4515885617975286e-06,
+      "loss": 0.5822,
+      "step": 32062
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.45104984394669e-06,
+      "loss": 0.6079,
+      "step": 32063
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4505111593729835e-06,
+      "loss": 0.5934,
+      "step": 32064
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4499725080791457e-06,
+      "loss": 0.5963,
+      "step": 32065
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.449433890067918e-06,
+      "loss": 0.6139,
+      "step": 32066
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.448895305342036e-06,
+      "loss": 0.5788,
+      "step": 32067
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4483567539042352e-06,
+      "loss": 0.6206,
+      "step": 32068
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4478182357572487e-06,
+      "loss": 0.6551,
+      "step": 32069
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4472797509038215e-06,
+      "loss": 0.6313,
+      "step": 32070
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.446741299346683e-06,
+      "loss": 0.5683,
+      "step": 32071
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.446202881088573e-06,
+      "loss": 0.5909,
+      "step": 32072
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4456644961322204e-06,
+      "loss": 0.5659,
+      "step": 32073
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.445126144480371e-06,
+      "loss": 0.6039,
+      "step": 32074
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.444587826135751e-06,
+      "loss": 0.6143,
+      "step": 32075
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.444049541101103e-06,
+      "loss": 0.5918,
+      "step": 32076
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.44351128937916e-06,
+      "loss": 0.6186,
+      "step": 32077
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.442973070972657e-06,
+      "loss": 0.6421,
+      "step": 32078
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4424348858843237e-06,
+      "loss": 0.5674,
+      "step": 32079
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.441896734116903e-06,
+      "loss": 0.6134,
+      "step": 32080
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.441358615673126e-06,
+      "loss": 0.5717,
+      "step": 32081
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4408205305557263e-06,
+      "loss": 0.6353,
+      "step": 32082
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.440282478767435e-06,
+      "loss": 0.5973,
+      "step": 32083
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.439744460310992e-06,
+      "loss": 0.6202,
+      "step": 32084
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4392064751891297e-06,
+      "loss": 0.5393,
+      "step": 32085
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.438668523404578e-06,
+      "loss": 0.6194,
+      "step": 32086
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.438130604960075e-06,
+      "loss": 0.5953,
+      "step": 32087
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4375927198583526e-06,
+      "loss": 0.6564,
+      "step": 32088
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4370548681021444e-06,
+      "loss": 0.5482,
+      "step": 32089
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.436517049694178e-06,
+      "loss": 0.6432,
+      "step": 32090
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.435979264637195e-06,
+      "loss": 0.624,
+      "step": 32091
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.435441512933919e-06,
+      "loss": 0.5845,
+      "step": 32092
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4349037945870946e-06,
+      "loss": 0.5549,
+      "step": 32093
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.434366109599441e-06,
+      "loss": 0.6316,
+      "step": 32094
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.433828457973698e-06,
+      "loss": 0.5839,
+      "step": 32095
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.433290839712592e-06,
+      "loss": 0.5852,
+      "step": 32096
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4327532548188624e-06,
+      "loss": 0.5817,
+      "step": 32097
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.432215703295234e-06,
+      "loss": 0.6427,
+      "step": 32098
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.431678185144447e-06,
+      "loss": 0.6389,
+      "step": 32099
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.431140700369221e-06,
+      "loss": 0.572,
+      "step": 32100
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4306032489722953e-06,
+      "loss": 0.6231,
+      "step": 32101
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4300658309563984e-06,
+      "loss": 0.6411,
+      "step": 32102
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4295284463242586e-06,
+      "loss": 0.6129,
+      "step": 32103
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.428991095078611e-06,
+      "loss": 0.6289,
+      "step": 32104
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.428453777222185e-06,
+      "loss": 0.5212,
+      "step": 32105
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.4279164927577103e-06,
+      "loss": 0.573,
+      "step": 32106
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4273792416879127e-06,
+      "loss": 0.6474,
+      "step": 32107
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.42684202401553e-06,
+      "loss": 0.673,
+      "step": 32108
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4263048397432862e-06,
+      "loss": 0.6293,
+      "step": 32109
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4257676888739144e-06,
+      "loss": 0.5585,
+      "step": 32110
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4252305714101387e-06,
+      "loss": 0.5958,
+      "step": 32111
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4246934873546957e-06,
+      "loss": 0.5977,
+      "step": 32112
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4241564367103063e-06,
+      "loss": 0.6132,
+      "step": 32113
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4236194194797078e-06,
+      "loss": 0.5112,
+      "step": 32114
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4230824356656254e-06,
+      "loss": 0.5866,
+      "step": 32115
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4225454852707875e-06,
+      "loss": 0.5982,
+      "step": 32116
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4220085682979178e-06,
+      "loss": 0.5774,
+      "step": 32117
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4214716847497533e-06,
+      "loss": 0.601,
+      "step": 32118
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.420934834629014e-06,
+      "loss": 0.5846,
+      "step": 32119
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4203980179384378e-06,
+      "loss": 0.5722,
+      "step": 32120
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4198612346807403e-06,
+      "loss": 0.6596,
+      "step": 32121
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4193244848586584e-06,
+      "loss": 0.6376,
+      "step": 32122
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4187877684749147e-06,
+      "loss": 0.5558,
+      "step": 32123
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.418251085532236e-06,
+      "loss": 0.5938,
+      "step": 32124
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.417714436033354e-06,
+      "loss": 0.6357,
+      "step": 32125
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4171778199809922e-06,
+      "loss": 0.6019,
+      "step": 32126
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.416641237377878e-06,
+      "loss": 0.5761,
+      "step": 32127
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4161046882267345e-06,
+      "loss": 0.5602,
+      "step": 32128
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.415568172530295e-06,
+      "loss": 0.5898,
+      "step": 32129
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.415031690291277e-06,
+      "loss": 0.6189,
+      "step": 32130
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.41449524151242e-06,
+      "loss": 0.6129,
+      "step": 32131
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.413958826196434e-06,
+      "loss": 0.6151,
+      "step": 32132
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.413422444346055e-06,
+      "loss": 0.5394,
+      "step": 32133
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.412886095964002e-06,
+      "loss": 0.5565,
+      "step": 32134
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.412349781053008e-06,
+      "loss": 0.5866,
+      "step": 32135
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.411813499615794e-06,
+      "loss": 0.5935,
+      "step": 32136
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.411277251655084e-06,
+      "loss": 0.6132,
+      "step": 32137
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4107410371736014e-06,
+      "loss": 0.595,
+      "step": 32138
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4102048561740774e-06,
+      "loss": 0.6297,
+      "step": 32139
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4096687086592285e-06,
+      "loss": 0.537,
+      "step": 32140
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4091325946317853e-06,
+      "loss": 0.5708,
+      "step": 32141
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.408596514094471e-06,
+      "loss": 0.6305,
+      "step": 32142
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4080604670500074e-06,
+      "loss": 0.6133,
+      "step": 32143
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4075244535011187e-06,
+      "loss": 0.6187,
+      "step": 32144
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4069884734505264e-06,
+      "loss": 0.5926,
+      "step": 32145
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4064525269009596e-06,
+      "loss": 0.5324,
+      "step": 32146
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4059166138551394e-06,
+      "loss": 0.5392,
+      "step": 32147
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4053807343157864e-06,
+      "loss": 0.6,
+      "step": 32148
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4048448882856235e-06,
+      "loss": 0.6365,
+      "step": 32149
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.404309075767377e-06,
+      "loss": 0.613,
+      "step": 32150
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.403773296763766e-06,
+      "loss": 0.5666,
+      "step": 32151
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.403237551277516e-06,
+      "loss": 0.6601,
+      "step": 32152
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.40270183931135e-06,
+      "loss": 0.6492,
+      "step": 32153
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4021661608679867e-06,
+      "loss": 0.5876,
+      "step": 32154
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4016305159501474e-06,
+      "loss": 0.6574,
+      "step": 32155
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.401094904560558e-06,
+      "loss": 0.5843,
+      "step": 32156
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.4005593267019342e-06,
+      "loss": 0.6285,
+      "step": 32157
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.400023782377009e-06,
+      "loss": 0.5733,
+      "step": 32158
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.399488271588489e-06,
+      "loss": 0.5783,
+      "step": 32159
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3989527943391055e-06,
+      "loss": 0.606,
+      "step": 32160
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.398417350631572e-06,
+      "loss": 0.5621,
+      "step": 32161
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.397881940468617e-06,
+      "loss": 0.611,
+      "step": 32162
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.397346563852958e-06,
+      "loss": 0.5863,
+      "step": 32163
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.396811220787314e-06,
+      "loss": 0.6177,
+      "step": 32164
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.396275911274407e-06,
+      "loss": 0.6862,
+      "step": 32165
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.395740635316952e-06,
+      "loss": 0.6018,
+      "step": 32166
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3952053929176765e-06,
+      "loss": 0.6252,
+      "step": 32167
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.394670184079293e-06,
+      "loss": 0.5884,
+      "step": 32168
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.394135008804532e-06,
+      "loss": 0.6097,
+      "step": 32169
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.393599867096098e-06,
+      "loss": 0.561,
+      "step": 32170
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3930647589567225e-06,
+      "loss": 0.5494,
+      "step": 32171
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3925296843891153e-06,
+      "loss": 0.5931,
+      "step": 32172
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3919946433960036e-06,
+      "loss": 0.6607,
+      "step": 32173
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3914596359801034e-06,
+      "loss": 0.5786,
+      "step": 32174
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3909246621441314e-06,
+      "loss": 0.5721,
+      "step": 32175
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3903897218908034e-06,
+      "loss": 0.5726,
+      "step": 32176
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.389854815222844e-06,
+      "loss": 0.581,
+      "step": 32177
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3893199421429644e-06,
+      "loss": 0.5872,
+      "step": 32178
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.388785102653891e-06,
+      "loss": 0.5008,
+      "step": 32179
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3882502967583354e-06,
+      "loss": 0.6338,
+      "step": 32180
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3877155244590175e-06,
+      "loss": 0.5571,
+      "step": 32181
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3871807857586527e-06,
+      "loss": 0.5597,
+      "step": 32182
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3866460806599553e-06,
+      "loss": 0.6092,
+      "step": 32183
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.38611140916565e-06,
+      "loss": 0.5478,
+      "step": 32184
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3855767712784493e-06,
+      "loss": 0.5999,
+      "step": 32185
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.38504216700107e-06,
+      "loss": 0.5915,
+      "step": 32186
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3845075963362263e-06,
+      "loss": 0.5802,
+      "step": 32187
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3839730592866393e-06,
+      "loss": 0.6236,
+      "step": 32188
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.38343855585502e-06,
+      "loss": 0.6132,
+      "step": 32189
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.382904086044091e-06,
+      "loss": 0.5946,
+      "step": 32190
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.382369649856563e-06,
+      "loss": 0.5768,
+      "step": 32191
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3818352472951533e-06,
+      "loss": 0.6031,
+      "step": 32192
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3813008783625734e-06,
+      "loss": 0.6015,
+      "step": 32193
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3807665430615464e-06,
+      "loss": 0.6137,
+      "step": 32194
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3802322413947786e-06,
+      "loss": 0.6439,
+      "step": 32195
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3796979733649973e-06,
+      "loss": 0.5865,
+      "step": 32196
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.379163738974902e-06,
+      "loss": 0.6654,
+      "step": 32197
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3786295382272182e-06,
+      "loss": 0.5622,
+      "step": 32198
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3780953711246533e-06,
+      "loss": 0.6231,
+      "step": 32199
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3775612376699285e-06,
+      "loss": 0.5955,
+      "step": 32200
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.377027137865755e-06,
+      "loss": 0.5812,
+      "step": 32201
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.376493071714846e-06,
+      "loss": 0.6706,
+      "step": 32202
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.375959039219916e-06,
+      "loss": 0.584,
+      "step": 32203
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3754250403836743e-06,
+      "loss": 0.6373,
+      "step": 32204
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.37489107520884e-06,
+      "loss": 0.589,
+      "step": 32205
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3743571436981236e-06,
+      "loss": 0.6594,
+      "step": 32206
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3738232458542443e-06,
+      "loss": 0.5866,
+      "step": 32207
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3732893816799026e-06,
+      "loss": 0.5528,
+      "step": 32208
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3727555511778222e-06,
+      "loss": 0.6247,
+      "step": 32209
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3722217543507085e-06,
+      "loss": 0.5636,
+      "step": 32210
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.37168799120128e-06,
+      "loss": 0.5481,
+      "step": 32211
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.371154261732247e-06,
+      "loss": 0.6321,
+      "step": 32212
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.37062056594632e-06,
+      "loss": 0.5603,
+      "step": 32213
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.370086903846207e-06,
+      "loss": 0.5945,
+      "step": 32214
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.369553275434628e-06,
+      "loss": 0.5587,
+      "step": 32215
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.369019680714288e-06,
+      "loss": 0.611,
+      "step": 32216
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3684861196879036e-06,
+      "loss": 0.5801,
+      "step": 32217
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3679525923581833e-06,
+      "loss": 0.6008,
+      "step": 32218
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3674190987278387e-06,
+      "loss": 0.6004,
+      "step": 32219
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3668856387995776e-06,
+      "loss": 0.5828,
+      "step": 32220
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.366352212576115e-06,
+      "loss": 0.5696,
+      "step": 32221
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3658188200601606e-06,
+      "loss": 0.5948,
+      "step": 32222
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3652854612544227e-06,
+      "loss": 0.6116,
+      "step": 32223
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.364752136161613e-06,
+      "loss": 0.5979,
+      "step": 32224
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.364218844784437e-06,
+      "loss": 0.6036,
+      "step": 32225
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3636855871256124e-06,
+      "loss": 0.5455,
+      "step": 32226
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3631523631878413e-06,
+      "loss": 0.5925,
+      "step": 32227
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3626191729738434e-06,
+      "loss": 0.6122,
+      "step": 32228
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3620860164863147e-06,
+      "loss": 0.5882,
+      "step": 32229
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.361552893727974e-06,
+      "loss": 0.5747,
+      "step": 32230
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.361019804701524e-06,
+      "loss": 0.6124,
+      "step": 32231
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3604867494096806e-06,
+      "loss": 0.5942,
+      "step": 32232
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3599537278551454e-06,
+      "loss": 0.5706,
+      "step": 32233
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3594207400406354e-06,
+      "loss": 0.5634,
+      "step": 32234
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3588877859688475e-06,
+      "loss": 0.5757,
+      "step": 32235
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.358354865642498e-06,
+      "loss": 0.5719,
+      "step": 32236
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3578219790642907e-06,
+      "loss": 0.601,
+      "step": 32237
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3572891262369377e-06,
+      "loss": 0.5995,
+      "step": 32238
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.356756307163144e-06,
+      "loss": 0.6063,
+      "step": 32239
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3562235218456174e-06,
+      "loss": 0.6306,
+      "step": 32240
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3556907702870613e-06,
+      "loss": 0.6291,
+      "step": 32241
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3551580524901894e-06,
+      "loss": 0.5827,
+      "step": 32242
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.354625368457706e-06,
+      "loss": 0.5244,
+      "step": 32243
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.354092718192313e-06,
+      "loss": 0.5318,
+      "step": 32244
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.353560101696729e-06,
+      "loss": 0.6312,
+      "step": 32245
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.353027518973646e-06,
+      "loss": 0.5329,
+      "step": 32246
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3524949700257793e-06,
+      "loss": 0.6499,
+      "step": 32247
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.351962454855829e-06,
+      "loss": 0.5874,
+      "step": 32248
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.351429973466509e-06,
+      "loss": 0.5468,
+      "step": 32249
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.350897525860519e-06,
+      "loss": 0.6495,
+      "step": 32250
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3503651120405676e-06,
+      "loss": 0.6386,
+      "step": 32251
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3498327320093536e-06,
+      "loss": 0.6207,
+      "step": 32252
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.3493003857695916e-06,
+      "loss": 0.5953,
+      "step": 32253
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.348768073323978e-06,
+      "loss": 0.5747,
+      "step": 32254
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.348235794675229e-06,
+      "loss": 0.6086,
+      "step": 32255
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.347703549826035e-06,
+      "loss": 0.5959,
+      "step": 32256
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3471713387791106e-06,
+      "loss": 0.6653,
+      "step": 32257
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.346639161537153e-06,
+      "loss": 0.5206,
+      "step": 32258
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3461070181028754e-06,
+      "loss": 0.5643,
+      "step": 32259
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3455749084789758e-06,
+      "loss": 0.5619,
+      "step": 32260
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3450428326681583e-06,
+      "loss": 0.6552,
+      "step": 32261
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3445107906731276e-06,
+      "loss": 0.5758,
+      "step": 32262
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.343978782496584e-06,
+      "loss": 0.6319,
+      "step": 32263
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3434468081412364e-06,
+      "loss": 0.6201,
+      "step": 32264
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.342914867609781e-06,
+      "loss": 0.5546,
+      "step": 32265
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.342382960904933e-06,
+      "loss": 0.5904,
+      "step": 32266
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.34185108802938e-06,
+      "loss": 0.6064,
+      "step": 32267
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3413192489858337e-06,
+      "loss": 0.5715,
+      "step": 32268
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3407874437769915e-06,
+      "loss": 0.5621,
+      "step": 32269
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3402556724055625e-06,
+      "loss": 0.6505,
+      "step": 32270
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3397239348742407e-06,
+      "loss": 0.5807,
+      "step": 32271
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3391922311857395e-06,
+      "loss": 0.5922,
+      "step": 32272
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.338660561342746e-06,
+      "loss": 0.5559,
+      "step": 32273
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.338128925347972e-06,
+      "loss": 0.6201,
+      "step": 32274
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3375973232041135e-06,
+      "loss": 0.6057,
+      "step": 32275
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3370657549138773e-06,
+      "loss": 0.5979,
+      "step": 32276
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3365342204799613e-06,
+      "loss": 0.6071,
+      "step": 32277
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.336002719905067e-06,
+      "loss": 0.6397,
+      "step": 32278
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3354712531918897e-06,
+      "loss": 0.5642,
+      "step": 32279
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3349398203431395e-06,
+      "loss": 0.5678,
+      "step": 32280
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3344084213615113e-06,
+      "loss": 0.5906,
+      "step": 32281
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3338770562497038e-06,
+      "loss": 0.6565,
+      "step": 32282
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.333345725010424e-06,
+      "loss": 0.5854,
+      "step": 32283
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3328144276463615e-06,
+      "loss": 0.6316,
+      "step": 32284
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.332283164160225e-06,
+      "loss": 0.5804,
+      "step": 32285
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3317519345547066e-06,
+      "loss": 0.538,
+      "step": 32286
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3312207388325136e-06,
+      "loss": 0.6141,
+      "step": 32287
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.33068957699634e-06,
+      "loss": 0.6164,
+      "step": 32288
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.330158449048886e-06,
+      "loss": 0.6196,
+      "step": 32289
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.329627354992847e-06,
+      "loss": 0.5901,
+      "step": 32290
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.329096294830929e-06,
+      "loss": 0.529,
+      "step": 32291
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.328565268565823e-06,
+      "loss": 0.6709,
+      "step": 32292
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3280342762002373e-06,
+      "loss": 0.566,
+      "step": 32293
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3275033177368555e-06,
+      "loss": 0.5723,
+      "step": 32294
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3269723931783883e-06,
+      "loss": 0.6357,
+      "step": 32295
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.326441502527524e-06,
+      "loss": 0.5574,
+      "step": 32296
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3259106457869696e-06,
+      "loss": 0.6573,
+      "step": 32297
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.325379822959417e-06,
+      "loss": 0.5681,
+      "step": 32298
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.324849034047565e-06,
+      "loss": 0.5967,
+      "step": 32299
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.324318279054105e-06,
+      "loss": 0.6259,
+      "step": 32300
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3237875579817437e-06,
+      "loss": 0.5474,
+      "step": 32301
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3232568708331726e-06,
+      "loss": 0.6337,
+      "step": 32302
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3227262176110852e-06,
+      "loss": 0.6004,
+      "step": 32303
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3221955983181884e-06,
+      "loss": 0.6039,
+      "step": 32304
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3216650129571646e-06,
+      "loss": 0.6304,
+      "step": 32305
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3211344615307206e-06,
+      "loss": 0.586,
+      "step": 32306
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.320603944041544e-06,
+      "loss": 0.5693,
+      "step": 32307
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3200734604923392e-06,
+      "loss": 0.6476,
+      "step": 32308
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.319543010885794e-06,
+      "loss": 0.6534,
+      "step": 32309
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.319012595224613e-06,
+      "loss": 0.588,
+      "step": 32310
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3184822135114802e-06,
+      "loss": 0.6419,
+      "step": 32311
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.317951865749098e-06,
+      "loss": 0.644,
+      "step": 32312
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3174215519401576e-06,
+      "loss": 0.6397,
+      "step": 32313
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.316891272087357e-06,
+      "loss": 0.6212,
+      "step": 32314
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.31636102619339e-06,
+      "loss": 0.5919,
+      "step": 32315
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3158308142609508e-06,
+      "loss": 0.6024,
+      "step": 32316
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3153006362927274e-06,
+      "loss": 0.5915,
+      "step": 32317
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3147704922914248e-06,
+      "loss": 0.6298,
+      "step": 32318
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.31424038225973e-06,
+      "loss": 0.5885,
+      "step": 32319
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3137103062003373e-06,
+      "loss": 0.6257,
+      "step": 32320
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3131802641159384e-06,
+      "loss": 0.6771,
+      "step": 32321
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3126502560092323e-06,
+      "loss": 0.5941,
+      "step": 32322
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.312120281882908e-06,
+      "loss": 0.6409,
+      "step": 32323
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3115903417396556e-06,
+      "loss": 0.6224,
+      "step": 32324
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.311060435582175e-06,
+      "loss": 0.585,
+      "step": 32325
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.310530563413156e-06,
+      "loss": 0.6014,
+      "step": 32326
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3100007252352906e-06,
+      "loss": 0.6264,
+      "step": 32327
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3094709210512664e-06,
+      "loss": 0.6245,
+      "step": 32328
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3089411508637836e-06,
+      "loss": 0.5813,
+      "step": 32329
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.308411414675526e-06,
+      "loss": 0.6415,
+      "step": 32330
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3078817124891982e-06,
+      "loss": 0.5831,
+      "step": 32331
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.307352044307476e-06,
+      "loss": 0.5603,
+      "step": 32332
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3068224101330604e-06,
+      "loss": 0.5835,
+      "step": 32333
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3062928099686386e-06,
+      "loss": 0.5907,
+      "step": 32334
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.305763243816905e-06,
+      "loss": 0.5703,
+      "step": 32335
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.30523371168055e-06,
+      "loss": 0.5864,
+      "step": 32336
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3047042135622633e-06,
+      "loss": 0.5828,
+      "step": 32337
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3041747494647312e-06,
+      "loss": 0.5721,
+      "step": 32338
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3036453193906526e-06,
+      "loss": 0.5621,
+      "step": 32339
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3031159233427125e-06,
+      "loss": 0.5477,
+      "step": 32340
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3025865613235986e-06,
+      "loss": 0.6057,
+      "step": 32341
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.302057233336007e-06,
+      "loss": 0.6182,
+      "step": 32342
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3015279393826237e-06,
+      "loss": 0.6707,
+      "step": 32343
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.3009986794661385e-06,
+      "loss": 0.5945,
+      "step": 32344
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.300469453589239e-06,
+      "loss": 0.5969,
+      "step": 32345
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2999402617546185e-06,
+      "loss": 0.6013,
+      "step": 32346
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2994111039649648e-06,
+      "loss": 0.6028,
+      "step": 32347
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.298881980222964e-06,
+      "loss": 0.5687,
+      "step": 32348
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2983528905313044e-06,
+      "loss": 0.6222,
+      "step": 32349
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.297823834892678e-06,
+      "loss": 0.5514,
+      "step": 32350
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2972948133097704e-06,
+      "loss": 0.5942,
+      "step": 32351
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.296765825785273e-06,
+      "loss": 0.6088,
+      "step": 32352
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.296236872321872e-06,
+      "loss": 0.6059,
+      "step": 32353
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2957079529222547e-06,
+      "loss": 0.5831,
+      "step": 32354
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.295179067589105e-06,
+      "loss": 0.6167,
+      "step": 32355
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.294650216325117e-06,
+      "loss": 0.5682,
+      "step": 32356
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.294121399132976e-06,
+      "loss": 0.6274,
+      "step": 32357
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2935926160153673e-06,
+      "loss": 0.5891,
+      "step": 32358
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.293063866974976e-06,
+      "loss": 0.5938,
+      "step": 32359
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2925351520144943e-06,
+      "loss": 0.5289,
+      "step": 32360
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2920064711366064e-06,
+      "loss": 0.6224,
+      "step": 32361
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.291477824343994e-06,
+      "loss": 0.5929,
+      "step": 32362
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2909492116393506e-06,
+      "loss": 0.5604,
+      "step": 32363
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.29042063302536e-06,
+      "loss": 0.6264,
+      "step": 32364
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.289892088504707e-06,
+      "loss": 0.5897,
+      "step": 32365
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.289363578080074e-06,
+      "loss": 0.5681,
+      "step": 32366
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.288835101754153e-06,
+      "loss": 0.5858,
+      "step": 32367
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.288306659529623e-06,
+      "loss": 0.6314,
+      "step": 32368
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2877782514091793e-06,
+      "loss": 0.6204,
+      "step": 32369
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2872498773954933e-06,
+      "loss": 0.6157,
+      "step": 32370
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.28672153749126e-06,
+      "loss": 0.6304,
+      "step": 32371
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.286193231699157e-06,
+      "loss": 0.6097,
+      "step": 32372
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.285664960021876e-06,
+      "loss": 0.6331,
+      "step": 32373
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2851367224620977e-06,
+      "loss": 0.5445,
+      "step": 32374
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2846085190225065e-06,
+      "loss": 0.5964,
+      "step": 32375
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.284080349705783e-06,
+      "loss": 0.6228,
+      "step": 32376
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.283552214514617e-06,
+      "loss": 0.6357,
+      "step": 32377
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2830241134516895e-06,
+      "loss": 0.5841,
+      "step": 32378
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.28249604651968e-06,
+      "loss": 0.6021,
+      "step": 32379
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.281968013721278e-06,
+      "loss": 0.6753,
+      "step": 32380
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2814400150591653e-06,
+      "loss": 0.6236,
+      "step": 32381
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2809120505360224e-06,
+      "loss": 0.6038,
+      "step": 32382
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.280384120154532e-06,
+      "loss": 0.5563,
+      "step": 32383
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2798562239173793e-06,
+      "loss": 0.5658,
+      "step": 32384
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2793283618272454e-06,
+      "loss": 0.6129,
+      "step": 32385
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.278800533886812e-06,
+      "loss": 0.5443,
+      "step": 32386
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.278272740098758e-06,
+      "loss": 0.5597,
+      "step": 32387
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2777449804657733e-06,
+      "loss": 0.5791,
+      "step": 32388
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.277217254990531e-06,
+      "loss": 0.6359,
+      "step": 32389
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.27668956367572e-06,
+      "loss": 0.624,
+      "step": 32390
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.276161906524018e-06,
+      "loss": 0.5848,
+      "step": 32391
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2756342835381073e-06,
+      "loss": 0.6084,
+      "step": 32392
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2751066947206647e-06,
+      "loss": 0.5576,
+      "step": 32393
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2745791400743776e-06,
+      "loss": 0.5727,
+      "step": 32394
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2740516196019233e-06,
+      "loss": 0.6497,
+      "step": 32395
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.273524133305982e-06,
+      "loss": 0.5973,
+      "step": 32396
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2729966811892322e-06,
+      "loss": 0.5511,
+      "step": 32397
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2724692632543598e-06,
+      "loss": 0.5497,
+      "step": 32398
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2719418795040405e-06,
+      "loss": 0.6593,
+      "step": 32399
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.2714145299409518e-06,
+      "loss": 0.5486,
+      "step": 32400
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2708872145677796e-06,
+      "loss": 0.5713,
+      "step": 32401
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2703599333872007e-06,
+      "loss": 0.5795,
+      "step": 32402
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2698326864018936e-06,
+      "loss": 0.6181,
+      "step": 32403
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2693054736145347e-06,
+      "loss": 0.6094,
+      "step": 32404
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.268778295027809e-06,
+      "loss": 0.5776,
+      "step": 32405
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.268251150644388e-06,
+      "loss": 0.5123,
+      "step": 32406
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.267724040466962e-06,
+      "loss": 0.6149,
+      "step": 32407
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.267196964498195e-06,
+      "loss": 0.5398,
+      "step": 32408
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.266669922740775e-06,
+      "loss": 0.6643,
+      "step": 32409
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2661429151973734e-06,
+      "loss": 0.57,
+      "step": 32410
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2656159418706756e-06,
+      "loss": 0.6035,
+      "step": 32411
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2650890027633554e-06,
+      "loss": 0.6297,
+      "step": 32412
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.26456209787809e-06,
+      "loss": 0.5984,
+      "step": 32413
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2640352272175545e-06,
+      "loss": 0.6694,
+      "step": 32414
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2635083907844325e-06,
+      "loss": 0.6048,
+      "step": 32415
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2629815885813963e-06,
+      "loss": 0.6158,
+      "step": 32416
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2624548206111204e-06,
+      "loss": 0.5864,
+      "step": 32417
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2619280868762883e-06,
+      "loss": 0.6084,
+      "step": 32418
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2614013873795723e-06,
+      "loss": 0.5778,
+      "step": 32419
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2608747221236503e-06,
+      "loss": 0.6028,
+      "step": 32420
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.260348091111194e-06,
+      "loss": 0.6295,
+      "step": 32421
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.259821494344886e-06,
+      "loss": 0.6315,
+      "step": 32422
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.259294931827398e-06,
+      "loss": 0.612,
+      "step": 32423
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2587684035614074e-06,
+      "loss": 0.6363,
+      "step": 32424
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2582419095495853e-06,
+      "loss": 0.6049,
+      "step": 32425
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.257715449794614e-06,
+      "loss": 0.5593,
+      "step": 32426
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.257189024299162e-06,
+      "loss": 0.6041,
+      "step": 32427
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.25666263306591e-06,
+      "loss": 0.6497,
+      "step": 32428
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.256136276097531e-06,
+      "loss": 0.638,
+      "step": 32429
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.255609953396698e-06,
+      "loss": 0.6497,
+      "step": 32430
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.255083664966082e-06,
+      "loss": 0.5459,
+      "step": 32431
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2545574108083655e-06,
+      "loss": 0.5499,
+      "step": 32432
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2540311909262157e-06,
+      "loss": 0.6315,
+      "step": 32433
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.253505005322314e-06,
+      "loss": 0.6857,
+      "step": 32434
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2529788539993234e-06,
+      "loss": 0.5627,
+      "step": 32435
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2524527369599266e-06,
+      "loss": 0.5829,
+      "step": 32436
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.251926654206793e-06,
+      "loss": 0.623,
+      "step": 32437
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.251400605742594e-06,
+      "loss": 0.6172,
+      "step": 32438
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.250874591570007e-06,
+      "loss": 0.5902,
+      "step": 32439
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.250348611691704e-06,
+      "loss": 0.5803,
+      "step": 32440
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.249822666110355e-06,
+      "loss": 0.5438,
+      "step": 32441
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2492967548286314e-06,
+      "loss": 0.5551,
+      "step": 32442
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.248770877849211e-06,
+      "loss": 0.5722,
+      "step": 32443
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2482450351747608e-06,
+      "loss": 0.5552,
+      "step": 32444
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2477192268079605e-06,
+      "loss": 0.6019,
+      "step": 32445
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2471934527514702e-06,
+      "loss": 0.5987,
+      "step": 32446
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2466677130079704e-06,
+      "loss": 0.6312,
+      "step": 32447
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2461420075801263e-06,
+      "loss": 0.6266,
+      "step": 32448
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.245616336470616e-06,
+      "loss": 0.573,
+      "step": 32449
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2450906996821087e-06,
+      "loss": 0.6449,
+      "step": 32450
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2445650972172726e-06,
+      "loss": 0.6491,
+      "step": 32451
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.244039529078776e-06,
+      "loss": 0.5618,
+      "step": 32452
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2435139952692963e-06,
+      "loss": 0.5982,
+      "step": 32453
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2429884957915015e-06,
+      "loss": 0.5679,
+      "step": 32454
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2424630306480575e-06,
+      "loss": 0.6599,
+      "step": 32455
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2419375998416415e-06,
+      "loss": 0.5783,
+      "step": 32456
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2414122033749196e-06,
+      "loss": 0.5499,
+      "step": 32457
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.240886841250561e-06,
+      "loss": 0.5711,
+      "step": 32458
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2403615134712328e-06,
+      "loss": 0.5546,
+      "step": 32459
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2398362200396115e-06,
+      "loss": 0.5438,
+      "step": 32460
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.239310960958362e-06,
+      "loss": 0.6033,
+      "step": 32461
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.238785736230152e-06,
+      "loss": 0.5959,
+      "step": 32462
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.23826054585765e-06,
+      "loss": 0.6384,
+      "step": 32463
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2377353898435295e-06,
+      "loss": 0.5756,
+      "step": 32464
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2372102681904516e-06,
+      "loss": 0.5642,
+      "step": 32465
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2366851809010934e-06,
+      "loss": 0.5884,
+      "step": 32466
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2361601279781184e-06,
+      "loss": 0.5777,
+      "step": 32467
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2356351094241943e-06,
+      "loss": 0.5736,
+      "step": 32468
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.235110125241986e-06,
+      "loss": 0.5608,
+      "step": 32469
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.234585175434167e-06,
+      "loss": 0.6106,
+      "step": 32470
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2340602600034e-06,
+      "loss": 0.6039,
+      "step": 32471
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2335353789523605e-06,
+      "loss": 0.6375,
+      "step": 32472
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2330105322837026e-06,
+      "loss": 0.5996,
+      "step": 32473
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2324857200001036e-06,
+      "loss": 0.527,
+      "step": 32474
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2319609421042263e-06,
+      "loss": 0.58,
+      "step": 32475
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.231436198598734e-06,
+      "loss": 0.5971,
+      "step": 32476
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2309114894863e-06,
+      "loss": 0.6176,
+      "step": 32477
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2303868147695873e-06,
+      "loss": 0.5426,
+      "step": 32478
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2298621744512627e-06,
+      "loss": 0.5845,
+      "step": 32479
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2293375685339866e-06,
+      "loss": 0.651,
+      "step": 32480
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.228812997020433e-06,
+      "loss": 0.5846,
+      "step": 32481
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.22828845991326e-06,
+      "loss": 0.5201,
+      "step": 32482
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2277639572151432e-06,
+      "loss": 0.594,
+      "step": 32483
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.227239488928735e-06,
+      "loss": 0.6186,
+      "step": 32484
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2267150550567085e-06,
+      "loss": 0.5701,
+      "step": 32485
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2261906556017244e-06,
+      "loss": 0.6084,
+      "step": 32486
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2256662905664526e-06,
+      "loss": 0.5987,
+      "step": 32487
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.225141959953554e-06,
+      "loss": 0.5778,
+      "step": 32488
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.224617663765692e-06,
+      "loss": 0.6272,
+      "step": 32489
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2240934020055305e-06,
+      "loss": 0.6114,
+      "step": 32490
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.223569174675738e-06,
+      "loss": 0.6423,
+      "step": 32491
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.223044981778971e-06,
+      "loss": 0.5874,
+      "step": 32492
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2225208233179007e-06,
+      "loss": 0.6253,
+      "step": 32493
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.221996699295187e-06,
+      "loss": 0.566,
+      "step": 32494
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.221472609713493e-06,
+      "loss": 0.6302,
+      "step": 32495
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2209485545754815e-06,
+      "loss": 0.5986,
+      "step": 32496
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2204245338838137e-06,
+      "loss": 0.6224,
+      "step": 32497
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.219900547641156e-06,
+      "loss": 0.5937,
+      "step": 32498
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.21937659585017e-06,
+      "loss": 0.6473,
+      "step": 32499
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.218852678513518e-06,
+      "loss": 0.5427,
+      "step": 32500
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2183287956338573e-06,
+      "loss": 0.6609,
+      "step": 32501
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2178049472138574e-06,
+      "loss": 0.5609,
+      "step": 32502
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.217281133256174e-06,
+      "loss": 0.5679,
+      "step": 32503
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.216757353763478e-06,
+      "loss": 0.6354,
+      "step": 32504
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.216233608738418e-06,
+      "loss": 0.662,
+      "step": 32505
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2157098981836644e-06,
+      "loss": 0.5732,
+      "step": 32506
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.215186222101874e-06,
+      "loss": 0.585,
+      "step": 32507
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.214662580495711e-06,
+      "loss": 0.5335,
+      "step": 32508
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2141389733678318e-06,
+      "loss": 0.5629,
+      "step": 32509
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2136154007209063e-06,
+      "loss": 0.644,
+      "step": 32510
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2130918625575824e-06,
+      "loss": 0.6193,
+      "step": 32511
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.212568358880529e-06,
+      "loss": 0.5892,
+      "step": 32512
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2120448896924007e-06,
+      "loss": 0.6682,
+      "step": 32513
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2115214549958628e-06,
+      "loss": 0.5626,
+      "step": 32514
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.210998054793574e-06,
+      "loss": 0.6241,
+      "step": 32515
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.210474689088191e-06,
+      "loss": 0.6741,
+      "step": 32516
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2099513578823748e-06,
+      "loss": 0.5364,
+      "step": 32517
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.209428061178781e-06,
+      "loss": 0.5931,
+      "step": 32518
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2089047989800757e-06,
+      "loss": 0.6588,
+      "step": 32519
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.208381571288911e-06,
+      "loss": 0.5971,
+      "step": 32520
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.207858378107954e-06,
+      "loss": 0.6274,
+      "step": 32521
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.207335219439852e-06,
+      "loss": 0.6113,
+      "step": 32522
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2068120952872715e-06,
+      "loss": 0.5953,
+      "step": 32523
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2062890056528663e-06,
+      "loss": 0.6224,
+      "step": 32524
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.205765950539298e-06,
+      "loss": 0.6189,
+      "step": 32525
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2052429299492237e-06,
+      "loss": 0.5887,
+      "step": 32526
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2047199438852994e-06,
+      "loss": 0.6115,
+      "step": 32527
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.20419699235018e-06,
+      "loss": 0.6185,
+      "step": 32528
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.20367407534653e-06,
+      "loss": 0.5559,
+      "step": 32529
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2031511928769985e-06,
+      "loss": 0.5993,
+      "step": 32530
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.202628344944253e-06,
+      "loss": 0.6481,
+      "step": 32531
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2021055315509374e-06,
+      "loss": 0.6297,
+      "step": 32532
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2015827526997166e-06,
+      "loss": 0.6133,
+      "step": 32533
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.201060008393243e-06,
+      "loss": 0.6276,
+      "step": 32534
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.2005372986341775e-06,
+      "loss": 0.6095,
+      "step": 32535
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.200014623425173e-06,
+      "loss": 0.5893,
+      "step": 32536
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1994919827688852e-06,
+      "loss": 0.6113,
+      "step": 32537
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1989693766679707e-06,
+      "loss": 0.6028,
+      "step": 32538
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1984468051250818e-06,
+      "loss": 0.5345,
+      "step": 32539
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.197924268142879e-06,
+      "loss": 0.5983,
+      "step": 32540
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.197401765724012e-06,
+      "loss": 0.5637,
+      "step": 32541
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.196879297871146e-06,
+      "loss": 0.5779,
+      "step": 32542
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1963568645869212e-06,
+      "loss": 0.5765,
+      "step": 32543
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.195834465874005e-06,
+      "loss": 0.5781,
+      "step": 32544
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1953121017350407e-06,
+      "loss": 0.6252,
+      "step": 32545
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.1947897721726927e-06,
+      "loss": 0.6052,
+      "step": 32546
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.194267477189609e-06,
+      "loss": 0.5674,
+      "step": 32547
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1937452167884497e-06,
+      "loss": 0.5982,
+      "step": 32548
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1932229909718593e-06,
+      "loss": 0.6379,
+      "step": 32549
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1927007997424985e-06,
+      "loss": 0.6085,
+      "step": 32550
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.192178643103017e-06,
+      "loss": 0.6519,
+      "step": 32551
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1916565210560715e-06,
+      "loss": 0.6145,
+      "step": 32552
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.191134433604313e-06,
+      "loss": 0.5688,
+      "step": 32553
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.190612380750394e-06,
+      "loss": 0.6562,
+      "step": 32554
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.190090362496969e-06,
+      "loss": 0.6457,
+      "step": 32555
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.189568378846686e-06,
+      "loss": 0.6145,
+      "step": 32556
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.189046429802203e-06,
+      "loss": 0.6348,
+      "step": 32557
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.188524515366167e-06,
+      "loss": 0.5677,
+      "step": 32558
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1880026355412385e-06,
+      "loss": 0.6655,
+      "step": 32559
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1874807903300586e-06,
+      "loss": 0.6072,
+      "step": 32560
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.186958979735286e-06,
+      "loss": 0.5459,
+      "step": 32561
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.186437203759567e-06,
+      "loss": 0.6721,
+      "step": 32562
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.18591546240556e-06,
+      "loss": 0.6342,
+      "step": 32563
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.185393755675912e-06,
+      "loss": 0.5699,
+      "step": 32564
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1848720835732726e-06,
+      "loss": 0.5892,
+      "step": 32565
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.184350446100293e-06,
+      "loss": 0.6115,
+      "step": 32566
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1838288432596266e-06,
+      "loss": 0.5745,
+      "step": 32567
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1833072750539186e-06,
+      "loss": 0.5782,
+      "step": 32568
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.18278574148583e-06,
+      "loss": 0.5709,
+      "step": 32569
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.182264242557996e-06,
+      "loss": 0.6096,
+      "step": 32570
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.181742778273079e-06,
+      "loss": 0.618,
+      "step": 32571
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1812213486337196e-06,
+      "loss": 0.5597,
+      "step": 32572
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1806999536425753e-06,
+      "loss": 0.6592,
+      "step": 32573
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.180178593302291e-06,
+      "loss": 0.5864,
+      "step": 32574
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1796572676155167e-06,
+      "loss": 0.5799,
+      "step": 32575
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1791359765849006e-06,
+      "loss": 0.5875,
+      "step": 32576
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.17861472021309e-06,
+      "loss": 0.569,
+      "step": 32577
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.178093498502739e-06,
+      "loss": 0.6303,
+      "step": 32578
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1775723114564884e-06,
+      "loss": 0.6603,
+      "step": 32579
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1770511590769983e-06,
+      "loss": 0.5561,
+      "step": 32580
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.176530041366903e-06,
+      "loss": 0.6096,
+      "step": 32581
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.17600895832886e-06,
+      "loss": 0.5883,
+      "step": 32582
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1754879099655113e-06,
+      "loss": 0.5789,
+      "step": 32583
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.174966896279509e-06,
+      "loss": 0.5959,
+      "step": 32584
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1744459172734964e-06,
+      "loss": 0.5846,
+      "step": 32585
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1739249729501297e-06,
+      "loss": 0.6428,
+      "step": 32586
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1734040633120423e-06,
+      "loss": 0.566,
+      "step": 32587
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1728831883618916e-06,
+      "loss": 0.5298,
+      "step": 32588
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.172362348102318e-06,
+      "loss": 0.6289,
+      "step": 32589
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.171841542535974e-06,
+      "loss": 0.6039,
+      "step": 32590
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1713207716655024e-06,
+      "loss": 0.5884,
+      "step": 32591
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1708000354935496e-06,
+      "loss": 0.5844,
+      "step": 32592
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.170279334022759e-06,
+      "loss": 0.5934,
+      "step": 32593
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1697586672557834e-06,
+      "loss": 0.5547,
+      "step": 32594
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.169238035195263e-06,
+      "loss": 0.6159,
+      "step": 32595
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.168717437843846e-06,
+      "loss": 0.6267,
+      "step": 32596
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1681968752041745e-06,
+      "loss": 0.6441,
+      "step": 32597
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.167676347278893e-06,
+      "loss": 0.5838,
+      "step": 32598
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.167155854070653e-06,
+      "loss": 0.5902,
+      "step": 32599
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1666353955820904e-06,
+      "loss": 0.615,
+      "step": 32600
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.166114971815859e-06,
+      "loss": 0.555,
+      "step": 32601
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1655945827745984e-06,
+      "loss": 0.5976,
+      "step": 32602
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1650742284609537e-06,
+      "loss": 0.6392,
+      "step": 32603
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1645539088775645e-06,
+      "loss": 0.5885,
+      "step": 32604
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.164033624027083e-06,
+      "loss": 0.7078,
+      "step": 32605
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1635133739121448e-06,
+      "loss": 0.6385,
+      "step": 32606
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1629931585354044e-06,
+      "loss": 0.5356,
+      "step": 32607
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1624729778994913e-06,
+      "loss": 0.592,
+      "step": 32608
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.161952832007058e-06,
+      "loss": 0.6191,
+      "step": 32609
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.161432720860742e-06,
+      "loss": 0.5469,
+      "step": 32610
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1609126444631934e-06,
+      "loss": 0.5452,
+      "step": 32611
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1603926028170496e-06,
+      "loss": 0.5907,
+      "step": 32612
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1598725959249543e-06,
+      "loss": 0.6214,
+      "step": 32613
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1593526237895456e-06,
+      "loss": 0.6396,
+      "step": 32614
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.158832686413473e-06,
+      "loss": 0.5683,
+      "step": 32615
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1583127837993765e-06,
+      "loss": 0.5921,
+      "step": 32616
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1577929159498923e-06,
+      "loss": 0.5622,
+      "step": 32617
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1572730828676724e-06,
+      "loss": 0.5753,
+      "step": 32618
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.156753284555345e-06,
+      "loss": 0.6366,
+      "step": 32619
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1562335210155624e-06,
+      "loss": 0.6423,
+      "step": 32620
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1557137922509583e-06,
+      "loss": 0.6214,
+      "step": 32621
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1551940982641793e-06,
+      "loss": 0.6006,
+      "step": 32622
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1546744390578644e-06,
+      "loss": 0.6175,
+      "step": 32623
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1541548146346534e-06,
+      "loss": 0.571,
+      "step": 32624
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1536352249971836e-06,
+      "loss": 0.5937,
+      "step": 32625
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1531156701481014e-06,
+      "loss": 0.6071,
+      "step": 32626
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1525961500900404e-06,
+      "loss": 0.5575,
+      "step": 32627
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.152076664825646e-06,
+      "loss": 0.5262,
+      "step": 32628
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.151557214357557e-06,
+      "loss": 0.5968,
+      "step": 32629
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1510377986884113e-06,
+      "loss": 0.6049,
+      "step": 32630
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1505184178208435e-06,
+      "loss": 0.608,
+      "step": 32631
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1499990717575024e-06,
+      "loss": 0.5907,
+      "step": 32632
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.149479760501022e-06,
+      "loss": 0.559,
+      "step": 32633
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.148960484054041e-06,
+      "loss": 0.5434,
+      "step": 32634
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.148441242419198e-06,
+      "loss": 0.505,
+      "step": 32635
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1479220355991292e-06,
+      "loss": 0.6445,
+      "step": 32636
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1474028635964773e-06,
+      "loss": 0.5574,
+      "step": 32637
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.146883726413875e-06,
+      "loss": 0.6007,
+      "step": 32638
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.146364624053968e-06,
+      "loss": 0.6593,
+      "step": 32639
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.145845556519389e-06,
+      "loss": 0.6248,
+      "step": 32640
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1453265238127763e-06,
+      "loss": 0.6168,
+      "step": 32641
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1448075259367626e-06,
+      "loss": 0.5938,
+      "step": 32642
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1442885628939932e-06,
+      "loss": 0.5914,
+      "step": 32643
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.143769634687099e-06,
+      "loss": 0.58,
+      "step": 32644
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1432507413187253e-06,
+      "loss": 0.587,
+      "step": 32645
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1427318827914966e-06,
+      "loss": 0.6246,
+      "step": 32646
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.142213059108058e-06,
+      "loss": 0.5969,
+      "step": 32647
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1416942702710396e-06,
+      "loss": 0.6048,
+      "step": 32648
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.141175516283086e-06,
+      "loss": 0.5735,
+      "step": 32649
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.140656797146827e-06,
+      "loss": 0.6509,
+      "step": 32650
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1401381128649e-06,
+      "loss": 0.5844,
+      "step": 32651
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.139619463439937e-06,
+      "loss": 0.5723,
+      "step": 32652
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.13910084887458e-06,
+      "loss": 0.6263,
+      "step": 32653
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1385822691714605e-06,
+      "loss": 0.5983,
+      "step": 32654
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.138063724333211e-06,
+      "loss": 0.5728,
+      "step": 32655
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.137545214362475e-06,
+      "loss": 0.5351,
+      "step": 32656
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.137026739261877e-06,
+      "loss": 0.5691,
+      "step": 32657
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.136508299034058e-06,
+      "loss": 0.6279,
+      "step": 32658
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.135989893681647e-06,
+      "loss": 0.6201,
+      "step": 32659
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.135471523207285e-06,
+      "loss": 0.5898,
+      "step": 32660
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.134953187613603e-06,
+      "loss": 0.5869,
+      "step": 32661
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1344348869032337e-06,
+      "loss": 0.5941,
+      "step": 32662
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1339166210788086e-06,
+      "loss": 0.6454,
+      "step": 32663
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1333983901429675e-06,
+      "loss": 0.6483,
+      "step": 32664
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1328801940983368e-06,
+      "loss": 0.5849,
+      "step": 32665
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1323620329475556e-06,
+      "loss": 0.682,
+      "step": 32666
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.131843906693255e-06,
+      "loss": 0.6136,
+      "step": 32667
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1313258153380654e-06,
+      "loss": 0.5783,
+      "step": 32668
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1308077588846198e-06,
+      "loss": 0.61,
+      "step": 32669
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1302897373355532e-06,
+      "loss": 0.6664,
+      "step": 32670
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1297717506934975e-06,
+      "loss": 0.5529,
+      "step": 32671
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1292537989610827e-06,
+      "loss": 0.6137,
+      "step": 32672
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.128735882140939e-06,
+      "loss": 0.5049,
+      "step": 32673
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.128218000235703e-06,
+      "loss": 0.6066,
+      "step": 32674
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1277001532480046e-06,
+      "loss": 0.5589,
+      "step": 32675
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1271823411804715e-06,
+      "loss": 0.641,
+      "step": 32676
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.126664564035741e-06,
+      "loss": 0.5572,
+      "step": 32677
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1261468218164393e-06,
+      "loss": 0.6162,
+      "step": 32678
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1256291145251995e-06,
+      "loss": 0.6349,
+      "step": 32679
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.125111442164649e-06,
+      "loss": 0.6133,
+      "step": 32680
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1245938047374223e-06,
+      "loss": 0.6267,
+      "step": 32681
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.124076202246146e-06,
+      "loss": 0.5699,
+      "step": 32682
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.123558634693459e-06,
+      "loss": 0.5909,
+      "step": 32683
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1230411020819773e-06,
+      "loss": 0.6581,
+      "step": 32684
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1225236044143404e-06,
+      "loss": 0.595,
+      "step": 32685
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1220061416931737e-06,
+      "loss": 0.6178,
+      "step": 32686
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.121488713921109e-06,
+      "loss": 0.6063,
+      "step": 32687
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1209713211007765e-06,
+      "loss": 0.5963,
+      "step": 32688
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.120453963234803e-06,
+      "loss": 0.6204,
+      "step": 32689
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1199366403258137e-06,
+      "loss": 0.5975,
+      "step": 32690
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1194193523764448e-06,
+      "loss": 0.5988,
+      "step": 32691
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.1189020993893213e-06,
+      "loss": 0.6043,
+      "step": 32692
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.118384881367069e-06,
+      "loss": 0.5908,
+      "step": 32693
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.117867698312321e-06,
+      "loss": 0.6291,
+      "step": 32694
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1173505502277023e-06,
+      "loss": 0.6071,
+      "step": 32695
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1168334371158428e-06,
+      "loss": 0.5253,
+      "step": 32696
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.116316358979363e-06,
+      "loss": 0.5674,
+      "step": 32697
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.115799315820901e-06,
+      "loss": 0.6724,
+      "step": 32698
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.115282307643077e-06,
+      "loss": 0.6661,
+      "step": 32699
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1147653344485205e-06,
+      "loss": 0.5763,
+      "step": 32700
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.114248396239855e-06,
+      "loss": 0.6124,
+      "step": 32701
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.113731493019714e-06,
+      "loss": 0.6016,
+      "step": 32702
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1132146247907156e-06,
+      "loss": 0.6175,
+      "step": 32703
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1126977915554935e-06,
+      "loss": 0.6094,
+      "step": 32704
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1121809933166713e-06,
+      "loss": 0.5934,
+      "step": 32705
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1116642300768753e-06,
+      "loss": 0.6065,
+      "step": 32706
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1111475018387274e-06,
+      "loss": 0.619,
+      "step": 32707
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1106308086048585e-06,
+      "loss": 0.6124,
+      "step": 32708
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.110114150377893e-06,
+      "loss": 0.617,
+      "step": 32709
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.109597527160455e-06,
+      "loss": 0.6494,
+      "step": 32710
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.109080938955168e-06,
+      "loss": 0.5884,
+      "step": 32711
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1085643857646607e-06,
+      "loss": 0.5601,
+      "step": 32712
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1080478675915558e-06,
+      "loss": 0.5862,
+      "step": 32713
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1075313844384757e-06,
+      "loss": 0.5965,
+      "step": 32714
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1070149363080503e-06,
+      "loss": 0.6258,
+      "step": 32715
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1064985232029e-06,
+      "loss": 0.5742,
+      "step": 32716
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.10598214512565e-06,
+      "loss": 0.5625,
+      "step": 32717
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1054658020789206e-06,
+      "loss": 0.6674,
+      "step": 32718
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1049494940653425e-06,
+      "loss": 0.5501,
+      "step": 32719
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1044332210875307e-06,
+      "loss": 0.5836,
+      "step": 32720
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.103916983148121e-06,
+      "loss": 0.5911,
+      "step": 32721
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.103400780249721e-06,
+      "loss": 0.6504,
+      "step": 32722
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.102884612394965e-06,
+      "loss": 0.5991,
+      "step": 32723
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1023684795864694e-06,
+      "loss": 0.6293,
+      "step": 32724
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1018523818268632e-06,
+      "loss": 0.681,
+      "step": 32725
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.101336319118764e-06,
+      "loss": 0.6357,
+      "step": 32726
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.1008202914647967e-06,
+      "loss": 0.5873,
+      "step": 32727
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.100304298867578e-06,
+      "loss": 0.6008,
+      "step": 32728
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.099788341329737e-06,
+      "loss": 0.5326,
+      "step": 32729
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0992724188538925e-06,
+      "loss": 0.5928,
+      "step": 32730
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0987565314426627e-06,
+      "loss": 0.6361,
+      "step": 32731
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0982406790986752e-06,
+      "loss": 0.65,
+      "step": 32732
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0977248618245483e-06,
+      "loss": 0.6064,
+      "step": 32733
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0972090796229014e-06,
+      "loss": 0.5717,
+      "step": 32734
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0966933324963557e-06,
+      "loss": 0.6153,
+      "step": 32735
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0961776204475348e-06,
+      "loss": 0.5217,
+      "step": 32736
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.095661943479057e-06,
+      "loss": 0.6301,
+      "step": 32737
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0951463015935435e-06,
+      "loss": 0.6513,
+      "step": 32738
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.094630694793611e-06,
+      "loss": 0.6162,
+      "step": 32739
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0941151230818844e-06,
+      "loss": 0.5987,
+      "step": 32740
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.093599586460979e-06,
+      "loss": 0.6259,
+      "step": 32741
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0930840849335196e-06,
+      "loss": 0.6458,
+      "step": 32742
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.092568618502123e-06,
+      "loss": 0.6322,
+      "step": 32743
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.092053187169408e-06,
+      "loss": 0.6239,
+      "step": 32744
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0915377909379908e-06,
+      "loss": 0.6278,
+      "step": 32745
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0910224298104963e-06,
+      "loss": 0.6345,
+      "step": 32746
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0905071037895405e-06,
+      "loss": 0.6274,
+      "step": 32747
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0899918128777417e-06,
+      "loss": 0.5848,
+      "step": 32748
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.089476557077715e-06,
+      "loss": 0.545,
+      "step": 32749
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0889613363920847e-06,
+      "loss": 0.6458,
+      "step": 32750
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0884461508234654e-06,
+      "loss": 0.6055,
+      "step": 32751
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.087931000374473e-06,
+      "loss": 0.6805,
+      "step": 32752
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0874158850477297e-06,
+      "loss": 0.6327,
+      "step": 32753
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0869008048458513e-06,
+      "loss": 0.589,
+      "step": 32754
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.086385759771455e-06,
+      "loss": 0.6275,
+      "step": 32755
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0858707498271536e-06,
+      "loss": 0.5791,
+      "step": 32756
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0853557750155718e-06,
+      "loss": 0.5874,
+      "step": 32757
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0848408353393177e-06,
+      "loss": 0.6068,
+      "step": 32758
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0843259308010197e-06,
+      "loss": 0.5971,
+      "step": 32759
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.083811061403281e-06,
+      "loss": 0.6258,
+      "step": 32760
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0832962271487264e-06,
+      "loss": 0.5598,
+      "step": 32761
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.082781428039967e-06,
+      "loss": 0.6133,
+      "step": 32762
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0822666640796227e-06,
+      "loss": 0.5794,
+      "step": 32763
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0817519352703087e-06,
+      "loss": 0.609,
+      "step": 32764
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0812372416146386e-06,
+      "loss": 0.6025,
+      "step": 32765
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0807225831152256e-06,
+      "loss": 0.6197,
+      "step": 32766
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0802079597746903e-06,
+      "loss": 0.5679,
+      "step": 32767
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0796933715956447e-06,
+      "loss": 0.5941,
+      "step": 32768
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0791788185807014e-06,
+      "loss": 0.5858,
+      "step": 32769
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.07866430073248e-06,
+      "loss": 0.564,
+      "step": 32770
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0781498180535917e-06,
+      "loss": 0.6123,
+      "step": 32771
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0776353705466524e-06,
+      "loss": 0.5877,
+      "step": 32772
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0771209582142715e-06,
+      "loss": 0.5623,
+      "step": 32773
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0766065810590695e-06,
+      "loss": 0.5656,
+      "step": 32774
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.076092239083658e-06,
+      "loss": 0.6007,
+      "step": 32775
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0755779322906486e-06,
+      "loss": 0.6662,
+      "step": 32776
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.075063660682652e-06,
+      "loss": 0.6177,
+      "step": 32777
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0745494242622898e-06,
+      "loss": 0.5524,
+      "step": 32778
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0740352230321658e-06,
+      "loss": 0.5796,
+      "step": 32779
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0735210569949038e-06,
+      "loss": 0.6909,
+      "step": 32780
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0730069261531037e-06,
+      "loss": 0.6197,
+      "step": 32781
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0724928305093872e-06,
+      "loss": 0.6716,
+      "step": 32782
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0719787700663594e-06,
+      "loss": 0.6058,
+      "step": 32783
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0714647448266412e-06,
+      "loss": 0.6298,
+      "step": 32784
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0709507547928353e-06,
+      "loss": 0.6133,
+      "step": 32785
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0704367999675645e-06,
+      "loss": 0.5684,
+      "step": 32786
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0699228803534275e-06,
+      "loss": 0.551,
+      "step": 32787
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.069408995953046e-06,
+      "loss": 0.5925,
+      "step": 32788
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0688951467690254e-06,
+      "loss": 0.5875,
+      "step": 32789
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0683813328039767e-06,
+      "loss": 0.5274,
+      "step": 32790
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.067867554060515e-06,
+      "loss": 0.5728,
+      "step": 32791
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.067353810541248e-06,
+      "loss": 0.5948,
+      "step": 32792
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.066840102248787e-06,
+      "loss": 0.6435,
+      "step": 32793
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0663264291857385e-06,
+      "loss": 0.5852,
+      "step": 32794
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.06581279135472e-06,
+      "loss": 0.6291,
+      "step": 32795
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.065299188758334e-06,
+      "loss": 0.5852,
+      "step": 32796
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0647856213992e-06,
+      "loss": 0.5643,
+      "step": 32797
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.064272089279915e-06,
+      "loss": 0.5784,
+      "step": 32798
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.063758592403098e-06,
+      "loss": 0.5556,
+      "step": 32799
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0632451307713507e-06,
+      "loss": 0.5871,
+      "step": 32800
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0627317043872907e-06,
+      "loss": 0.5751,
+      "step": 32801
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.062218313253522e-06,
+      "loss": 0.5114,
+      "step": 32802
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0617049573726555e-06,
+      "loss": 0.6022,
+      "step": 32803
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.061191636747293e-06,
+      "loss": 0.6196,
+      "step": 32804
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.060678351380052e-06,
+      "loss": 0.6198,
+      "step": 32805
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0601651012735335e-06,
+      "loss": 0.5792,
+      "step": 32806
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0596518864303536e-06,
+      "loss": 0.5693,
+      "step": 32807
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0591387068531096e-06,
+      "loss": 0.6292,
+      "step": 32808
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.058625562544418e-06,
+      "loss": 0.5907,
+      "step": 32809
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.058112453506883e-06,
+      "loss": 0.5453,
+      "step": 32810
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.057599379743108e-06,
+      "loss": 0.5923,
+      "step": 32811
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.057086341255706e-06,
+      "loss": 0.6491,
+      "step": 32812
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.056573338047283e-06,
+      "loss": 0.6246,
+      "step": 32813
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0560603701204437e-06,
+      "loss": 0.6089,
+      "step": 32814
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.055547437477793e-06,
+      "loss": 0.5399,
+      "step": 32815
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0550345401219417e-06,
+      "loss": 0.6218,
+      "step": 32816
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.054521678055491e-06,
+      "loss": 0.5444,
+      "step": 32817
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.054008851281056e-06,
+      "loss": 0.5435,
+      "step": 32818
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.05349605980123e-06,
+      "loss": 0.6049,
+      "step": 32819
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0529833036186283e-06,
+      "loss": 0.6171,
+      "step": 32820
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.052470582735849e-06,
+      "loss": 0.6026,
+      "step": 32821
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0519578971555053e-06,
+      "loss": 0.5844,
+      "step": 32822
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.051445246880195e-06,
+      "loss": 0.5607,
+      "step": 32823
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.050932631912532e-06,
+      "loss": 0.5864,
+      "step": 32824
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.050420052255111e-06,
+      "loss": 0.6086,
+      "step": 32825
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0499075079105422e-06,
+      "loss": 0.5416,
+      "step": 32826
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0493949988814266e-06,
+      "loss": 0.559,
+      "step": 32827
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0488825251703737e-06,
+      "loss": 0.6222,
+      "step": 32828
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0483700867799835e-06,
+      "loss": 0.5818,
+      "step": 32829
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0478576837128616e-06,
+      "loss": 0.5998,
+      "step": 32830
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0473453159716105e-06,
+      "loss": 0.6021,
+      "step": 32831
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0468329835588318e-06,
+      "loss": 0.5801,
+      "step": 32832
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0463206864771335e-06,
+      "loss": 0.648,
+      "step": 32833
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.045808424729113e-06,
+      "loss": 0.6206,
+      "step": 32834
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0452961983173823e-06,
+      "loss": 0.5785,
+      "step": 32835
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0447840072445335e-06,
+      "loss": 0.6473,
+      "step": 32836
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0442718515131765e-06,
+      "loss": 0.5938,
+      "step": 32837
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.043759731125907e-06,
+      "loss": 0.5928,
+      "step": 32838
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.0432476460853366e-06,
+      "loss": 0.62,
+      "step": 32839
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.042735596394061e-06,
+      "loss": 0.5723,
+      "step": 32840
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.042223582054684e-06,
+      "loss": 0.5993,
+      "step": 32841
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0417116030698034e-06,
+      "loss": 0.6046,
+      "step": 32842
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0411996594420266e-06,
+      "loss": 0.5803,
+      "step": 32843
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0406877511739497e-06,
+      "loss": 0.5338,
+      "step": 32844
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0401758782681834e-06,
+      "loss": 0.6378,
+      "step": 32845
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.039664040727315e-06,
+      "loss": 0.5756,
+      "step": 32846
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.039152238553954e-06,
+      "loss": 0.5379,
+      "step": 32847
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0386404717507003e-06,
+      "loss": 0.5806,
+      "step": 32848
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.03812874032015e-06,
+      "loss": 0.5977,
+      "step": 32849
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.037617044264909e-06,
+      "loss": 0.601,
+      "step": 32850
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.037105383587575e-06,
+      "loss": 0.6115,
+      "step": 32851
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0365937582907477e-06,
+      "loss": 0.5432,
+      "step": 32852
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.036082168377024e-06,
+      "loss": 0.6517,
+      "step": 32853
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0355706138490093e-06,
+      "loss": 0.5922,
+      "step": 32854
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.035059094709296e-06,
+      "loss": 0.5961,
+      "step": 32855
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0345476109604943e-06,
+      "loss": 0.5583,
+      "step": 32856
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0340361626051896e-06,
+      "loss": 0.6532,
+      "step": 32857
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0335247496459907e-06,
+      "loss": 0.599,
+      "step": 32858
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.033013372085489e-06,
+      "loss": 0.6434,
+      "step": 32859
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.032502029926291e-06,
+      "loss": 0.5577,
+      "step": 32860
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0319907231709866e-06,
+      "loss": 0.633,
+      "step": 32861
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0314794518221846e-06,
+      "loss": 0.6085,
+      "step": 32862
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0309682158824706e-06,
+      "loss": 0.6043,
+      "step": 32863
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.03045701535445e-06,
+      "loss": 0.5761,
+      "step": 32864
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0299458502407163e-06,
+      "loss": 0.6226,
+      "step": 32865
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0294347205438713e-06,
+      "loss": 0.644,
+      "step": 32866
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.02892362626651e-06,
+      "loss": 0.5601,
+      "step": 32867
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.02841256741123e-06,
+      "loss": 0.5733,
+      "step": 32868
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.027901543980627e-06,
+      "loss": 0.5581,
+      "step": 32869
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0273905559772953e-06,
+      "loss": 0.6188,
+      "step": 32870
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0268796034038374e-06,
+      "loss": 0.62,
+      "step": 32871
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0263686862628462e-06,
+      "loss": 0.5854,
+      "step": 32872
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0258578045569186e-06,
+      "loss": 0.5979,
+      "step": 32873
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0253469582886463e-06,
+      "loss": 0.6314,
+      "step": 32874
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0248361474606315e-06,
+      "loss": 0.6007,
+      "step": 32875
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.024325372075464e-06,
+      "loss": 0.5801,
+      "step": 32876
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0238146321357454e-06,
+      "loss": 0.5634,
+      "step": 32877
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.023303927644068e-06,
+      "loss": 0.6329,
+      "step": 32878
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.022793258603026e-06,
+      "loss": 0.6039,
+      "step": 32879
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0222826250152125e-06,
+      "loss": 0.5814,
+      "step": 32880
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0217720268832275e-06,
+      "loss": 0.5995,
+      "step": 32881
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0212614642096595e-06,
+      "loss": 0.5754,
+      "step": 32882
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.020750936997111e-06,
+      "loss": 0.676,
+      "step": 32883
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.020240445248166e-06,
+      "loss": 0.6295,
+      "step": 32884
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0197299889654253e-06,
+      "loss": 0.6181,
+      "step": 32885
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0192195681514792e-06,
+      "loss": 0.6183,
+      "step": 32886
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.018709182808924e-06,
+      "loss": 0.6065,
+      "step": 32887
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0181988329403533e-06,
+      "loss": 0.5986,
+      "step": 32888
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0176885185483594e-06,
+      "loss": 0.5761,
+      "step": 32889
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0171782396355344e-06,
+      "loss": 0.5712,
+      "step": 32890
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.016667996204469e-06,
+      "loss": 0.5641,
+      "step": 32891
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.016157788257761e-06,
+      "loss": 0.5704,
+      "step": 32892
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0156476157979986e-06,
+      "loss": 0.6224,
+      "step": 32893
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.015137478827782e-06,
+      "loss": 0.5604,
+      "step": 32894
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0146273773496904e-06,
+      "loss": 0.6253,
+      "step": 32895
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.014117311366327e-06,
+      "loss": 0.5433,
+      "step": 32896
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.013607280880275e-06,
+      "loss": 0.5606,
+      "step": 32897
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0130972858941354e-06,
+      "loss": 0.5679,
+      "step": 32898
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.012587326410493e-06,
+      "loss": 0.5962,
+      "step": 32899
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0120774024319422e-06,
+      "loss": 0.6002,
+      "step": 32900
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.011567513961069e-06,
+      "loss": 0.6053,
+      "step": 32901
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0110576610004704e-06,
+      "loss": 0.5517,
+      "step": 32902
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0105478435527324e-06,
+      "loss": 0.5833,
+      "step": 32903
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0100380616204507e-06,
+      "loss": 0.5983,
+      "step": 32904
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0095283152062128e-06,
+      "loss": 0.5565,
+      "step": 32905
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.009018604312608e-06,
+      "loss": 0.5797,
+      "step": 32906
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.008508928942224e-06,
+      "loss": 0.6108,
+      "step": 32907
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0079992890976583e-06,
+      "loss": 0.603,
+      "step": 32908
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.007489684781495e-06,
+      "loss": 0.5837,
+      "step": 32909
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.006980115996324e-06,
+      "loss": 0.6078,
+      "step": 32910
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.006470582744735e-06,
+      "loss": 0.5748,
+      "step": 32911
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.005961085029314e-06,
+      "loss": 0.6097,
+      "step": 32912
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0054516228526576e-06,
+      "loss": 0.6524,
+      "step": 32913
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0049421962173453e-06,
+      "loss": 0.5808,
+      "step": 32914
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.004432805125974e-06,
+      "loss": 0.5727,
+      "step": 32915
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0039234495811277e-06,
+      "loss": 0.5815,
+      "step": 32916
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0034141295853957e-06,
+      "loss": 0.5881,
+      "step": 32917
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.002904845141363e-06,
+      "loss": 0.578,
+      "step": 32918
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0023955962516225e-06,
+      "loss": 0.5591,
+      "step": 32919
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0018863829187563e-06,
+      "loss": 0.5431,
+      "step": 32920
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0013772051453614e-06,
+      "loss": 0.6579,
+      "step": 32921
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.0008680629340113e-06,
+      "loss": 0.5901,
+      "step": 32922
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.000358956287305e-06,
+      "loss": 0.5551,
+      "step": 32923
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9998498852078205e-06,
+      "loss": 0.5893,
+      "step": 32924
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9993408496981526e-06,
+      "loss": 0.5761,
+      "step": 32925
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9988318497608836e-06,
+      "loss": 0.5747,
+      "step": 32926
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9983228853986015e-06,
+      "loss": 0.6286,
+      "step": 32927
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.99781395661389e-06,
+      "loss": 0.635,
+      "step": 32928
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9973050634093337e-06,
+      "loss": 0.5247,
+      "step": 32929
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9967962057875246e-06,
+      "loss": 0.6319,
+      "step": 32930
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.996287383751041e-06,
+      "loss": 0.5815,
+      "step": 32931
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.99577859730248e-06,
+      "loss": 0.4856,
+      "step": 32932
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.995269846444412e-06,
+      "loss": 0.6186,
+      "step": 32933
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9947611311794323e-06,
+      "loss": 0.5951,
+      "step": 32934
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.99425245151012e-06,
+      "loss": 0.6329,
+      "step": 32935
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9937438074390667e-06,
+      "loss": 0.5943,
+      "step": 32936
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9932351989688524e-06,
+      "loss": 0.5776,
+      "step": 32937
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9927266261020615e-06,
+      "loss": 0.6416,
+      "step": 32938
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.992218088841277e-06,
+      "loss": 0.5615,
+      "step": 32939
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.991709587189087e-06,
+      "loss": 0.6301,
+      "step": 32940
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9912011211480694e-06,
+      "loss": 0.5676,
+      "step": 32941
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9906926907208144e-06,
+      "loss": 0.7006,
+      "step": 32942
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9901842959099036e-06,
+      "loss": 0.5844,
+      "step": 32943
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9896759367179196e-06,
+      "loss": 0.6087,
+      "step": 32944
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.989167613147441e-06,
+      "loss": 0.6316,
+      "step": 32945
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9886593252010585e-06,
+      "loss": 0.6848,
+      "step": 32946
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.988151072881352e-06,
+      "loss": 0.6417,
+      "step": 32947
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.987642856190903e-06,
+      "loss": 0.5365,
+      "step": 32948
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9871346751322937e-06,
+      "loss": 0.6299,
+      "step": 32949
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.986626529708104e-06,
+      "loss": 0.5717,
+      "step": 32950
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9861184199209227e-06,
+      "loss": 0.6099,
+      "step": 32951
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.985610345773324e-06,
+      "loss": 0.6196,
+      "step": 32952
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9851023072678975e-06,
+      "loss": 0.6117,
+      "step": 32953
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9845943044072203e-06,
+      "loss": 0.6321,
+      "step": 32954
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9840863371938734e-06,
+      "loss": 0.54,
+      "step": 32955
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.983578405630435e-06,
+      "loss": 0.6066,
+      "step": 32956
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9830705097194934e-06,
+      "loss": 0.6593,
+      "step": 32957
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.982562649463623e-06,
+      "loss": 0.5847,
+      "step": 32958
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9820548248654125e-06,
+      "loss": 0.6396,
+      "step": 32959
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.98154703592743e-06,
+      "loss": 0.647,
+      "step": 32960
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.981039282652266e-06,
+      "loss": 0.7228,
+      "step": 32961
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.980531565042495e-06,
+      "loss": 0.6139,
+      "step": 32962
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9800238831007e-06,
+      "loss": 0.5869,
+      "step": 32963
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9795162368294604e-06,
+      "loss": 0.6433,
+      "step": 32964
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9790086262313555e-06,
+      "loss": 0.5987,
+      "step": 32965
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.978501051308961e-06,
+      "loss": 0.5769,
+      "step": 32966
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9779935120648606e-06,
+      "loss": 0.5913,
+      "step": 32967
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9774860085016323e-06,
+      "loss": 0.5302,
+      "step": 32968
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.976978540621851e-06,
+      "loss": 0.6344,
+      "step": 32969
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.976471108428105e-06,
+      "loss": 0.574,
+      "step": 32970
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9759637119229603e-06,
+      "loss": 0.6792,
+      "step": 32971
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.975456351109004e-06,
+      "loss": 0.5754,
+      "step": 32972
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.974949025988808e-06,
+      "loss": 0.5873,
+      "step": 32973
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.974441736564957e-06,
+      "loss": 0.6013,
+      "step": 32974
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.973934482840024e-06,
+      "loss": 0.638,
+      "step": 32975
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9734272648165885e-06,
+      "loss": 0.6008,
+      "step": 32976
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9729200824972237e-06,
+      "loss": 0.6138,
+      "step": 32977
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.972412935884512e-06,
+      "loss": 0.5891,
+      "step": 32978
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.971905824981026e-06,
+      "loss": 0.6058,
+      "step": 32979
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9713987497893484e-06,
+      "loss": 0.5685,
+      "step": 32980
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9708917103120516e-06,
+      "loss": 0.606,
+      "step": 32981
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.970384706551712e-06,
+      "loss": 0.5862,
+      "step": 32982
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9698777385109045e-06,
+      "loss": 0.6459,
+      "step": 32983
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.96937080619221e-06,
+      "loss": 0.5946,
+      "step": 32984
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.968863909598201e-06,
+      "loss": 0.5936,
+      "step": 32985
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.968357048731454e-06,
+      "loss": 0.608,
+      "step": 32986
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.9678502235945404e-06,
+      "loss": 0.5644,
+      "step": 32987
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.967343434190042e-06,
+      "loss": 0.5739,
+      "step": 32988
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.966836680520533e-06,
+      "loss": 0.553,
+      "step": 32989
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9663299625885824e-06,
+      "loss": 0.5687,
+      "step": 32990
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9658232803967746e-06,
+      "loss": 0.5995,
+      "step": 32991
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9653166339476735e-06,
+      "loss": 0.597,
+      "step": 32992
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9648100232438616e-06,
+      "loss": 0.5985,
+      "step": 32993
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9643034482879073e-06,
+      "loss": 0.557,
+      "step": 32994
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9637969090823913e-06,
+      "loss": 0.6874,
+      "step": 32995
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9632904056298807e-06,
+      "loss": 0.6432,
+      "step": 32996
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.962783937932958e-06,
+      "loss": 0.6374,
+      "step": 32997
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.962277505994184e-06,
+      "loss": 0.5982,
+      "step": 32998
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.961771109816144e-06,
+      "loss": 0.6167,
+      "step": 32999
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9612647494014024e-06,
+      "loss": 0.6163,
+      "step": 33000
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.960758424752539e-06,
+      "loss": 0.6452,
+      "step": 33001
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9602521358721237e-06,
+      "loss": 0.6172,
+      "step": 33002
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.959745882762729e-06,
+      "loss": 0.5926,
+      "step": 33003
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.959239665426924e-06,
+      "loss": 0.6237,
+      "step": 33004
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9587334838672865e-06,
+      "loss": 0.6112,
+      "step": 33005
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.958227338086388e-06,
+      "loss": 0.6056,
+      "step": 33006
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.957721228086795e-06,
+      "loss": 0.5904,
+      "step": 33007
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9572151538710858e-06,
+      "loss": 0.5651,
+      "step": 33008
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.956709115441828e-06,
+      "loss": 0.5817,
+      "step": 33009
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9562031128015956e-06,
+      "loss": 0.6145,
+      "step": 33010
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9556971459529536e-06,
+      "loss": 0.5925,
+      "step": 33011
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.955191214898482e-06,
+      "loss": 0.6349,
+      "step": 33012
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9546853196407453e-06,
+      "loss": 0.579,
+      "step": 33013
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.954179460182317e-06,
+      "loss": 0.5914,
+      "step": 33014
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9536736365257625e-06,
+      "loss": 0.6066,
+      "step": 33015
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.953167848673659e-06,
+      "loss": 0.5657,
+      "step": 33016
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.952662096628571e-06,
+      "loss": 0.6528,
+      "step": 33017
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.952156380393074e-06,
+      "loss": 0.6237,
+      "step": 33018
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9516506999697338e-06,
+      "loss": 0.5815,
+      "step": 33019
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.951145055361121e-06,
+      "loss": 0.6342,
+      "step": 33020
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9506394465698007e-06,
+      "loss": 0.6332,
+      "step": 33021
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.95013387359835e-06,
+      "loss": 0.6153,
+      "step": 33022
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9496283364493327e-06,
+      "loss": 0.5962,
+      "step": 33023
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.94912283512532e-06,
+      "loss": 0.5731,
+      "step": 33024
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9486173696288744e-06,
+      "loss": 0.6262,
+      "step": 33025
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9481119399625733e-06,
+      "loss": 0.6042,
+      "step": 33026
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9476065461289804e-06,
+      "loss": 0.6485,
+      "step": 33027
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.947101188130661e-06,
+      "loss": 0.6412,
+      "step": 33028
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.946595865970192e-06,
+      "loss": 0.584,
+      "step": 33029
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9460905796501294e-06,
+      "loss": 0.5371,
+      "step": 33030
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9455853291730487e-06,
+      "loss": 0.5549,
+      "step": 33031
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.945080114541512e-06,
+      "loss": 0.6477,
+      "step": 33032
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.944574935758093e-06,
+      "loss": 0.5882,
+      "step": 33033
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9440697928253515e-06,
+      "loss": 0.5333,
+      "step": 33034
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9435646857458644e-06,
+      "loss": 0.5939,
+      "step": 33035
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9430596145221845e-06,
+      "loss": 0.6276,
+      "step": 33036
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.94255457915689e-06,
+      "loss": 0.5536,
+      "step": 33037
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9420495796525384e-06,
+      "loss": 0.6119,
+      "step": 33038
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9415446160117032e-06,
+      "loss": 0.6369,
+      "step": 33039
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9410396882369463e-06,
+      "loss": 0.5399,
+      "step": 33040
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.940534796330835e-06,
+      "loss": 0.6052,
+      "step": 33041
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.940029940295931e-06,
+      "loss": 0.55,
+      "step": 33042
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9395251201348053e-06,
+      "loss": 0.595,
+      "step": 33043
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.939020335850019e-06,
+      "loss": 0.5917,
+      "step": 33044
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9385155874441373e-06,
+      "loss": 0.594,
+      "step": 33045
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9380108749197278e-06,
+      "loss": 0.5639,
+      "step": 33046
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.937506198279353e-06,
+      "loss": 0.5446,
+      "step": 33047
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9370015575255785e-06,
+      "loss": 0.6217,
+      "step": 33048
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9364969526609632e-06,
+      "loss": 0.5356,
+      "step": 33049
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.93599238368808e-06,
+      "loss": 0.5771,
+      "step": 33050
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9354878506094875e-06,
+      "loss": 0.5801,
+      "step": 33051
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9349833534277507e-06,
+      "loss": 0.6307,
+      "step": 33052
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.934478892145428e-06,
+      "loss": 0.5742,
+      "step": 33053
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9339744667650915e-06,
+      "loss": 0.5449,
+      "step": 33054
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9334700772892975e-06,
+      "loss": 0.5769,
+      "step": 33055
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.932965723720618e-06,
+      "loss": 0.6804,
+      "step": 33056
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.932461406061602e-06,
+      "loss": 0.5954,
+      "step": 33057
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9319571243148226e-06,
+      "loss": 0.6625,
+      "step": 33058
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9314528784828365e-06,
+      "loss": 0.6414,
+      "step": 33059
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.930948668568211e-06,
+      "loss": 0.5516,
+      "step": 33060
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9304444945735054e-06,
+      "loss": 0.6189,
+      "step": 33061
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.929940356501283e-06,
+      "loss": 0.5784,
+      "step": 33062
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9294362543541e-06,
+      "loss": 0.6116,
+      "step": 33063
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9289321881345257e-06,
+      "loss": 0.5724,
+      "step": 33064
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.928428157845118e-06,
+      "loss": 0.6026,
+      "step": 33065
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9279241634884335e-06,
+      "loss": 0.6378,
+      "step": 33066
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.927420205067042e-06,
+      "loss": 0.5271,
+      "step": 33067
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9269162825834997e-06,
+      "loss": 0.6599,
+      "step": 33068
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9264123960403655e-06,
+      "loss": 0.6838,
+      "step": 33069
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9259085454402003e-06,
+      "loss": 0.5627,
+      "step": 33070
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.925404730785568e-06,
+      "loss": 0.5251,
+      "step": 33071
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.924900952079023e-06,
+      "loss": 0.6556,
+      "step": 33072
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9243972093231344e-06,
+      "loss": 0.612,
+      "step": 33073
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9238935025204497e-06,
+      "loss": 0.6221,
+      "step": 33074
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9233898316735376e-06,
+      "loss": 0.6202,
+      "step": 33075
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.922886196784951e-06,
+      "loss": 0.5679,
+      "step": 33076
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.922382597857255e-06,
+      "loss": 0.6249,
+      "step": 33077
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9218790348930072e-06,
+      "loss": 0.5573,
+      "step": 33078
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9213755078947636e-06,
+      "loss": 0.5624,
+      "step": 33079
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9208720168650807e-06,
+      "loss": 0.5613,
+      "step": 33080
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9203685618065247e-06,
+      "loss": 0.6274,
+      "step": 33081
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9198651427216485e-06,
+      "loss": 0.5979,
+      "step": 33082
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9193617596130107e-06,
+      "loss": 0.643,
+      "step": 33083
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9188584124831675e-06,
+      "loss": 0.6822,
+      "step": 33084
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9183551013346813e-06,
+      "loss": 0.5813,
+      "step": 33085
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9178518261701062e-06,
+      "loss": 0.5607,
+      "step": 33086
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.917348586991997e-06,
+      "loss": 0.5636,
+      "step": 33087
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9168453838029175e-06,
+      "loss": 0.6795,
+      "step": 33088
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.916342216605421e-06,
+      "loss": 0.6341,
+      "step": 33089
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9158390854020644e-06,
+      "loss": 0.6494,
+      "step": 33090
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9153359901954005e-06,
+      "loss": 0.5738,
+      "step": 33091
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.914832930987993e-06,
+      "loss": 0.5396,
+      "step": 33092
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9143299077823907e-06,
+      "loss": 0.6167,
+      "step": 33093
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9138269205811597e-06,
+      "loss": 0.6272,
+      "step": 33094
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.913323969386843e-06,
+      "loss": 0.5587,
+      "step": 33095
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9128210542020053e-06,
+      "loss": 0.6907,
+      "step": 33096
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9123181750291972e-06,
+      "loss": 0.5488,
+      "step": 33097
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9118153318709786e-06,
+      "loss": 0.6614,
+      "step": 33098
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.911312524729899e-06,
+      "loss": 0.5953,
+      "step": 33099
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9108097536085236e-06,
+      "loss": 0.6606,
+      "step": 33100
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.910307018509393e-06,
+      "loss": 0.6249,
+      "step": 33101
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9098043194350723e-06,
+      "loss": 0.6328,
+      "step": 33102
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9093016563881125e-06,
+      "loss": 0.557,
+      "step": 33103
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9087990293710633e-06,
+      "loss": 0.5644,
+      "step": 33104
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9082964383864878e-06,
+      "loss": 0.5755,
+      "step": 33105
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9077938834369333e-06,
+      "loss": 0.6107,
+      "step": 33106
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.907291364524957e-06,
+      "loss": 0.5935,
+      "step": 33107
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9067888816531052e-06,
+      "loss": 0.552,
+      "step": 33108
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9062864348239404e-06,
+      "loss": 0.6033,
+      "step": 33109
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9057840240400124e-06,
+      "loss": 0.5898,
+      "step": 33110
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.905281649303873e-06,
+      "loss": 0.6015,
+      "step": 33111
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.904779310618072e-06,
+      "loss": 0.553,
+      "step": 33112
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9042770079851688e-06,
+      "loss": 0.6126,
+      "step": 33113
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.903774741407709e-06,
+      "loss": 0.5587,
+      "step": 33114
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.903272510888251e-06,
+      "loss": 0.6727,
+      "step": 33115
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.902770316429343e-06,
+      "loss": 0.5184,
+      "step": 33116
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9022681580335387e-06,
+      "loss": 0.6261,
+      "step": 33117
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.901766035703385e-06,
+      "loss": 0.5825,
+      "step": 33118
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9012639494414397e-06,
+      "loss": 0.6081,
+      "step": 33119
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.900761899250252e-06,
+      "loss": 0.5425,
+      "step": 33120
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.9002598851323714e-06,
+      "loss": 0.5969,
+      "step": 33121
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.899757907090347e-06,
+      "loss": 0.5923,
+      "step": 33122
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.899255965126735e-06,
+      "loss": 0.542,
+      "step": 33123
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8987540592440823e-06,
+      "loss": 0.5501,
+      "step": 33124
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.898252189444938e-06,
+      "loss": 0.6661,
+      "step": 33125
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8977503557318564e-06,
+      "loss": 0.6482,
+      "step": 33126
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.897248558107385e-06,
+      "loss": 0.6321,
+      "step": 33127
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.896746796574075e-06,
+      "loss": 0.6449,
+      "step": 33128
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8962450711344716e-06,
+      "loss": 0.5881,
+      "step": 33129
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8957433817911294e-06,
+      "loss": 0.6021,
+      "step": 33130
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.895241728546593e-06,
+      "loss": 0.6285,
+      "step": 33131
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.89474011140342e-06,
+      "loss": 0.5898,
+      "step": 33132
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8942385303641483e-06,
+      "loss": 0.6065,
+      "step": 33133
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8937369854313335e-06,
+      "loss": 0.5993,
+      "step": 33134
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.89323547660752e-06,
+      "loss": 0.5577,
+      "step": 33135
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8927340038952614e-06,
+      "loss": 0.5915,
+      "step": 33136
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8922325672971e-06,
+      "loss": 0.5587,
+      "step": 33137
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.891731166815591e-06,
+      "loss": 0.5692,
+      "step": 33138
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8912298024532726e-06,
+      "loss": 0.6398,
+      "step": 33139
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.890728474212701e-06,
+      "loss": 0.5399,
+      "step": 33140
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.89022718209642e-06,
+      "loss": 0.5979,
+      "step": 33141
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8897259261069733e-06,
+      "loss": 0.6123,
+      "step": 33142
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8892247062469146e-06,
+      "loss": 0.617,
+      "step": 33143
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.888723522518788e-06,
+      "loss": 0.6375,
+      "step": 33144
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8882223749251393e-06,
+      "loss": 0.6404,
+      "step": 33145
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8877212634685135e-06,
+      "loss": 0.6313,
+      "step": 33146
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8872201881514605e-06,
+      "loss": 0.6339,
+      "step": 33147
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8867191489765255e-06,
+      "loss": 0.5819,
+      "step": 33148
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8862181459462545e-06,
+      "loss": 0.6504,
+      "step": 33149
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.885717179063189e-06,
+      "loss": 0.6313,
+      "step": 33150
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8852162483298808e-06,
+      "loss": 0.5766,
+      "step": 33151
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8847153537488692e-06,
+      "loss": 0.643,
+      "step": 33152
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.884214495322706e-06,
+      "loss": 0.5941,
+      "step": 33153
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8837136730539327e-06,
+      "loss": 0.6427,
+      "step": 33154
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8832128869450937e-06,
+      "loss": 0.6649,
+      "step": 33155
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8827121369987322e-06,
+      "loss": 0.6263,
+      "step": 33156
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.882211423217398e-06,
+      "loss": 0.5697,
+      "step": 33157
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8817107456036285e-06,
+      "loss": 0.6196,
+      "step": 33158
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8812101041599773e-06,
+      "loss": 0.6378,
+      "step": 33159
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8807094988889773e-06,
+      "loss": 0.6075,
+      "step": 33160
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.880208929793179e-06,
+      "loss": 0.6058,
+      "step": 33161
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.879708396875125e-06,
+      "loss": 0.6269,
+      "step": 33162
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.879207900137355e-06,
+      "loss": 0.6136,
+      "step": 33163
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.878707439582418e-06,
+      "loss": 0.5772,
+      "step": 33164
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.878207015212854e-06,
+      "loss": 0.5954,
+      "step": 33165
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.877706627031206e-06,
+      "loss": 0.629,
+      "step": 33166
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.877206275040013e-06,
+      "loss": 0.5638,
+      "step": 33167
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8767059592418246e-06,
+      "loss": 0.5349,
+      "step": 33168
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8762056796391758e-06,
+      "loss": 0.6385,
+      "step": 33169
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.875705436234618e-06,
+      "loss": 0.6031,
+      "step": 33170
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.875205229030682e-06,
+      "loss": 0.561,
+      "step": 33171
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8747050580299186e-06,
+      "loss": 0.6118,
+      "step": 33172
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.874204923234861e-06,
+      "loss": 0.6204,
+      "step": 33173
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.873704824648058e-06,
+      "loss": 0.5619,
+      "step": 33174
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.873204762272049e-06,
+      "loss": 0.6147,
+      "step": 33175
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.872704736109373e-06,
+      "loss": 0.611,
+      "step": 33176
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.872204746162569e-06,
+      "loss": 0.5776,
+      "step": 33177
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8717047924341836e-06,
+      "loss": 0.5885,
+      "step": 33178
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8712048749267496e-06,
+      "loss": 0.5926,
+      "step": 33179
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8707049936428143e-06,
+      "loss": 0.6474,
+      "step": 33180
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8702051485849157e-06,
+      "loss": 0.6251,
+      "step": 33181
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.869705339755593e-06,
+      "loss": 0.6276,
+      "step": 33182
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.869205567157385e-06,
+      "loss": 0.6213,
+      "step": 33183
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8687058307928295e-06,
+      "loss": 0.588,
+      "step": 33184
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8682061306644706e-06,
+      "loss": 0.6105,
+      "step": 33185
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8677064667748455e-06,
+      "loss": 0.6536,
+      "step": 33186
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.867206839126492e-06,
+      "loss": 0.6269,
+      "step": 33187
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.866707247721947e-06,
+      "loss": 0.5827,
+      "step": 33188
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.866207692563754e-06,
+      "loss": 0.6279,
+      "step": 33189
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8657081736544454e-06,
+      "loss": 0.5642,
+      "step": 33190
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.865208690996567e-06,
+      "loss": 0.5991,
+      "step": 33191
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.864709244592653e-06,
+      "loss": 0.6484,
+      "step": 33192
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8642098344452407e-06,
+      "loss": 0.616,
+      "step": 33193
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8637104605568645e-06,
+      "loss": 0.5789,
+      "step": 33194
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8632111229300696e-06,
+      "loss": 0.5727,
+      "step": 33195
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8627118215673856e-06,
+      "loss": 0.64,
+      "step": 33196
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8622125564713587e-06,
+      "loss": 0.6552,
+      "step": 33197
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.861713327644514e-06,
+      "loss": 0.5747,
+      "step": 33198
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8612141350893975e-06,
+      "loss": 0.5831,
+      "step": 33199
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.86071497880854e-06,
+      "loss": 0.6023,
+      "step": 33200
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.860215858804484e-06,
+      "loss": 0.6456,
+      "step": 33201
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8597167750797607e-06,
+      "loss": 0.61,
+      "step": 33202
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.859217727636909e-06,
+      "loss": 0.6187,
+      "step": 33203
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8587187164784626e-06,
+      "loss": 0.6177,
+      "step": 33204
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8582197416069545e-06,
+      "loss": 0.6324,
+      "step": 33205
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.857720803024928e-06,
+      "loss": 0.5675,
+      "step": 33206
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.857221900734909e-06,
+      "loss": 0.595,
+      "step": 33207
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.856723034739445e-06,
+      "loss": 0.6487,
+      "step": 33208
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.856224205041056e-06,
+      "loss": 0.6562,
+      "step": 33209
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8557254116422873e-06,
+      "loss": 0.5945,
+      "step": 33210
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8552266545456676e-06,
+      "loss": 0.5958,
+      "step": 33211
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.854727933753736e-06,
+      "loss": 0.5566,
+      "step": 33212
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8542292492690247e-06,
+      "loss": 0.6424,
+      "step": 33213
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.853730601094068e-06,
+      "loss": 0.586,
+      "step": 33214
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.853231989231394e-06,
+      "loss": 0.5877,
+      "step": 33215
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.852733413683546e-06,
+      "loss": 0.5672,
+      "step": 33216
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.852234874453049e-06,
+      "loss": 0.6905,
+      "step": 33217
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8517363715424438e-06,
+      "loss": 0.6075,
+      "step": 33218
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8512379049542595e-06,
+      "loss": 0.6016,
+      "step": 33219
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8507394746910276e-06,
+      "loss": 0.6326,
+      "step": 33220
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.850241080755284e-06,
+      "loss": 0.6107,
+      "step": 33221
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8497427231495543e-06,
+      "loss": 0.6515,
+      "step": 33222
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.849244401876381e-06,
+      "loss": 0.602,
+      "step": 33223
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.84874611693829e-06,
+      "loss": 0.5812,
+      "step": 33224
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8482478683378146e-06,
+      "loss": 0.6018,
+      "step": 33225
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.847749656077483e-06,
+      "loss": 0.5797,
+      "step": 33226
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8472514801598326e-06,
+      "loss": 0.5878,
+      "step": 33227
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8467533405873904e-06,
+      "loss": 0.5604,
+      "step": 33228
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8462552373626918e-06,
+      "loss": 0.5363,
+      "step": 33229
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8457571704882647e-06,
+      "loss": 0.6246,
+      "step": 33230
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.845259139966642e-06,
+      "loss": 0.6004,
+      "step": 33231
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.844761145800349e-06,
+      "loss": 0.5789,
+      "step": 33232
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8442631879919234e-06,
+      "loss": 0.5928,
+      "step": 33233
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8437652665438885e-06,
+      "loss": 0.5585,
+      "step": 33234
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8432673814587853e-06,
+      "loss": 0.6041,
+      "step": 33235
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8427695327391304e-06,
+      "loss": 0.5793,
+      "step": 33236
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.842271720387462e-06,
+      "loss": 0.5789,
+      "step": 33237
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8417739444063043e-06,
+      "loss": 0.5973,
+      "step": 33238
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8412762047981925e-06,
+      "loss": 0.6116,
+      "step": 33239
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.840778501565653e-06,
+      "loss": 0.6742,
+      "step": 33240
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8402808347112145e-06,
+      "loss": 0.5859,
+      "step": 33241
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8397832042374064e-06,
+      "loss": 0.5716,
+      "step": 33242
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8392856101467537e-06,
+      "loss": 0.5762,
+      "step": 33243
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.83878805244179e-06,
+      "loss": 0.6156,
+      "step": 33244
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8382905311250386e-06,
+      "loss": 0.5936,
+      "step": 33245
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8377930461990376e-06,
+      "loss": 0.6279,
+      "step": 33246
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8372955976663006e-06,
+      "loss": 0.582,
+      "step": 33247
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8367981855293657e-06,
+      "loss": 0.6032,
+      "step": 33248
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.836300809790754e-06,
+      "loss": 0.6204,
+      "step": 33249
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8358034704529993e-06,
+      "loss": 0.6108,
+      "step": 33250
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8353061675186254e-06,
+      "loss": 0.5884,
+      "step": 33251
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.834808900990158e-06,
+      "loss": 0.64,
+      "step": 33252
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8343116708701234e-06,
+      "loss": 0.6109,
+      "step": 33253
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.833814477161052e-06,
+      "loss": 0.5489,
+      "step": 33254
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.833317319865465e-06,
+      "loss": 0.5709,
+      "step": 33255
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.832820198985894e-06,
+      "loss": 0.6021,
+      "step": 33256
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8323231145248633e-06,
+      "loss": 0.6195,
+      "step": 33257
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8318260664848985e-06,
+      "loss": 0.6298,
+      "step": 33258
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8313290548685214e-06,
+      "loss": 0.5959,
+      "step": 33259
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.830832079678263e-06,
+      "loss": 0.6081,
+      "step": 33260
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.830335140916647e-06,
+      "loss": 0.6224,
+      "step": 33261
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8298382385861976e-06,
+      "loss": 0.6266,
+      "step": 33262
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8293413726894402e-06,
+      "loss": 0.6288,
+      "step": 33263
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.828844543228897e-06,
+      "loss": 0.6148,
+      "step": 33264
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8283477502070965e-06,
+      "loss": 0.5594,
+      "step": 33265
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8278509936265597e-06,
+      "loss": 0.6277,
+      "step": 33266
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8273542734898174e-06,
+      "loss": 0.5964,
+      "step": 33267
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8268575897993833e-06,
+      "loss": 0.6478,
+      "step": 33268
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8263609425577887e-06,
+      "loss": 0.6353,
+      "step": 33269
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8258643317675527e-06,
+      "loss": 0.6159,
+      "step": 33270
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8253677574312045e-06,
+      "loss": 0.6327,
+      "step": 33271
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8248712195512606e-06,
+      "loss": 0.6056,
+      "step": 33272
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.824374718130253e-06,
+      "loss": 0.6323,
+      "step": 33273
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.823878253170693e-06,
+      "loss": 0.6091,
+      "step": 33274
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.823381824675112e-06,
+      "loss": 0.628,
+      "step": 33275
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8228854326460276e-06,
+      "loss": 0.6564,
+      "step": 33276
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.822389077085966e-06,
+      "loss": 0.6232,
+      "step": 33277
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8218927579974485e-06,
+      "loss": 0.5556,
+      "step": 33278
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.821396475382996e-06,
+      "loss": 0.6712,
+      "step": 33279
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.820900229245127e-06,
+      "loss": 0.5725,
+      "step": 33280
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8204040195863704e-06,
+      "loss": 0.6422,
+      "step": 33281
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8199078464092433e-06,
+      "loss": 0.5385,
+      "step": 33282
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8194117097162634e-06,
+      "loss": 0.559,
+      "step": 33283
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8189156095099625e-06,
+      "loss": 0.6253,
+      "step": 33284
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8184195457928475e-06,
+      "loss": 0.5627,
+      "step": 33285
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.817923518567449e-06,
+      "loss": 0.5426,
+      "step": 33286
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.817427527836282e-06,
+      "loss": 0.6926,
+      "step": 33287
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.816931573601872e-06,
+      "loss": 0.5654,
+      "step": 33288
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.816435655866737e-06,
+      "loss": 0.6022,
+      "step": 33289
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.815939774633396e-06,
+      "loss": 0.6212,
+      "step": 33290
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.815443929904366e-06,
+      "loss": 0.6369,
+      "step": 33291
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.814948121682172e-06,
+      "loss": 0.6453,
+      "step": 33292
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8144523499693277e-06,
+      "loss": 0.5618,
+      "step": 33293
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.813956614768362e-06,
+      "loss": 0.6281,
+      "step": 33294
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8134609160817805e-06,
+      "loss": 0.5887,
+      "step": 33295
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8129652539121122e-06,
+      "loss": 0.5948,
+      "step": 33296
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8124696282618692e-06,
+      "loss": 0.6017,
+      "step": 33297
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8119740391335757e-06,
+      "loss": 0.5701,
+      "step": 33298
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8114784865297486e-06,
+      "loss": 0.6171,
+      "step": 33299
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8109829704529023e-06,
+      "loss": 0.6423,
+      "step": 33300
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.810487490905559e-06,
+      "loss": 0.5935,
+      "step": 33301
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8099920478902298e-06,
+      "loss": 0.6424,
+      "step": 33302
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8094966414094404e-06,
+      "loss": 0.6285,
+      "step": 33303
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8090012714657023e-06,
+      "loss": 0.5615,
+      "step": 33304
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.80850593806154e-06,
+      "loss": 0.6086,
+      "step": 33305
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.808010641199458e-06,
+      "loss": 0.5321,
+      "step": 33306
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.807515380881984e-06,
+      "loss": 0.6093,
+      "step": 33307
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8070201571116285e-06,
+      "loss": 0.601,
+      "step": 33308
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8065249698909126e-06,
+      "loss": 0.6258,
+      "step": 33309
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8060298192223468e-06,
+      "loss": 0.604,
+      "step": 33310
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.805534705108457e-06,
+      "loss": 0.5648,
+      "step": 33311
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8050396275517455e-06,
+      "loss": 0.6004,
+      "step": 33312
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.804544586554738e-06,
+      "loss": 0.5178,
+      "step": 33313
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8040495821199443e-06,
+      "loss": 0.5915,
+      "step": 33314
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.803554614249885e-06,
+      "loss": 0.6177,
+      "step": 33315
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8030596829470713e-06,
+      "loss": 0.5747,
+      "step": 33316
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8025647882140205e-06,
+      "loss": 0.6053,
+      "step": 33317
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8020699300532417e-06,
+      "loss": 0.6358,
+      "step": 33318
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.801575108467257e-06,
+      "loss": 0.6078,
+      "step": 33319
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.801080323458578e-06,
+      "loss": 0.5579,
+      "step": 33320
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.800585575029714e-06,
+      "loss": 0.6119,
+      "step": 33321
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8000908631831904e-06,
+      "loss": 0.5554,
+      "step": 33322
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.799596187921507e-06,
+      "loss": 0.5178,
+      "step": 33323
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7991015492471875e-06,
+      "loss": 0.6354,
+      "step": 33324
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7986069471627376e-06,
+      "loss": 0.6138,
+      "step": 33325
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.798112381670679e-06,
+      "loss": 0.6335,
+      "step": 33326
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7976178527735207e-06,
+      "loss": 0.6205,
+      "step": 33327
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7971233604737747e-06,
+      "loss": 0.675,
+      "step": 33328
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7966289047739512e-06,
+      "loss": 0.6432,
+      "step": 33329
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7961344856765704e-06,
+      "loss": 0.5569,
+      "step": 33330
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.795640103184135e-06,
+      "loss": 0.5823,
+      "step": 33331
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7951457572991703e-06,
+      "loss": 0.6115,
+      "step": 33332
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.794651448024173e-06,
+      "loss": 0.6183,
+      "step": 33333
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7941571753616636e-06,
+      "loss": 0.6716,
+      "step": 33334
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.79366293931415e-06,
+      "loss": 0.6044,
+      "step": 33335
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.793168739884149e-06,
+      "loss": 0.6245,
+      "step": 33336
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7926745770741683e-06,
+      "loss": 0.648,
+      "step": 33337
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.792180450886718e-06,
+      "loss": 0.5802,
+      "step": 33338
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.791686361324307e-06,
+      "loss": 0.5833,
+      "step": 33339
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7911923083894523e-06,
+      "loss": 0.5721,
+      "step": 33340
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7906982920846592e-06,
+      "loss": 0.6187,
+      "step": 33341
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7902043124124377e-06,
+      "loss": 0.5466,
+      "step": 33342
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7897103693753046e-06,
+      "loss": 0.6128,
+      "step": 33343
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.78921646297576e-06,
+      "loss": 0.6511,
+      "step": 33344
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.788722593216321e-06,
+      "loss": 0.5533,
+      "step": 33345
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7882287600994908e-06,
+      "loss": 0.576,
+      "step": 33346
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.787734963627785e-06,
+      "loss": 0.6216,
+      "step": 33347
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.787241203803708e-06,
+      "loss": 0.562,
+      "step": 33348
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.786747480629777e-06,
+      "loss": 0.5883,
+      "step": 33349
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7862537941084878e-06,
+      "loss": 0.6262,
+      "step": 33350
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.785760144242359e-06,
+      "loss": 0.628,
+      "step": 33351
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7852665310338924e-06,
+      "loss": 0.6348,
+      "step": 33352
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7847729544856027e-06,
+      "loss": 0.5783,
+      "step": 33353
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.784279414599995e-06,
+      "loss": 0.5798,
+      "step": 33354
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7837859113795764e-06,
+      "loss": 0.5652,
+      "step": 33355
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7832924448268518e-06,
+      "loss": 0.61,
+      "step": 33356
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7827990149443353e-06,
+      "loss": 0.5225,
+      "step": 33357
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.78230562173453e-06,
+      "loss": 0.5971,
+      "step": 33358
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7818122651999448e-06,
+      "loss": 0.5628,
+      "step": 33359
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.781318945343081e-06,
+      "loss": 0.6307,
+      "step": 33360
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.780825662166454e-06,
+      "loss": 0.6362,
+      "step": 33361
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.780332415672565e-06,
+      "loss": 0.5922,
+      "step": 33362
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.779839205863919e-06,
+      "loss": 0.5746,
+      "step": 33363
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.779346032743028e-06,
+      "loss": 0.5908,
+      "step": 33364
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.778852896312394e-06,
+      "loss": 0.6397,
+      "step": 33365
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7783597965745223e-06,
+      "loss": 0.6096,
+      "step": 33366
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7778667335319177e-06,
+      "loss": 0.5774,
+      "step": 33367
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.77737370718709e-06,
+      "loss": 0.5782,
+      "step": 33368
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7768807175425384e-06,
+      "loss": 0.5572,
+      "step": 33369
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7763877646007775e-06,
+      "loss": 0.6222,
+      "step": 33370
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7758948483643e-06,
+      "loss": 0.6291,
+      "step": 33371
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7754019688356193e-06,
+      "loss": 0.6048,
+      "step": 33372
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7749091260172335e-06,
+      "loss": 0.6219,
+      "step": 33373
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7744163199116536e-06,
+      "loss": 0.5754,
+      "step": 33374
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7739235505213813e-06,
+      "loss": 0.5783,
+      "step": 33375
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7734308178489187e-06,
+      "loss": 0.6348,
+      "step": 33376
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.772938121896769e-06,
+      "loss": 0.6281,
+      "step": 33377
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.772445462667439e-06,
+      "loss": 0.613,
+      "step": 33378
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.77195284016343e-06,
+      "loss": 0.6638,
+      "step": 33379
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7714602543872436e-06,
+      "loss": 0.6027,
+      "step": 33380
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7709677053413863e-06,
+      "loss": 0.5468,
+      "step": 33381
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.77047519302836e-06,
+      "loss": 0.5536,
+      "step": 33382
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7699827174506665e-06,
+      "loss": 0.5693,
+      "step": 33383
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7694902786108056e-06,
+      "loss": 0.6427,
+      "step": 33384
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7689978765112848e-06,
+      "loss": 0.6178,
+      "step": 33385
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7685055111546034e-06,
+      "loss": 0.5632,
+      "step": 33386
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7680131825432634e-06,
+      "loss": 0.6116,
+      "step": 33387
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7675208906797634e-06,
+      "loss": 0.6155,
+      "step": 33388
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.767028635566611e-06,
+      "loss": 0.6266,
+      "step": 33389
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7665364172063003e-06,
+      "loss": 0.6327,
+      "step": 33390
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.766044235601341e-06,
+      "loss": 0.6259,
+      "step": 33391
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.765552090754229e-06,
+      "loss": 0.6066,
+      "step": 33392
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7650599826674652e-06,
+      "loss": 0.5841,
+      "step": 33393
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7645679113435476e-06,
+      "loss": 0.5578,
+      "step": 33394
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.764075876784983e-06,
+      "loss": 0.6248,
+      "step": 33395
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.763583878994267e-06,
+      "loss": 0.6149,
+      "step": 33396
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.763091917973901e-06,
+      "loss": 0.6026,
+      "step": 33397
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7625999937263815e-06,
+      "loss": 0.5711,
+      "step": 33398
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7621081062542142e-06,
+      "loss": 0.5609,
+      "step": 33399
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7616162555598957e-06,
+      "loss": 0.602,
+      "step": 33400
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7611244416459217e-06,
+      "loss": 0.6013,
+      "step": 33401
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7606326645147963e-06,
+      "loss": 0.6364,
+      "step": 33402
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.760140924169017e-06,
+      "loss": 0.588,
+      "step": 33403
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7596492206110826e-06,
+      "loss": 0.6284,
+      "step": 33404
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.759157553843488e-06,
+      "loss": 0.6182,
+      "step": 33405
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.758665923868736e-06,
+      "loss": 0.57,
+      "step": 33406
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7581743306893217e-06,
+      "loss": 0.5879,
+      "step": 33407
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7576827743077494e-06,
+      "loss": 0.6288,
+      "step": 33408
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.757191254726507e-06,
+      "loss": 0.6378,
+      "step": 33409
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.756699771948099e-06,
+      "loss": 0.5741,
+      "step": 33410
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7562083259750183e-06,
+      "loss": 0.6004,
+      "step": 33411
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7557169168097674e-06,
+      "loss": 0.5506,
+      "step": 33412
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7552255444548404e-06,
+      "loss": 0.5959,
+      "step": 33413
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.754734208912735e-06,
+      "loss": 0.5943,
+      "step": 33414
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7542429101859426e-06,
+      "loss": 0.5469,
+      "step": 33415
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7537516482769676e-06,
+      "loss": 0.5908,
+      "step": 33416
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.753260423188303e-06,
+      "loss": 0.5938,
+      "step": 33417
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.752769234922441e-06,
+      "loss": 0.57,
+      "step": 33418
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.752278083481884e-06,
+      "loss": 0.5328,
+      "step": 33419
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7517869688691246e-06,
+      "loss": 0.5737,
+      "step": 33420
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.751295891086658e-06,
+      "loss": 0.6424,
+      "step": 33421
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.750804850136978e-06,
+      "loss": 0.6168,
+      "step": 33422
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7503138460225842e-06,
+      "loss": 0.5828,
+      "step": 33423
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.749822878745968e-06,
+      "loss": 0.6421,
+      "step": 33424
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.749331948309626e-06,
+      "loss": 0.6649,
+      "step": 33425
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.748841054716048e-06,
+      "loss": 0.5907,
+      "step": 33426
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7483501979677363e-06,
+      "loss": 0.5967,
+      "step": 33427
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.747859378067177e-06,
+      "loss": 0.5694,
+      "step": 33428
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7473685950168715e-06,
+      "loss": 0.6157,
+      "step": 33429
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7468778488193104e-06,
+      "loss": 0.6527,
+      "step": 33430
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.746387139476987e-06,
+      "loss": 0.6267,
+      "step": 33431
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7458964669923914e-06,
+      "loss": 0.6065,
+      "step": 33432
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7454058313680245e-06,
+      "loss": 0.5894,
+      "step": 33433
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7449152326063743e-06,
+      "loss": 0.6453,
+      "step": 33434
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.744424670709934e-06,
+      "loss": 0.5687,
+      "step": 33435
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.743934145681195e-06,
+      "loss": 0.5787,
+      "step": 33436
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.743443657522654e-06,
+      "loss": 0.615,
+      "step": 33437
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7429532062368013e-06,
+      "loss": 0.6711,
+      "step": 33438
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.742462791826125e-06,
+      "loss": 0.5966,
+      "step": 33439
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.741972414293125e-06,
+      "loss": 0.6043,
+      "step": 33440
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.741482073640287e-06,
+      "loss": 0.6037,
+      "step": 33441
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.740991769870106e-06,
+      "loss": 0.5944,
+      "step": 33442
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.740501502985068e-06,
+      "loss": 0.6093,
+      "step": 33443
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7400112729876703e-06,
+      "loss": 0.6296,
+      "step": 33444
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7395210798803994e-06,
+      "loss": 0.6026,
+      "step": 33445
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7390309236657543e-06,
+      "loss": 0.6256,
+      "step": 33446
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7385408043462123e-06,
+      "loss": 0.6401,
+      "step": 33447
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.738050721924276e-06,
+      "loss": 0.5298,
+      "step": 33448
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7375606764024265e-06,
+      "loss": 0.5754,
+      "step": 33449
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7370706677831616e-06,
+      "loss": 0.6331,
+      "step": 33450
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7365806960689667e-06,
+      "loss": 0.6327,
+      "step": 33451
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7360907612623335e-06,
+      "loss": 0.6305,
+      "step": 33452
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.735600863365747e-06,
+      "loss": 0.5627,
+      "step": 33453
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.735111002381703e-06,
+      "loss": 0.6461,
+      "step": 33454
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7346211783126873e-06,
+      "loss": 0.5798,
+      "step": 33455
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7341313911611877e-06,
+      "loss": 0.5292,
+      "step": 33456
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.733641640929696e-06,
+      "loss": 0.5637,
+      "step": 33457
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7331519276206995e-06,
+      "loss": 0.6175,
+      "step": 33458
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.732662251236686e-06,
+      "loss": 0.5579,
+      "step": 33459
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7321726117801416e-06,
+      "loss": 0.5715,
+      "step": 33460
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7316830092535594e-06,
+      "loss": 0.6013,
+      "step": 33461
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7311934436594247e-06,
+      "loss": 0.6015,
+      "step": 33462
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7307039150002235e-06,
+      "loss": 0.6377,
+      "step": 33463
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7302144232784435e-06,
+      "loss": 0.6355,
+      "step": 33464
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.729724968496574e-06,
+      "loss": 0.644,
+      "step": 33465
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7292355506571e-06,
+      "loss": 0.632,
+      "step": 33466
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7287461697625117e-06,
+      "loss": 0.67,
+      "step": 33467
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.728256825815294e-06,
+      "loss": 0.588,
+      "step": 33468
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.727767518817933e-06,
+      "loss": 0.5668,
+      "step": 33469
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.727278248772911e-06,
+      "loss": 0.5994,
+      "step": 33470
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7267890156827225e-06,
+      "loss": 0.6373,
+      "step": 33471
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7262998195498446e-06,
+      "loss": 0.5842,
+      "step": 33472
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7258106603767754e-06,
+      "loss": 0.5371,
+      "step": 33473
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.725321538165985e-06,
+      "loss": 0.5948,
+      "step": 33474
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.72483245291997e-06,
+      "loss": 0.5739,
+      "step": 33475
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7243434046412122e-06,
+      "loss": 0.6172,
+      "step": 33476
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7238543933321927e-06,
+      "loss": 0.5591,
+      "step": 33477
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.723365418995403e-06,
+      "loss": 0.5844,
+      "step": 33478
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.722876481633325e-06,
+      "loss": 0.5575,
+      "step": 33479
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7223875812484425e-06,
+      "loss": 0.6273,
+      "step": 33480
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.721898717843238e-06,
+      "loss": 0.5832,
+      "step": 33481
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7214098914201994e-06,
+      "loss": 0.5583,
+      "step": 33482
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.720921101981806e-06,
+      "loss": 0.5328,
+      "step": 33483
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7204323495305507e-06,
+      "loss": 0.5799,
+      "step": 33484
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7199436340689045e-06,
+      "loss": 0.6094,
+      "step": 33485
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.719454955599359e-06,
+      "loss": 0.5984,
+      "step": 33486
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.718966314124393e-06,
+      "loss": 0.5855,
+      "step": 33487
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7184777096464943e-06,
+      "loss": 0.6199,
+      "step": 33488
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7179891421681427e-06,
+      "loss": 0.6048,
+      "step": 33489
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7175006116918214e-06,
+      "loss": 0.7184,
+      "step": 33490
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.717012118220008e-06,
+      "loss": 0.5774,
+      "step": 33491
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7165236617551925e-06,
+      "loss": 0.5562,
+      "step": 33492
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7160352422998506e-06,
+      "loss": 0.639,
+      "step": 33493
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7155468598564705e-06,
+      "loss": 0.5953,
+      "step": 33494
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7150585144275286e-06,
+      "loss": 0.5986,
+      "step": 33495
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7145702060155087e-06,
+      "loss": 0.5986,
+      "step": 33496
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.714081934622892e-06,
+      "loss": 0.5858,
+      "step": 33497
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.713593700252155e-06,
+      "loss": 0.5682,
+      "step": 33498
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.713105502905785e-06,
+      "loss": 0.571,
+      "step": 33499
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.71261734258626e-06,
+      "loss": 0.6641,
+      "step": 33500
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7121292192960614e-06,
+      "loss": 0.6134,
+      "step": 33501
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7116411330376646e-06,
+      "loss": 0.5888,
+      "step": 33502
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.711153083813558e-06,
+      "loss": 0.5904,
+      "step": 33503
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.710665071626213e-06,
+      "loss": 0.6256,
+      "step": 33504
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7101770964781173e-06,
+      "loss": 0.6101,
+      "step": 33505
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7096891583717453e-06,
+      "loss": 0.634,
+      "step": 33506
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7092012573095796e-06,
+      "loss": 0.623,
+      "step": 33507
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7087133932940934e-06,
+      "loss": 0.6086,
+      "step": 33508
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7082255663277733e-06,
+      "loss": 0.6597,
+      "step": 33509
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7077377764130918e-06,
+      "loss": 0.6174,
+      "step": 33510
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7072500235525357e-06,
+      "loss": 0.623,
+      "step": 33511
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7067623077485715e-06,
+      "loss": 0.5511,
+      "step": 33512
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.706274629003688e-06,
+      "loss": 0.5879,
+      "step": 33513
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.705786987320359e-06,
+      "loss": 0.5832,
+      "step": 33514
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7052993827010598e-06,
+      "loss": 0.5993,
+      "step": 33515
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.704811815148273e-06,
+      "loss": 0.6534,
+      "step": 33516
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7043242846644746e-06,
+      "loss": 0.5758,
+      "step": 33517
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.70383679125214e-06,
+      "loss": 0.5628,
+      "step": 33518
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7033493349137453e-06,
+      "loss": 0.6192,
+      "step": 33519
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.702861915651772e-06,
+      "loss": 0.5345,
+      "step": 33520
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7023745334686926e-06,
+      "loss": 0.5968,
+      "step": 33521
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7018871883669906e-06,
+      "loss": 0.6135,
+      "step": 33522
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7013998803491314e-06,
+      "loss": 0.5452,
+      "step": 33523
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7009126094175997e-06,
+      "loss": 0.6233,
+      "step": 33524
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.7004253755748654e-06,
+      "loss": 0.5655,
+      "step": 33525
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6999381788234104e-06,
+      "loss": 0.5876,
+      "step": 33526
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6994510191657076e-06,
+      "loss": 0.6101,
+      "step": 33527
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.698963896604233e-06,
+      "loss": 0.6171,
+      "step": 33528
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6984768111414574e-06,
+      "loss": 0.6529,
+      "step": 33529
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6979897627798626e-06,
+      "loss": 0.6193,
+      "step": 33530
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6975027515219175e-06,
+      "loss": 0.6128,
+      "step": 33531
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6970157773701023e-06,
+      "loss": 0.5851,
+      "step": 33532
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6965288403268886e-06,
+      "loss": 0.5637,
+      "step": 33533
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6960419403947514e-06,
+      "loss": 0.6318,
+      "step": 33534
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.695555077576163e-06,
+      "loss": 0.6006,
+      "step": 33535
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.695068251873596e-06,
+      "loss": 0.6511,
+      "step": 33536
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6945814632895296e-06,
+      "loss": 0.6415,
+      "step": 33537
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.694094711826435e-06,
+      "loss": 0.6034,
+      "step": 33538
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6936079974867848e-06,
+      "loss": 0.6248,
+      "step": 33539
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6931213202730478e-06,
+      "loss": 0.5874,
+      "step": 33540
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6926346801877045e-06,
+      "loss": 0.5632,
+      "step": 33541
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.692148077233222e-06,
+      "loss": 0.5961,
+      "step": 33542
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6916615114120813e-06,
+      "loss": 0.6412,
+      "step": 33543
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.691174982726743e-06,
+      "loss": 0.6525,
+      "step": 33544
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.690688491179688e-06,
+      "loss": 0.5486,
+      "step": 33545
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.690202036773383e-06,
+      "loss": 0.6362,
+      "step": 33546
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6897156195103038e-06,
+      "loss": 0.595,
+      "step": 33547
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6892292393929186e-06,
+      "loss": 0.5827,
+      "step": 33548
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.688742896423706e-06,
+      "loss": 0.6672,
+      "step": 33549
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6882565906051263e-06,
+      "loss": 0.7035,
+      "step": 33550
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.687770321939659e-06,
+      "loss": 0.6043,
+      "step": 33551
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.687284090429769e-06,
+      "loss": 0.5835,
+      "step": 33552
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.686797896077934e-06,
+      "loss": 0.6321,
+      "step": 33553
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.68631173888662e-06,
+      "loss": 0.6137,
+      "step": 33554
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6858256188582986e-06,
+      "loss": 0.5986,
+      "step": 33555
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.685339535995438e-06,
+      "loss": 0.6445,
+      "step": 33556
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6848534903005073e-06,
+      "loss": 0.5917,
+      "step": 33557
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6843674817759815e-06,
+      "loss": 0.582,
+      "step": 33558
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6838815104243244e-06,
+      "loss": 0.6122,
+      "step": 33559
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6833955762480134e-06,
+      "loss": 0.5848,
+      "step": 33560
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6829096792495057e-06,
+      "loss": 0.6562,
+      "step": 33561
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.68242381943128e-06,
+      "loss": 0.6132,
+      "step": 33562
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.681937996795799e-06,
+      "loss": 0.635,
+      "step": 33563
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.681452211345538e-06,
+      "loss": 0.5883,
+      "step": 33564
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.680966463082961e-06,
+      "loss": 0.6433,
+      "step": 33565
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6804807520105367e-06,
+      "loss": 0.5765,
+      "step": 33566
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6799950781307314e-06,
+      "loss": 0.5854,
+      "step": 33567
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6795094414460164e-06,
+      "loss": 0.5535,
+      "step": 33568
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6790238419588554e-06,
+      "loss": 0.6108,
+      "step": 33569
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6785382796717254e-06,
+      "loss": 0.5933,
+      "step": 33570
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6780527545870796e-06,
+      "loss": 0.6654,
+      "step": 33571
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.677567266707395e-06,
+      "loss": 0.5827,
+      "step": 33572
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.677081816035132e-06,
+      "loss": 0.5245,
+      "step": 33573
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.676596402572765e-06,
+      "loss": 0.5407,
+      "step": 33574
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.6761110263227565e-06,
+      "loss": 0.6272,
+      "step": 33575
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.675625687287572e-06,
+      "loss": 0.6125,
+      "step": 33576
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.675140385469679e-06,
+      "loss": 0.5785,
+      "step": 33577
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.67465512087154e-06,
+      "loss": 0.6321,
+      "step": 33578
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.674169893495626e-06,
+      "loss": 0.569,
+      "step": 33579
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.673684703344397e-06,
+      "loss": 0.5615,
+      "step": 33580
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6731995504203278e-06,
+      "loss": 0.599,
+      "step": 33581
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.672714434725873e-06,
+      "loss": 0.5984,
+      "step": 33582
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6722293562635037e-06,
+      "loss": 0.5871,
+      "step": 33583
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6717443150356805e-06,
+      "loss": 0.5671,
+      "step": 33584
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6712593110448725e-06,
+      "loss": 0.5543,
+      "step": 33585
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.670774344293541e-06,
+      "loss": 0.5685,
+      "step": 33586
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6702894147841564e-06,
+      "loss": 0.5447,
+      "step": 33587
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6698045225191714e-06,
+      "loss": 0.64,
+      "step": 33588
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.66931966750106e-06,
+      "loss": 0.595,
+      "step": 33589
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6688348497322793e-06,
+      "loss": 0.5933,
+      "step": 33590
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6683500692153e-06,
+      "loss": 0.6,
+      "step": 33591
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.667865325952579e-06,
+      "loss": 0.5858,
+      "step": 33592
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6673806199465824e-06,
+      "loss": 0.6032,
+      "step": 33593
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6668959511997728e-06,
+      "loss": 0.5919,
+      "step": 33594
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.666411319714609e-06,
+      "loss": 0.6079,
+      "step": 33595
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6659267254935606e-06,
+      "loss": 0.5921,
+      "step": 33596
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.665442168539083e-06,
+      "loss": 0.5763,
+      "step": 33597
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.664957648853648e-06,
+      "loss": 0.5851,
+      "step": 33598
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.664473166439705e-06,
+      "loss": 0.6308,
+      "step": 33599
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.663988721299725e-06,
+      "loss": 0.5976,
+      "step": 33600
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.663504313436164e-06,
+      "loss": 0.5388,
+      "step": 33601
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.663019942851489e-06,
+      "loss": 0.5822,
+      "step": 33602
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6625356095481593e-06,
+      "loss": 0.5805,
+      "step": 33603
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6620513135286343e-06,
+      "loss": 0.6219,
+      "step": 33604
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.661567054795372e-06,
+      "loss": 0.6192,
+      "step": 33605
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6610828333508407e-06,
+      "loss": 0.5713,
+      "step": 33606
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6605986491974934e-06,
+      "loss": 0.6309,
+      "step": 33607
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6601145023377993e-06,
+      "loss": 0.6595,
+      "step": 33608
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6596303927742072e-06,
+      "loss": 0.5653,
+      "step": 33609
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.659146320509187e-06,
+      "loss": 0.6114,
+      "step": 33610
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6586622855451906e-06,
+      "loss": 0.6026,
+      "step": 33611
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.658178287884683e-06,
+      "loss": 0.6297,
+      "step": 33612
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6576943275301226e-06,
+      "loss": 0.5979,
+      "step": 33613
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6572104044839673e-06,
+      "loss": 0.6036,
+      "step": 33614
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.656726518748677e-06,
+      "loss": 0.5172,
+      "step": 33615
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.656242670326706e-06,
+      "loss": 0.5654,
+      "step": 33616
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.655758859220521e-06,
+      "loss": 0.6381,
+      "step": 33617
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6552750854325725e-06,
+      "loss": 0.6388,
+      "step": 33618
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.654791348965328e-06,
+      "loss": 0.5845,
+      "step": 33619
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.654307649821234e-06,
+      "loss": 0.5598,
+      "step": 33620
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.653823988002756e-06,
+      "loss": 0.633,
+      "step": 33621
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.653340363512348e-06,
+      "loss": 0.627,
+      "step": 33622
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.652856776352473e-06,
+      "loss": 0.581,
+      "step": 33623
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6523732265255796e-06,
+      "loss": 0.6094,
+      "step": 33624
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.651889714034136e-06,
+      "loss": 0.6013,
+      "step": 33625
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6514062388805874e-06,
+      "loss": 0.612,
+      "step": 33626
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.650922801067399e-06,
+      "loss": 0.5724,
+      "step": 33627
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.65043940059702e-06,
+      "loss": 0.5584,
+      "step": 33628
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.649956037471915e-06,
+      "loss": 0.6024,
+      "step": 33629
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6494727116945353e-06,
+      "loss": 0.5518,
+      "step": 33630
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6489894232673387e-06,
+      "loss": 0.5523,
+      "step": 33631
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.648506172192775e-06,
+      "loss": 0.5609,
+      "step": 33632
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6480229584733076e-06,
+      "loss": 0.5792,
+      "step": 33633
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6475397821113892e-06,
+      "loss": 0.5788,
+      "step": 33634
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6470566431094748e-06,
+      "loss": 0.6422,
+      "step": 33635
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.646573541470019e-06,
+      "loss": 0.5786,
+      "step": 33636
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6460904771954732e-06,
+      "loss": 0.6066,
+      "step": 33637
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6456074502882977e-06,
+      "loss": 0.6003,
+      "step": 33638
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.645124460750942e-06,
+      "loss": 0.6248,
+      "step": 33639
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6446415085858667e-06,
+      "loss": 0.591,
+      "step": 33640
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6441585937955216e-06,
+      "loss": 0.6139,
+      "step": 33641
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.64367571638236e-06,
+      "loss": 0.5913,
+      "step": 33642
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.643192876348835e-06,
+      "loss": 0.6557,
+      "step": 33643
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.642710073697403e-06,
+      "loss": 0.6559,
+      "step": 33644
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.642227308430514e-06,
+      "loss": 0.6277,
+      "step": 33645
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.641744580550628e-06,
+      "loss": 0.6104,
+      "step": 33646
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6412618900601873e-06,
+      "loss": 0.6023,
+      "step": 33647
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6407792369616537e-06,
+      "loss": 0.5649,
+      "step": 33648
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.640296621257472e-06,
+      "loss": 0.6081,
+      "step": 33649
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6398140429501016e-06,
+      "loss": 0.6128,
+      "step": 33650
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6393315020419917e-06,
+      "loss": 0.6104,
+      "step": 33651
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6388489985355947e-06,
+      "loss": 0.5883,
+      "step": 33652
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6383665324333583e-06,
+      "loss": 0.6601,
+      "step": 33653
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.63788410373774e-06,
+      "loss": 0.6337,
+      "step": 33654
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6374017124511897e-06,
+      "loss": 0.6404,
+      "step": 33655
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.636919358576154e-06,
+      "loss": 0.5759,
+      "step": 33656
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.636437042115093e-06,
+      "loss": 0.6244,
+      "step": 33657
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.635954763070446e-06,
+      "loss": 0.5623,
+      "step": 33658
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.635472521444673e-06,
+      "loss": 0.5726,
+      "step": 33659
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6349903172402182e-06,
+      "loss": 0.6073,
+      "step": 33660
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.634508150459537e-06,
+      "loss": 0.6712,
+      "step": 33661
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6340260211050774e-06,
+      "loss": 0.6043,
+      "step": 33662
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6335439291792888e-06,
+      "loss": 0.5681,
+      "step": 33663
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.633061874684617e-06,
+      "loss": 0.6082,
+      "step": 33664
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6325798576235184e-06,
+      "loss": 0.5903,
+      "step": 33665
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6320978779984375e-06,
+      "loss": 0.5424,
+      "step": 33666
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6316159358118267e-06,
+      "loss": 0.5959,
+      "step": 33667
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6311340310661338e-06,
+      "loss": 0.623,
+      "step": 33668
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.630652163763807e-06,
+      "loss": 0.6006,
+      "step": 33669
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6301703339072914e-06,
+      "loss": 0.5741,
+      "step": 33670
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6296885414990424e-06,
+      "loss": 0.6114,
+      "step": 33671
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.629206786541504e-06,
+      "loss": 0.573,
+      "step": 33672
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6287250690371247e-06,
+      "loss": 0.5825,
+      "step": 33673
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6282433889883485e-06,
+      "loss": 0.5533,
+      "step": 33674
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6277617463976303e-06,
+      "loss": 0.6806,
+      "step": 33675
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.627280141267413e-06,
+      "loss": 0.5831,
+      "step": 33676
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.626798573600142e-06,
+      "loss": 0.6514,
+      "step": 33677
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.626317043398271e-06,
+      "loss": 0.5778,
+      "step": 33678
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6258355506642406e-06,
+      "loss": 0.6067,
+      "step": 33679
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.625354095400501e-06,
+      "loss": 0.5666,
+      "step": 33680
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6248726776094922e-06,
+      "loss": 0.5745,
+      "step": 33681
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.624391297293669e-06,
+      "loss": 0.6002,
+      "step": 33682
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6239099544554715e-06,
+      "loss": 0.6195,
+      "step": 33683
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6234286490973527e-06,
+      "loss": 0.6647,
+      "step": 33684
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6229473812217474e-06,
+      "loss": 0.6367,
+      "step": 33685
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.62246615083111e-06,
+      "loss": 0.6568,
+      "step": 33686
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.62198495792788e-06,
+      "loss": 0.6252,
+      "step": 33687
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.621503802514508e-06,
+      "loss": 0.6115,
+      "step": 33688
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.621022684593436e-06,
+      "loss": 0.5752,
+      "step": 33689
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6205416041671093e-06,
+      "loss": 0.5598,
+      "step": 33690
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.620060561237968e-06,
+      "loss": 0.5951,
+      "step": 33691
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6195795558084645e-06,
+      "loss": 0.6507,
+      "step": 33692
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.619098587881038e-06,
+      "loss": 0.604,
+      "step": 33693
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6186176574581303e-06,
+      "loss": 0.5807,
+      "step": 33694
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.618136764542194e-06,
+      "loss": 0.5458,
+      "step": 33695
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.61765590913566e-06,
+      "loss": 0.6591,
+      "step": 33696
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6171750912409822e-06,
+      "loss": 0.6515,
+      "step": 33697
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6166943108605967e-06,
+      "loss": 0.5863,
+      "step": 33698
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6162135679969525e-06,
+      "loss": 0.6122,
+      "step": 33699
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.615732862652489e-06,
+      "loss": 0.5842,
+      "step": 33700
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6152521948296503e-06,
+      "loss": 0.5944,
+      "step": 33701
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6147715645308746e-06,
+      "loss": 0.5881,
+      "step": 33702
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.614290971758611e-06,
+      "loss": 0.5872,
+      "step": 33703
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6138104165152946e-06,
+      "loss": 0.6516,
+      "step": 33704
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6133298988033738e-06,
+      "loss": 0.5905,
+      "step": 33705
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6128494186252875e-06,
+      "loss": 0.5869,
+      "step": 33706
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.612368975983476e-06,
+      "loss": 0.6355,
+      "step": 33707
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6118885708803774e-06,
+      "loss": 0.6188,
+      "step": 33708
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6114082033184407e-06,
+      "loss": 0.6185,
+      "step": 33709
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.610927873300103e-06,
+      "loss": 0.587,
+      "step": 33710
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.610447580827805e-06,
+      "loss": 0.5992,
+      "step": 33711
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6099673259039836e-06,
+      "loss": 0.5379,
+      "step": 33712
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6094871085310846e-06,
+      "loss": 0.6704,
+      "step": 33713
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.609006928711547e-06,
+      "loss": 0.5922,
+      "step": 33714
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.608526786447807e-06,
+      "loss": 0.6237,
+      "step": 33715
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.60804668174231e-06,
+      "loss": 0.5897,
+      "step": 33716
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.607566614597492e-06,
+      "loss": 0.5574,
+      "step": 33717
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6070865850157933e-06,
+      "loss": 0.5268,
+      "step": 33718
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6066065929996496e-06,
+      "loss": 0.58,
+      "step": 33719
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.606126638551506e-06,
+      "loss": 0.5747,
+      "step": 33720
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6056467216737957e-06,
+      "loss": 0.5776,
+      "step": 33721
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.6051668423689657e-06,
+      "loss": 0.6185,
+      "step": 33722
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.604687000639443e-06,
+      "loss": 0.5384,
+      "step": 33723
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6042071964876735e-06,
+      "loss": 0.5649,
+      "step": 33724
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.603727429916092e-06,
+      "loss": 0.6147,
+      "step": 33725
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6032477009271395e-06,
+      "loss": 0.5852,
+      "step": 33726
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.602768009523251e-06,
+      "loss": 0.5886,
+      "step": 33727
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6022883557068658e-06,
+      "loss": 0.6207,
+      "step": 33728
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6018087394804158e-06,
+      "loss": 0.6575,
+      "step": 33729
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6013291608463463e-06,
+      "loss": 0.6324,
+      "step": 33730
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.6008496198070897e-06,
+      "loss": 0.5695,
+      "step": 33731
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.60037011636508e-06,
+      "loss": 0.5974,
+      "step": 33732
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5998906505227604e-06,
+      "loss": 0.5919,
+      "step": 33733
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.599411222282563e-06,
+      "loss": 0.5242,
+      "step": 33734
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5989318316469246e-06,
+      "loss": 0.6086,
+      "step": 33735
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5984524786182785e-06,
+      "loss": 0.5942,
+      "step": 33736
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5979731631990656e-06,
+      "loss": 0.586,
+      "step": 33737
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5974938853917187e-06,
+      "loss": 0.5787,
+      "step": 33738
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5970146451986732e-06,
+      "loss": 0.5943,
+      "step": 33739
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.596535442622362e-06,
+      "loss": 0.5677,
+      "step": 33740
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5960562776652255e-06,
+      "loss": 0.5983,
+      "step": 33741
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.595577150329691e-06,
+      "loss": 0.6151,
+      "step": 33742
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.595098060618202e-06,
+      "loss": 0.5712,
+      "step": 33743
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.594619008533188e-06,
+      "loss": 0.6361,
+      "step": 33744
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.594139994077084e-06,
+      "loss": 0.5379,
+      "step": 33745
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.593661017252319e-06,
+      "loss": 0.658,
+      "step": 33746
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5931820780613358e-06,
+      "loss": 0.6374,
+      "step": 33747
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5927031765065637e-06,
+      "loss": 0.5959,
+      "step": 33748
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5922243125904358e-06,
+      "loss": 0.6118,
+      "step": 33749
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5917454863153823e-06,
+      "loss": 0.6066,
+      "step": 33750
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.591266697683843e-06,
+      "loss": 0.6267,
+      "step": 33751
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5907879466982465e-06,
+      "loss": 0.6275,
+      "step": 33752
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.590309233361025e-06,
+      "loss": 0.571,
+      "step": 33753
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5898305576746143e-06,
+      "loss": 0.6009,
+      "step": 33754
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.589351919641444e-06,
+      "loss": 0.6116,
+      "step": 33755
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.588873319263948e-06,
+      "loss": 0.6108,
+      "step": 33756
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.588394756544554e-06,
+      "loss": 0.6016,
+      "step": 33757
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5879162314856996e-06,
+      "loss": 0.5689,
+      "step": 33758
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.58743774408981e-06,
+      "loss": 0.6452,
+      "step": 33759
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.586959294359327e-06,
+      "loss": 0.5875,
+      "step": 33760
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5864808822966682e-06,
+      "loss": 0.6309,
+      "step": 33761
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5860025079042748e-06,
+      "loss": 0.5992,
+      "step": 33762
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5855241711845704e-06,
+      "loss": 0.5968,
+      "step": 33763
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5850458721399927e-06,
+      "loss": 0.5482,
+      "step": 33764
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5845676107729678e-06,
+      "loss": 0.5886,
+      "step": 33765
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5840893870859274e-06,
+      "loss": 0.5435,
+      "step": 33766
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5836112010812965e-06,
+      "loss": 0.5951,
+      "step": 33767
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.583133052761513e-06,
+      "loss": 0.5826,
+      "step": 33768
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.582654942129003e-06,
+      "loss": 0.6466,
+      "step": 33769
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5821768691861917e-06,
+      "loss": 0.6046,
+      "step": 33770
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5816988339355154e-06,
+      "loss": 0.6317,
+      "step": 33771
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.581220836379399e-06,
+      "loss": 0.603,
+      "step": 33772
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5807428765202737e-06,
+      "loss": 0.6273,
+      "step": 33773
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.580264954360563e-06,
+      "loss": 0.623,
+      "step": 33774
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5797870699027015e-06,
+      "loss": 0.6223,
+      "step": 33775
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5793092231491156e-06,
+      "loss": 0.5995,
+      "step": 33776
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.578831414102233e-06,
+      "loss": 0.6428,
+      "step": 33777
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5783536427644772e-06,
+      "loss": 0.6005,
+      "step": 33778
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.577875909138284e-06,
+      "loss": 0.5476,
+      "step": 33779
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5773982132260742e-06,
+      "loss": 0.5502,
+      "step": 33780
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.576920555030281e-06,
+      "loss": 0.6254,
+      "step": 33781
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.576442934553328e-06,
+      "loss": 0.5457,
+      "step": 33782
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5759653517976422e-06,
+      "loss": 0.6119,
+      "step": 33783
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5754878067656485e-06,
+      "loss": 0.5798,
+      "step": 33784
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.575010299459778e-06,
+      "loss": 0.6217,
+      "step": 33785
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5745328298824547e-06,
+      "loss": 0.6092,
+      "step": 33786
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.574055398036105e-06,
+      "loss": 0.6165,
+      "step": 33787
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5735780039231517e-06,
+      "loss": 0.6594,
+      "step": 33788
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.573100647546026e-06,
+      "loss": 0.6007,
+      "step": 33789
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.572623328907151e-06,
+      "loss": 0.5851,
+      "step": 33790
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.57214604800895e-06,
+      "loss": 0.62,
+      "step": 33791
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5716688048538518e-06,
+      "loss": 0.6112,
+      "step": 33792
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5711915994442817e-06,
+      "loss": 0.6088,
+      "step": 33793
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5707144317826616e-06,
+      "loss": 0.6437,
+      "step": 33794
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5702373018714144e-06,
+      "loss": 0.6208,
+      "step": 33795
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5697602097129704e-06,
+      "loss": 0.5525,
+      "step": 33796
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5692831553097473e-06,
+      "loss": 0.5928,
+      "step": 33797
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.568806138664179e-06,
+      "loss": 0.5426,
+      "step": 33798
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5683291597786774e-06,
+      "loss": 0.5495,
+      "step": 33799
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5678522186556743e-06,
+      "loss": 0.6258,
+      "step": 33800
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5673753152975878e-06,
+      "loss": 0.5891,
+      "step": 33801
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.566898449706847e-06,
+      "loss": 0.5836,
+      "step": 33802
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.566421621885873e-06,
+      "loss": 0.5939,
+      "step": 33803
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.565944831837087e-06,
+      "loss": 0.6473,
+      "step": 33804
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5654680795629095e-06,
+      "loss": 0.6376,
+      "step": 33805
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5649913650657686e-06,
+      "loss": 0.5614,
+      "step": 33806
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.564514688348084e-06,
+      "loss": 0.6183,
+      "step": 33807
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.564038049412276e-06,
+      "loss": 0.6666,
+      "step": 33808
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5635614482607708e-06,
+      "loss": 0.6699,
+      "step": 33809
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5630848848959875e-06,
+      "loss": 0.5315,
+      "step": 33810
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5626083593203476e-06,
+      "loss": 0.5781,
+      "step": 33811
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.562131871536271e-06,
+      "loss": 0.5493,
+      "step": 33812
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.561655421546182e-06,
+      "loss": 0.5883,
+      "step": 33813
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5611790093525014e-06,
+      "loss": 0.5786,
+      "step": 33814
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.560702634957648e-06,
+      "loss": 0.6244,
+      "step": 33815
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.560226298364041e-06,
+      "loss": 0.5755,
+      "step": 33816
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.559749999574106e-06,
+      "loss": 0.6103,
+      "step": 33817
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5592737385902566e-06,
+      "loss": 0.6223,
+      "step": 33818
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5587975154149224e-06,
+      "loss": 0.5995,
+      "step": 33819
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5583213300505118e-06,
+      "loss": 0.5945,
+      "step": 33820
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.557845182499453e-06,
+      "loss": 0.5911,
+      "step": 33821
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5573690727641587e-06,
+      "loss": 0.6102,
+      "step": 33822
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.556893000847056e-06,
+      "loss": 0.5083,
+      "step": 33823
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5564169667505556e-06,
+      "loss": 0.6206,
+      "step": 33824
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5559409704770864e-06,
+      "loss": 0.5904,
+      "step": 33825
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.555465012029056e-06,
+      "loss": 0.5702,
+      "step": 33826
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5549890914088916e-06,
+      "loss": 0.5874,
+      "step": 33827
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5545132086190073e-06,
+      "loss": 0.6019,
+      "step": 33828
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5540373636618187e-06,
+      "loss": 0.5719,
+      "step": 33829
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5535615565397496e-06,
+      "loss": 0.6201,
+      "step": 33830
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.553085787255216e-06,
+      "loss": 0.6907,
+      "step": 33831
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.552610055810634e-06,
+      "loss": 0.5949,
+      "step": 33832
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5521343622084184e-06,
+      "loss": 0.562,
+      "step": 33833
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5516587064509925e-06,
+      "loss": 0.6113,
+      "step": 33834
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.551183088540766e-06,
+      "loss": 0.6369,
+      "step": 33835
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.550707508480167e-06,
+      "loss": 0.6026,
+      "step": 33836
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.550231966271598e-06,
+      "loss": 0.6155,
+      "step": 33837
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5497564619174854e-06,
+      "loss": 0.5883,
+      "step": 33838
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.549280995420239e-06,
+      "loss": 0.5639,
+      "step": 33839
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.548805566782282e-06,
+      "loss": 0.5527,
+      "step": 33840
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5483301760060254e-06,
+      "loss": 0.5742,
+      "step": 33841
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.547854823093885e-06,
+      "loss": 0.5404,
+      "step": 33842
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5473795080482734e-06,
+      "loss": 0.6266,
+      "step": 33843
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.546904230871613e-06,
+      "loss": 0.6235,
+      "step": 33844
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.546428991566312e-06,
+      "loss": 0.6007,
+      "step": 33845
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.545953790134793e-06,
+      "loss": 0.5996,
+      "step": 33846
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5454786265794605e-06,
+      "loss": 0.6145,
+      "step": 33847
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5450035009027374e-06,
+      "loss": 0.6062,
+      "step": 33848
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5445284131070357e-06,
+      "loss": 0.5783,
+      "step": 33849
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.544053363194764e-06,
+      "loss": 0.5961,
+      "step": 33850
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5435783511683444e-06,
+      "loss": 0.5693,
+      "step": 33851
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.543103377030187e-06,
+      "loss": 0.6088,
+      "step": 33852
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5426284407827053e-06,
+      "loss": 0.5934,
+      "step": 33853
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5421535424283096e-06,
+      "loss": 0.6356,
+      "step": 33854
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5416786819694184e-06,
+      "loss": 0.6593,
+      "step": 33855
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.541203859408439e-06,
+      "loss": 0.5362,
+      "step": 33856
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.540729074747794e-06,
+      "loss": 0.5928,
+      "step": 33857
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.540254327989883e-06,
+      "loss": 0.5989,
+      "step": 33858
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.539779619137126e-06,
+      "loss": 0.6394,
+      "step": 33859
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.539304948191932e-06,
+      "loss": 0.5608,
+      "step": 33860
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5388303151567173e-06,
+      "loss": 0.5812,
+      "step": 33861
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5383557200338873e-06,
+      "loss": 0.5599,
+      "step": 33862
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.537881162825864e-06,
+      "loss": 0.5888,
+      "step": 33863
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5374066435350454e-06,
+      "loss": 0.6056,
+      "step": 33864
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.536932162163852e-06,
+      "loss": 0.6106,
+      "step": 33865
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5364577187146886e-06,
+      "loss": 0.5942,
+      "step": 33866
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5359833131899737e-06,
+      "loss": 0.5887,
+      "step": 33867
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5355089455921123e-06,
+      "loss": 0.6056,
+      "step": 33868
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5350346159235164e-06,
+      "loss": 0.6113,
+      "step": 33869
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.5345603241865958e-06,
+      "loss": 0.5712,
+      "step": 33870
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5340860703837577e-06,
+      "loss": 0.6279,
+      "step": 33871
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5336118545174183e-06,
+      "loss": 0.6346,
+      "step": 33872
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5331376765899794e-06,
+      "loss": 0.5414,
+      "step": 33873
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5326635366038623e-06,
+      "loss": 0.6318,
+      "step": 33874
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.532189434561462e-06,
+      "loss": 0.541,
+      "step": 33875
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5317153704651963e-06,
+      "loss": 0.656,
+      "step": 33876
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.53124134431747e-06,
+      "loss": 0.612,
+      "step": 33877
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.530767356120696e-06,
+      "loss": 0.6627,
+      "step": 33878
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.530293405877281e-06,
+      "loss": 0.6478,
+      "step": 33879
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5298194935896325e-06,
+      "loss": 0.6093,
+      "step": 33880
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.529345619260156e-06,
+      "loss": 0.6173,
+      "step": 33881
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5288717828912667e-06,
+      "loss": 0.5685,
+      "step": 33882
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5283979844853635e-06,
+      "loss": 0.5652,
+      "step": 33883
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.527924224044864e-06,
+      "loss": 0.5327,
+      "step": 33884
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.527450501572165e-06,
+      "loss": 0.5985,
+      "step": 33885
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5269768170696805e-06,
+      "loss": 0.6132,
+      "step": 33886
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.526503170539816e-06,
+      "loss": 0.5812,
+      "step": 33887
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.526029561984974e-06,
+      "loss": 0.6077,
+      "step": 33888
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.525555991407568e-06,
+      "loss": 0.6096,
+      "step": 33889
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.525082458810001e-06,
+      "loss": 0.6019,
+      "step": 33890
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.524608964194678e-06,
+      "loss": 0.6343,
+      "step": 33891
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.524135507564005e-06,
+      "loss": 0.6214,
+      "step": 33892
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5236620889203913e-06,
+      "loss": 0.6411,
+      "step": 33893
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5231887082662364e-06,
+      "loss": 0.5413,
+      "step": 33894
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.522715365603956e-06,
+      "loss": 0.5611,
+      "step": 33895
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.522242060935942e-06,
+      "loss": 0.6019,
+      "step": 33896
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5217687942646095e-06,
+      "loss": 0.6346,
+      "step": 33897
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5212955655923566e-06,
+      "loss": 0.6216,
+      "step": 33898
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5208223749215944e-06,
+      "loss": 0.5921,
+      "step": 33899
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.520349222254722e-06,
+      "loss": 0.6484,
+      "step": 33900
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.51987610759415e-06,
+      "loss": 0.5729,
+      "step": 33901
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5194030309422735e-06,
+      "loss": 0.6034,
+      "step": 33902
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.518929992301503e-06,
+      "loss": 0.649,
+      "step": 33903
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5184569916742387e-06,
+      "loss": 0.6433,
+      "step": 33904
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5179840290628865e-06,
+      "loss": 0.5486,
+      "step": 33905
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5175111044698497e-06,
+      "loss": 0.6894,
+      "step": 33906
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5170382178975308e-06,
+      "loss": 0.636,
+      "step": 33907
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5165653693483315e-06,
+      "loss": 0.6444,
+      "step": 33908
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.516092558824652e-06,
+      "loss": 0.6166,
+      "step": 33909
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5156197863289013e-06,
+      "loss": 0.616,
+      "step": 33910
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.515147051863478e-06,
+      "loss": 0.6388,
+      "step": 33911
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.514674355430785e-06,
+      "loss": 0.5619,
+      "step": 33912
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5142016970332208e-06,
+      "loss": 0.5968,
+      "step": 33913
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5137290766731928e-06,
+      "loss": 0.6399,
+      "step": 33914
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5132564943530968e-06,
+      "loss": 0.6235,
+      "step": 33915
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5127839500753405e-06,
+      "loss": 0.5958,
+      "step": 33916
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.512311443842321e-06,
+      "loss": 0.6071,
+      "step": 33917
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.511838975656441e-06,
+      "loss": 0.5726,
+      "step": 33918
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5113665455200963e-06,
+      "loss": 0.6683,
+      "step": 33919
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5108941534356958e-06,
+      "loss": 0.6087,
+      "step": 33920
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.510421799405631e-06,
+      "loss": 0.5894,
+      "step": 33921
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5099494834323124e-06,
+      "loss": 0.6044,
+      "step": 33922
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5094772055181284e-06,
+      "loss": 0.6252,
+      "step": 33923
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.509004965665487e-06,
+      "loss": 0.5782,
+      "step": 33924
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.508532763876783e-06,
+      "loss": 0.6699,
+      "step": 33925
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5080606001544204e-06,
+      "loss": 0.6376,
+      "step": 33926
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.507588474500796e-06,
+      "loss": 0.6435,
+      "step": 33927
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.507116386918309e-06,
+      "loss": 0.5857,
+      "step": 33928
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5066443374093584e-06,
+      "loss": 0.5533,
+      "step": 33929
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5061723259763394e-06,
+      "loss": 0.6204,
+      "step": 33930
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.505700352621656e-06,
+      "loss": 0.6088,
+      "step": 33931
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5052284173477005e-06,
+      "loss": 0.6001,
+      "step": 33932
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5047565201568815e-06,
+      "loss": 0.5638,
+      "step": 33933
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5042846610515827e-06,
+      "loss": 0.5122,
+      "step": 33934
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.503812840034212e-06,
+      "loss": 0.5787,
+      "step": 33935
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5033410571071615e-06,
+      "loss": 0.5938,
+      "step": 33936
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.502869312272833e-06,
+      "loss": 0.6336,
+      "step": 33937
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5023976055336207e-06,
+      "loss": 0.5941,
+      "step": 33938
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5019259368919225e-06,
+      "loss": 0.6224,
+      "step": 33939
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.501454306350132e-06,
+      "loss": 0.5859,
+      "step": 33940
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5009827139106503e-06,
+      "loss": 0.6495,
+      "step": 33941
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.500511159575868e-06,
+      "loss": 0.6071,
+      "step": 33942
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.5000396433481887e-06,
+      "loss": 0.6301,
+      "step": 33943
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4995681652300052e-06,
+      "loss": 0.5976,
+      "step": 33944
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.499096725223711e-06,
+      "loss": 0.5856,
+      "step": 33945
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4986253233317017e-06,
+      "loss": 0.5506,
+      "step": 33946
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4981539595563753e-06,
+      "loss": 0.5998,
+      "step": 33947
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.497682633900127e-06,
+      "loss": 0.595,
+      "step": 33948
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4972113463653493e-06,
+      "loss": 0.6445,
+      "step": 33949
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4967400969544388e-06,
+      "loss": 0.5611,
+      "step": 33950
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.496268885669786e-06,
+      "loss": 0.6461,
+      "step": 33951
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.495797712513792e-06,
+      "loss": 0.5904,
+      "step": 33952
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4953265774888437e-06,
+      "loss": 0.5932,
+      "step": 33953
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4948554805973424e-06,
+      "loss": 0.5871,
+      "step": 33954
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.494384421841679e-06,
+      "loss": 0.6186,
+      "step": 33955
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.493913401224245e-06,
+      "loss": 0.6363,
+      "step": 33956
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4934424187474338e-06,
+      "loss": 0.6306,
+      "step": 33957
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4929714744136424e-06,
+      "loss": 0.5372,
+      "step": 33958
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.492500568225258e-06,
+      "loss": 0.5874,
+      "step": 33959
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.492029700184684e-06,
+      "loss": 0.6083,
+      "step": 33960
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4915588702942983e-06,
+      "loss": 0.6056,
+      "step": 33961
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4910880785565052e-06,
+      "loss": 0.6976,
+      "step": 33962
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.49061732497369e-06,
+      "loss": 0.604,
+      "step": 33963
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4901466095482498e-06,
+      "loss": 0.632,
+      "step": 33964
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4896759322825736e-06,
+      "loss": 0.6409,
+      "step": 33965
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4892052931790533e-06,
+      "loss": 0.5923,
+      "step": 33966
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4887346922400814e-06,
+      "loss": 0.5846,
+      "step": 33967
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4882641294680456e-06,
+      "loss": 0.6182,
+      "step": 33968
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.487793604865342e-06,
+      "loss": 0.5637,
+      "step": 33969
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.487323118434356e-06,
+      "loss": 0.5031,
+      "step": 33970
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4868526701774876e-06,
+      "loss": 0.6738,
+      "step": 33971
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4863822600971145e-06,
+      "loss": 0.6096,
+      "step": 33972
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.485911888195638e-06,
+      "loss": 0.5852,
+      "step": 33973
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4854415544754397e-06,
+      "loss": 0.6017,
+      "step": 33974
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4849712589389174e-06,
+      "loss": 0.6571,
+      "step": 33975
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4845010015884565e-06,
+      "loss": 0.6254,
+      "step": 33976
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4840307824264463e-06,
+      "loss": 0.6217,
+      "step": 33977
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.483560601455275e-06,
+      "loss": 0.6111,
+      "step": 33978
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.483090458677336e-06,
+      "loss": 0.581,
+      "step": 33979
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4826203540950135e-06,
+      "loss": 0.55,
+      "step": 33980
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4821502877107008e-06,
+      "loss": 0.6383,
+      "step": 33981
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.481680259526784e-06,
+      "loss": 0.59,
+      "step": 33982
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.481210269545653e-06,
+      "loss": 0.5978,
+      "step": 33983
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.48074031776969e-06,
+      "loss": 0.5487,
+      "step": 33984
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4802704042012914e-06,
+      "loss": 0.5938,
+      "step": 33985
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4798005288428407e-06,
+      "loss": 0.6035,
+      "step": 33986
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4793306916967263e-06,
+      "loss": 0.6067,
+      "step": 33987
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.478860892765336e-06,
+      "loss": 0.659,
+      "step": 33988
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4783911320510514e-06,
+      "loss": 0.5732,
+      "step": 33989
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4779214095562686e-06,
+      "loss": 0.6119,
+      "step": 33990
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4774517252833665e-06,
+      "loss": 0.6393,
+      "step": 33991
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4769820792347387e-06,
+      "loss": 0.6031,
+      "step": 33992
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4765124714127676e-06,
+      "loss": 0.6535,
+      "step": 33993
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4760429018198396e-06,
+      "loss": 0.5897,
+      "step": 33994
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4755733704583395e-06,
+      "loss": 0.6315,
+      "step": 33995
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.475103877330657e-06,
+      "loss": 0.6039,
+      "step": 33996
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4746344224391715e-06,
+      "loss": 0.5824,
+      "step": 33997
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.47416500578628e-06,
+      "loss": 0.5879,
+      "step": 33998
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.473695627374353e-06,
+      "loss": 0.594,
+      "step": 33999
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4732262872057867e-06,
+      "loss": 0.5662,
+      "step": 34000
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.472756985282958e-06,
+      "loss": 0.5994,
+      "step": 34001
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4722877216082597e-06,
+      "loss": 0.6365,
+      "step": 34002
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.471818496184071e-06,
+      "loss": 0.6198,
+      "step": 34003
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.471349309012777e-06,
+      "loss": 0.6244,
+      "step": 34004
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4708801600967603e-06,
+      "loss": 0.6334,
+      "step": 34005
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.470411049438408e-06,
+      "loss": 0.6056,
+      "step": 34006
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4699419770401036e-06,
+      "loss": 0.6114,
+      "step": 34007
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4694729429042264e-06,
+      "loss": 0.5605,
+      "step": 34008
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4690039470331673e-06,
+      "loss": 0.5677,
+      "step": 34009
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4685349894293e-06,
+      "loss": 0.5798,
+      "step": 34010
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4680660700950143e-06,
+      "loss": 0.6073,
+      "step": 34011
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4675971890326876e-06,
+      "loss": 0.5724,
+      "step": 34012
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4671283462447093e-06,
+      "loss": 0.5892,
+      "step": 34013
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.466659541733457e-06,
+      "loss": 0.6084,
+      "step": 34014
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4661907755013146e-06,
+      "loss": 0.6147,
+      "step": 34015
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.4657220475506603e-06,
+      "loss": 0.5732,
+      "step": 34016
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.465253357883881e-06,
+      "loss": 0.6052,
+      "step": 34017
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4647847065033535e-06,
+      "loss": 0.5984,
+      "step": 34018
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.464316093411465e-06,
+      "loss": 0.6525,
+      "step": 34019
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4638475186105937e-06,
+      "loss": 0.5915,
+      "step": 34020
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4633789821031197e-06,
+      "loss": 0.5704,
+      "step": 34021
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.462910483891421e-06,
+      "loss": 0.5678,
+      "step": 34022
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4624420239778847e-06,
+      "loss": 0.5955,
+      "step": 34023
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.461973602364888e-06,
+      "loss": 0.5858,
+      "step": 34024
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4615052190548107e-06,
+      "loss": 0.5906,
+      "step": 34025
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.46103687405003e-06,
+      "loss": 0.6931,
+      "step": 34026
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.460568567352932e-06,
+      "loss": 0.6471,
+      "step": 34027
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.460100298965894e-06,
+      "loss": 0.6232,
+      "step": 34028
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4596320688912913e-06,
+      "loss": 0.6344,
+      "step": 34029
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.459163877131512e-06,
+      "loss": 0.6492,
+      "step": 34030
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.458695723688923e-06,
+      "loss": 0.6067,
+      "step": 34031
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.458227608565913e-06,
+      "loss": 0.5884,
+      "step": 34032
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.457759531764855e-06,
+      "loss": 0.6147,
+      "step": 34033
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.457291493288132e-06,
+      "loss": 0.6645,
+      "step": 34034
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4568234931381163e-06,
+      "loss": 0.5781,
+      "step": 34035
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.456355531317196e-06,
+      "loss": 0.5606,
+      "step": 34036
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4558876078277372e-06,
+      "loss": 0.5246,
+      "step": 34037
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4554197226721245e-06,
+      "loss": 0.5982,
+      "step": 34038
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4549518758527324e-06,
+      "loss": 0.577,
+      "step": 34039
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4544840673719418e-06,
+      "loss": 0.6096,
+      "step": 34040
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4540162972321267e-06,
+      "loss": 0.6885,
+      "step": 34041
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.453548565435665e-06,
+      "loss": 0.5993,
+      "step": 34042
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4530808719849308e-06,
+      "loss": 0.5746,
+      "step": 34043
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4526132168823056e-06,
+      "loss": 0.6019,
+      "step": 34044
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4521456001301623e-06,
+      "loss": 0.5631,
+      "step": 34045
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4516780217308755e-06,
+      "loss": 0.5775,
+      "step": 34046
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4512104816868266e-06,
+      "loss": 0.6465,
+      "step": 34047
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.450742980000387e-06,
+      "loss": 0.6269,
+      "step": 34048
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4502755166739343e-06,
+      "loss": 0.5134,
+      "step": 34049
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4498080917098387e-06,
+      "loss": 0.6497,
+      "step": 34050
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4493407051104833e-06,
+      "loss": 0.615,
+      "step": 34051
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4488733568782384e-06,
+      "loss": 0.577,
+      "step": 34052
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4484060470154793e-06,
+      "loss": 0.605,
+      "step": 34053
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4479387755245786e-06,
+      "loss": 0.6448,
+      "step": 34054
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4474715424079156e-06,
+      "loss": 0.5745,
+      "step": 34055
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4470043476678573e-06,
+      "loss": 0.6192,
+      "step": 34056
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4465371913067857e-06,
+      "loss": 0.6491,
+      "step": 34057
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4460700733270703e-06,
+      "loss": 0.6053,
+      "step": 34058
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4456029937310865e-06,
+      "loss": 0.6373,
+      "step": 34059
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4451359525212027e-06,
+      "loss": 0.573,
+      "step": 34060
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.444668949699798e-06,
+      "loss": 0.5745,
+      "step": 34061
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4442019852692435e-06,
+      "loss": 0.5657,
+      "step": 34062
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.443735059231911e-06,
+      "loss": 0.5896,
+      "step": 34063
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.443268171590171e-06,
+      "loss": 0.631,
+      "step": 34064
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.442801322346402e-06,
+      "loss": 0.6548,
+      "step": 34065
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.442334511502972e-06,
+      "loss": 0.5662,
+      "step": 34066
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4418677390622512e-06,
+      "loss": 0.5929,
+      "step": 34067
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.441401005026619e-06,
+      "loss": 0.576,
+      "step": 34068
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.440934309398437e-06,
+      "loss": 0.5384,
+      "step": 34069
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.440467652180084e-06,
+      "loss": 0.6271,
+      "step": 34070
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4400010333739267e-06,
+      "loss": 0.6297,
+      "step": 34071
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4395344529823394e-06,
+      "loss": 0.6492,
+      "step": 34072
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.439067911007691e-06,
+      "loss": 0.5863,
+      "step": 34073
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4386014074523566e-06,
+      "loss": 0.5831,
+      "step": 34074
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4381349423186986e-06,
+      "loss": 0.592,
+      "step": 34075
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.437668515609094e-06,
+      "loss": 0.6026,
+      "step": 34076
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4372021273259073e-06,
+      "loss": 0.6249,
+      "step": 34077
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.436735777471514e-06,
+      "loss": 0.5794,
+      "step": 34078
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.436269466048282e-06,
+      "loss": 0.6116,
+      "step": 34079
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.435803193058579e-06,
+      "loss": 0.6698,
+      "step": 34080
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.435336958504774e-06,
+      "loss": 0.6174,
+      "step": 34081
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4348707623892385e-06,
+      "loss": 0.5995,
+      "step": 34082
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.434404604714341e-06,
+      "loss": 0.6052,
+      "step": 34083
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4339384854824465e-06,
+      "loss": 0.5894,
+      "step": 34084
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.433472404695929e-06,
+      "loss": 0.5971,
+      "step": 34085
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4330063623571542e-06,
+      "loss": 0.5905,
+      "step": 34086
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.43254035846849e-06,
+      "loss": 0.5874,
+      "step": 34087
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.432074393032303e-06,
+      "loss": 0.6199,
+      "step": 34088
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4316084660509644e-06,
+      "loss": 0.5753,
+      "step": 34089
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4311425775268393e-06,
+      "loss": 0.5797,
+      "step": 34090
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4306767274622956e-06,
+      "loss": 0.6505,
+      "step": 34091
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4302109158596977e-06,
+      "loss": 0.6065,
+      "step": 34092
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.429745142721417e-06,
+      "loss": 0.576,
+      "step": 34093
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.429279408049816e-06,
+      "loss": 0.6448,
+      "step": 34094
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4288137118472687e-06,
+      "loss": 0.662,
+      "step": 34095
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4283480541161318e-06,
+      "loss": 0.629,
+      "step": 34096
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.427882434858778e-06,
+      "loss": 0.5499,
+      "step": 34097
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4274168540775677e-06,
+      "loss": 0.6064,
+      "step": 34098
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4269513117748745e-06,
+      "loss": 0.5746,
+      "step": 34099
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4264858079530584e-06,
+      "loss": 0.5598,
+      "step": 34100
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4260203426144856e-06,
+      "loss": 0.6331,
+      "step": 34101
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.42555491576152e-06,
+      "loss": 0.64,
+      "step": 34102
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.42508952739653e-06,
+      "loss": 0.5584,
+      "step": 34103
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.424624177521878e-06,
+      "loss": 0.5969,
+      "step": 34104
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4241588661399274e-06,
+      "loss": 0.6159,
+      "step": 34105
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.423693593253047e-06,
+      "loss": 0.6673,
+      "step": 34106
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4232283588635985e-06,
+      "loss": 0.5576,
+      "step": 34107
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4227631629739446e-06,
+      "loss": 0.5705,
+      "step": 34108
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4222980055864474e-06,
+      "loss": 0.5127,
+      "step": 34109
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.421832886703477e-06,
+      "loss": 0.6372,
+      "step": 34110
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4213678063273906e-06,
+      "loss": 0.5407,
+      "step": 34111
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4209027644605588e-06,
+      "loss": 0.6432,
+      "step": 34112
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4204377611053343e-06,
+      "loss": 0.5722,
+      "step": 34113
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.419972796264087e-06,
+      "loss": 0.6335,
+      "step": 34114
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.419507869939176e-06,
+      "loss": 0.6029,
+      "step": 34115
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.419042982132969e-06,
+      "loss": 0.6171,
+      "step": 34116
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.418578132847824e-06,
+      "loss": 0.6048,
+      "step": 34117
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4181133220861043e-06,
+      "loss": 0.5598,
+      "step": 34118
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.417648549850168e-06,
+      "loss": 0.5968,
+      "step": 34119
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.417183816142382e-06,
+      "loss": 0.5852,
+      "step": 34120
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4167191209651076e-06,
+      "loss": 0.6726,
+      "step": 34121
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4162544643207032e-06,
+      "loss": 0.6172,
+      "step": 34122
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4157898462115283e-06,
+      "loss": 0.6041,
+      "step": 34123
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4153252666399486e-06,
+      "loss": 0.6244,
+      "step": 34124
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4148607256083222e-06,
+      "loss": 0.6436,
+      "step": 34125
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.414396223119008e-06,
+      "loss": 0.5332,
+      "step": 34126
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4139317591743706e-06,
+      "loss": 0.5967,
+      "step": 34127
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4134673337767676e-06,
+      "loss": 0.6129,
+      "step": 34128
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4130029469285598e-06,
+      "loss": 0.6217,
+      "step": 34129
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4125385986321015e-06,
+      "loss": 0.6132,
+      "step": 34130
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4120742888897607e-06,
+      "loss": 0.6102,
+      "step": 34131
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.411610017703889e-06,
+      "loss": 0.6023,
+      "step": 34132
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.411145785076856e-06,
+      "loss": 0.5484,
+      "step": 34133
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.410681591011007e-06,
+      "loss": 0.5947,
+      "step": 34134
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4102174355087115e-06,
+      "loss": 0.5986,
+      "step": 34135
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.40975331857232e-06,
+      "loss": 0.5543,
+      "step": 34136
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.409289240204198e-06,
+      "loss": 0.5617,
+      "step": 34137
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.408825200406697e-06,
+      "loss": 0.6238,
+      "step": 34138
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4083611991821854e-06,
+      "loss": 0.6289,
+      "step": 34139
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4078972365330065e-06,
+      "loss": 0.655,
+      "step": 34140
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.407433312461529e-06,
+      "loss": 0.5858,
+      "step": 34141
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.406969426970107e-06,
+      "loss": 0.6047,
+      "step": 34142
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4065055800610936e-06,
+      "loss": 0.6013,
+      "step": 34143
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.406041771736852e-06,
+      "loss": 0.6039,
+      "step": 34144
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4055780019997356e-06,
+      "loss": 0.5981,
+      "step": 34145
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4051142708521024e-06,
+      "loss": 0.5942,
+      "step": 34146
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.404650578296305e-06,
+      "loss": 0.6667,
+      "step": 34147
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.404186924334705e-06,
+      "loss": 0.5818,
+      "step": 34148
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.403723308969652e-06,
+      "loss": 0.551,
+      "step": 34149
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4032597322035123e-06,
+      "loss": 0.5715,
+      "step": 34150
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.402796194038629e-06,
+      "loss": 0.6029,
+      "step": 34151
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4023326944773662e-06,
+      "loss": 0.608,
+      "step": 34152
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.401869233522073e-06,
+      "loss": 0.578,
+      "step": 34153
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4014058111751092e-06,
+      "loss": 0.6052,
+      "step": 34154
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.400942427438829e-06,
+      "loss": 0.6183,
+      "step": 34155
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4004790823155853e-06,
+      "loss": 0.5566,
+      "step": 34156
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.4000157758077315e-06,
+      "loss": 0.6417,
+      "step": 34157
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.399552507917624e-06,
+      "loss": 0.5613,
+      "step": 34158
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.399089278647617e-06,
+      "loss": 0.6138,
+      "step": 34159
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3986260880000644e-06,
+      "loss": 0.6033,
+      "step": 34160
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.398162935977315e-06,
+      "loss": 0.628,
+      "step": 34161
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3976998225817294e-06,
+      "loss": 0.6245,
+      "step": 34162
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.397236747815657e-06,
+      "loss": 0.5866,
+      "step": 34163
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3967737116814483e-06,
+      "loss": 0.613,
+      "step": 34164
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3963107141814623e-06,
+      "loss": 0.6073,
+      "step": 34165
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3958477553180483e-06,
+      "loss": 0.5951,
+      "step": 34166
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3953848350935583e-06,
+      "loss": 0.5915,
+      "step": 34167
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.394921953510343e-06,
+      "loss": 0.6187,
+      "step": 34168
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.394459110570758e-06,
+      "loss": 0.6224,
+      "step": 34169
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.393996306277152e-06,
+      "loss": 0.6082,
+      "step": 34170
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3935335406318838e-06,
+      "loss": 0.5502,
+      "step": 34171
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.393070813637294e-06,
+      "loss": 0.5199,
+      "step": 34172
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3926081252957423e-06,
+      "loss": 0.6474,
+      "step": 34173
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.392145475609573e-06,
+      "loss": 0.5745,
+      "step": 34174
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.391682864581143e-06,
+      "loss": 0.6257,
+      "step": 34175
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.391220292212798e-06,
+      "loss": 0.5522,
+      "step": 34176
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3907577585068975e-06,
+      "loss": 0.584,
+      "step": 34177
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3902952634657793e-06,
+      "loss": 0.6285,
+      "step": 34178
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.389832807091802e-06,
+      "loss": 0.6712,
+      "step": 34179
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3893703893873134e-06,
+      "loss": 0.6102,
+      "step": 34180
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.388908010354659e-06,
+      "loss": 0.5704,
+      "step": 34181
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3884456699961957e-06,
+      "loss": 0.6563,
+      "step": 34182
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3879833683142696e-06,
+      "loss": 0.608,
+      "step": 34183
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.387521105311228e-06,
+      "loss": 0.6213,
+      "step": 34184
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.387058880989419e-06,
+      "loss": 0.5934,
+      "step": 34185
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3865966953511952e-06,
+      "loss": 0.5719,
+      "step": 34186
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.386134548398905e-06,
+      "loss": 0.5839,
+      "step": 34187
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.385672440134894e-06,
+      "loss": 0.5735,
+      "step": 34188
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.385210370561508e-06,
+      "loss": 0.6409,
+      "step": 34189
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.384748339681102e-06,
+      "loss": 0.6282,
+      "step": 34190
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3842863474960165e-06,
+      "loss": 0.6062,
+      "step": 34191
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.383824394008606e-06,
+      "loss": 0.6865,
+      "step": 34192
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3833624792212127e-06,
+      "loss": 0.5817,
+      "step": 34193
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3829006031361866e-06,
+      "loss": 0.5962,
+      "step": 34194
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.38243876575587e-06,
+      "loss": 0.5766,
+      "step": 34195
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3819769670826164e-06,
+      "loss": 0.5797,
+      "step": 34196
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3815152071187652e-06,
+      "loss": 0.5411,
+      "step": 34197
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3810534858666724e-06,
+      "loss": 0.6079,
+      "step": 34198
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3805918033286722e-06,
+      "loss": 0.5754,
+      "step": 34199
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3801301595071204e-06,
+      "loss": 0.5968,
+      "step": 34200
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3796685544043574e-06,
+      "loss": 0.6025,
+      "step": 34201
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.379206988022729e-06,
+      "loss": 0.5497,
+      "step": 34202
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3787454603645832e-06,
+      "loss": 0.6184,
+      "step": 34203
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3782839714322635e-06,
+      "loss": 0.6002,
+      "step": 34204
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.377822521228116e-06,
+      "loss": 0.6473,
+      "step": 34205
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.377361109754481e-06,
+      "loss": 0.5682,
+      "step": 34206
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3768997370137103e-06,
+      "loss": 0.6585,
+      "step": 34207
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.376438403008141e-06,
+      "loss": 0.5472,
+      "step": 34208
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3759771077401262e-06,
+      "loss": 0.5917,
+      "step": 34209
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.375515851212e-06,
+      "loss": 0.613,
+      "step": 34210
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3750546334261126e-06,
+      "loss": 0.6006,
+      "step": 34211
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3745934543848035e-06,
+      "loss": 0.7177,
+      "step": 34212
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3741323140904203e-06,
+      "loss": 0.6253,
+      "step": 34213
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.373671212545304e-06,
+      "loss": 0.6079,
+      "step": 34214
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3732101497517988e-06,
+      "loss": 0.6394,
+      "step": 34215
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3727491257122437e-06,
+      "loss": 0.5781,
+      "step": 34216
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3722881404289857e-06,
+      "loss": 0.5646,
+      "step": 34217
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3718271939043637e-06,
+      "loss": 0.5554,
+      "step": 34218
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3713662861407237e-06,
+      "loss": 0.634,
+      "step": 34219
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.370905417140407e-06,
+      "loss": 0.6213,
+      "step": 34220
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3704445869057535e-06,
+      "loss": 0.6428,
+      "step": 34221
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.369983795439105e-06,
+      "loss": 0.5774,
+      "step": 34222
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3695230427428008e-06,
+      "loss": 0.6271,
+      "step": 34223
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.369062328819187e-06,
+      "loss": 0.6253,
+      "step": 34224
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3686016536706036e-06,
+      "loss": 0.5943,
+      "step": 34225
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.36814101729939e-06,
+      "loss": 0.5553,
+      "step": 34226
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.367680419707883e-06,
+      "loss": 0.6238,
+      "step": 34227
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.367219860898431e-06,
+      "loss": 0.6218,
+      "step": 34228
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.366759340873367e-06,
+      "loss": 0.5832,
+      "step": 34229
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3662988596350366e-06,
+      "loss": 0.5806,
+      "step": 34230
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3658384171857782e-06,
+      "loss": 0.6151,
+      "step": 34231
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3653780135279302e-06,
+      "loss": 0.563,
+      "step": 34232
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.364917648663829e-06,
+      "loss": 0.6222,
+      "step": 34233
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.364457322595821e-06,
+      "loss": 0.622,
+      "step": 34234
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3639970353262386e-06,
+      "loss": 0.5857,
+      "step": 34235
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3635367868574287e-06,
+      "loss": 0.5666,
+      "step": 34236
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3630765771917195e-06,
+      "loss": 0.5891,
+      "step": 34237
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.362616406331457e-06,
+      "loss": 0.5931,
+      "step": 34238
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3621562742789762e-06,
+      "loss": 0.6242,
+      "step": 34239
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3616961810366166e-06,
+      "loss": 0.6153,
+      "step": 34240
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3612361266067173e-06,
+      "loss": 0.5913,
+      "step": 34241
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3607761109916137e-06,
+      "loss": 0.5956,
+      "step": 34242
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3603161341936452e-06,
+      "loss": 0.5959,
+      "step": 34243
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3598561962151434e-06,
+      "loss": 0.5787,
+      "step": 34244
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.359396297058454e-06,
+      "loss": 0.5462,
+      "step": 34245
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.358936436725906e-06,
+      "loss": 0.5811,
+      "step": 34246
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.358476615219845e-06,
+      "loss": 0.566,
+      "step": 34247
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.358016832542597e-06,
+      "loss": 0.5791,
+      "step": 34248
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.357557088696506e-06,
+      "loss": 0.644,
+      "step": 34249
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.357097383683904e-06,
+      "loss": 0.6456,
+      "step": 34250
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3566377175071296e-06,
+      "loss": 0.6077,
+      "step": 34251
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3561780901685192e-06,
+      "loss": 0.5599,
+      "step": 34252
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3557185016704052e-06,
+      "loss": 0.583,
+      "step": 34253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3552589520151224e-06,
+      "loss": 0.5955,
+      "step": 34254
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3547994412050102e-06,
+      "loss": 0.6129,
+      "step": 34255
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.354339969242397e-06,
+      "loss": 0.6045,
+      "step": 34256
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.353880536129626e-06,
+      "loss": 0.6152,
+      "step": 34257
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3534211418690267e-06,
+      "loss": 0.5886,
+      "step": 34258
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.352961786462934e-06,
+      "loss": 0.5517,
+      "step": 34259
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3525024699136825e-06,
+      "loss": 0.5418,
+      "step": 34260
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.352043192223602e-06,
+      "loss": 0.6324,
+      "step": 34261
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3515839533950322e-06,
+      "loss": 0.559,
+      "step": 34262
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.351124753430305e-06,
+      "loss": 0.5822,
+      "step": 34263
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3506655923317535e-06,
+      "loss": 0.5573,
+      "step": 34264
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.350206470101707e-06,
+      "loss": 0.7006,
+      "step": 34265
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3497473867425047e-06,
+      "loss": 0.5932,
+      "step": 34266
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.349288342256474e-06,
+      "loss": 0.6188,
+      "step": 34267
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3488293366459515e-06,
+      "loss": 0.652,
+      "step": 34268
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3483703699132686e-06,
+      "loss": 0.5288,
+      "step": 34269
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3479114420607575e-06,
+      "loss": 0.6262,
+      "step": 34270
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3474525530907453e-06,
+      "loss": 0.5308,
+      "step": 34271
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3469937030055712e-06,
+      "loss": 0.5621,
+      "step": 34272
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3465348918075603e-06,
+      "loss": 0.6066,
+      "step": 34273
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3460761194990533e-06,
+      "loss": 0.5856,
+      "step": 34274
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3456173860823682e-06,
+      "loss": 0.5951,
+      "step": 34275
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3451586915598466e-06,
+      "loss": 0.5882,
+      "step": 34276
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3447000359338136e-06,
+      "loss": 0.6195,
+      "step": 34277
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.344241419206603e-06,
+      "loss": 0.5786,
+      "step": 34278
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.343782841380544e-06,
+      "loss": 0.573,
+      "step": 34279
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3433243024579678e-06,
+      "loss": 0.5814,
+      "step": 34280
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.342865802441202e-06,
+      "loss": 0.6104,
+      "step": 34281
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3424073413325744e-06,
+      "loss": 0.6538,
+      "step": 34282
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3419489191344212e-06,
+      "loss": 0.6357,
+      "step": 34283
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.341490535849066e-06,
+      "loss": 0.5754,
+      "step": 34284
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3410321914788448e-06,
+      "loss": 0.5716,
+      "step": 34285
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.340573886026077e-06,
+      "loss": 0.6097,
+      "step": 34286
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3401156194930997e-06,
+      "loss": 0.5726,
+      "step": 34287
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.339657391882235e-06,
+      "loss": 0.598,
+      "step": 34288
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3391992031958177e-06,
+      "loss": 0.622,
+      "step": 34289
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.338741053436172e-06,
+      "loss": 0.5981,
+      "step": 34290
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3382829426056276e-06,
+      "loss": 0.5803,
+      "step": 34291
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3378248707065077e-06,
+      "loss": 0.6459,
+      "step": 34292
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.337366837741147e-06,
+      "loss": 0.588,
+      "step": 34293
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.336908843711867e-06,
+      "loss": 0.6339,
+      "step": 34294
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3364508886210002e-06,
+      "loss": 0.6475,
+      "step": 34295
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3359929724708695e-06,
+      "loss": 0.672,
+      "step": 34296
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3355350952638045e-06,
+      "loss": 0.5966,
+      "step": 34297
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.335077257002127e-06,
+      "loss": 0.6014,
+      "step": 34298
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3346194576881696e-06,
+      "loss": 0.6484,
+      "step": 34299
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3341616973242555e-06,
+      "loss": 0.5649,
+      "step": 34300
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.333703975912711e-06,
+      "loss": 0.6191,
+      "step": 34301
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3332462934558618e-06,
+      "loss": 0.6032,
+      "step": 34302
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3327886499560306e-06,
+      "loss": 0.5869,
+      "step": 34303
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.332331045415548e-06,
+      "loss": 0.5843,
+      "step": 34304
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.331873479836735e-06,
+      "loss": 0.5736,
+      "step": 34305
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.331415953221924e-06,
+      "loss": 0.5847,
+      "step": 34306
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3309584655734284e-06,
+      "loss": 0.6178,
+      "step": 34307
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3305010168935815e-06,
+      "loss": 0.6523,
+      "step": 34308
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.330043607184703e-06,
+      "loss": 0.5473,
+      "step": 34309
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3295862364491216e-06,
+      "loss": 0.6722,
+      "step": 34310
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.329128904689155e-06,
+      "loss": 0.6311,
+      "step": 34311
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3286716119071383e-06,
+      "loss": 0.5732,
+      "step": 34312
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3282143581053817e-06,
+      "loss": 0.58,
+      "step": 34313
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3277571432862167e-06,
+      "loss": 0.5641,
+      "step": 34314
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3272999674519625e-06,
+      "loss": 0.6335,
+      "step": 34315
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.326842830604947e-06,
+      "loss": 0.6403,
+      "step": 34316
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3263857327474903e-06,
+      "loss": 0.6411,
+      "step": 34317
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3259286738819152e-06,
+      "loss": 0.6255,
+      "step": 34318
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3254716540105404e-06,
+      "loss": 0.6053,
+      "step": 34319
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.325014673135696e-06,
+      "loss": 0.5907,
+      "step": 34320
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.324557731259698e-06,
+      "loss": 0.6169,
+      "step": 34321
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.324100828384869e-06,
+      "loss": 0.5135,
+      "step": 34322
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3236439645135356e-06,
+      "loss": 0.5706,
+      "step": 34323
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.32318713964801e-06,
+      "loss": 0.6172,
+      "step": 34324
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.322730353790622e-06,
+      "loss": 0.6381,
+      "step": 34325
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.322273606943687e-06,
+      "loss": 0.621,
+      "step": 34326
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3218168991095315e-06,
+      "loss": 0.6525,
+      "step": 34327
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3213602302904716e-06,
+      "loss": 0.5967,
+      "step": 34328
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3209036004888295e-06,
+      "loss": 0.5912,
+      "step": 34329
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.320447009706923e-06,
+      "loss": 0.5775,
+      "step": 34330
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3199904579470767e-06,
+      "loss": 0.6527,
+      "step": 34331
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.319533945211606e-06,
+      "loss": 0.6204,
+      "step": 34332
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.319077471502834e-06,
+      "loss": 0.6031,
+      "step": 34333
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.31862103682308e-06,
+      "loss": 0.575,
+      "step": 34334
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3181646411746616e-06,
+      "loss": 0.573,
+      "step": 34335
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3177082845598962e-06,
+      "loss": 0.6543,
+      "step": 34336
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3172519669811067e-06,
+      "loss": 0.6448,
+      "step": 34337
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3167956884406105e-06,
+      "loss": 0.6597,
+      "step": 34338
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3163394489407263e-06,
+      "loss": 0.5792,
+      "step": 34339
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.315883248483768e-06,
+      "loss": 0.6265,
+      "step": 34340
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3154270870720596e-06,
+      "loss": 0.698,
+      "step": 34341
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3149709647079175e-06,
+      "loss": 0.5772,
+      "step": 34342
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3145148813936557e-06,
+      "loss": 0.6015,
+      "step": 34343
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3140588371315998e-06,
+      "loss": 0.5454,
+      "step": 34344
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3136028319240566e-06,
+      "loss": 0.6453,
+      "step": 34345
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3131468657733512e-06,
+      "loss": 0.628,
+      "step": 34346
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3126909386817953e-06,
+      "loss": 0.5428,
+      "step": 34347
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.31223505065171e-06,
+      "loss": 0.6067,
+      "step": 34348
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.311779201685408e-06,
+      "loss": 0.5747,
+      "step": 34349
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3113233917852127e-06,
+      "loss": 0.6132,
+      "step": 34350
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3108676209534296e-06,
+      "loss": 0.6505,
+      "step": 34351
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.310411889192382e-06,
+      "loss": 0.6361,
+      "step": 34352
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3099561965043817e-06,
+      "loss": 0.6432,
+      "step": 34353
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3095005428917484e-06,
+      "loss": 0.6432,
+      "step": 34354
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.309044928356795e-06,
+      "loss": 0.6191,
+      "step": 34355
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.308589352901838e-06,
+      "loss": 0.61,
+      "step": 34356
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3081338165291866e-06,
+      "loss": 0.6113,
+      "step": 34357
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.307678319241163e-06,
+      "loss": 0.6299,
+      "step": 34358
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3072228610400794e-06,
+      "loss": 0.6164,
+      "step": 34359
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.306767441928246e-06,
+      "loss": 0.6147,
+      "step": 34360
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.306312061907986e-06,
+      "loss": 0.55,
+      "step": 34361
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.305856720981603e-06,
+      "loss": 0.6588,
+      "step": 34362
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3054014191514175e-06,
+      "loss": 0.6002,
+      "step": 34363
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3049461564197373e-06,
+      "loss": 0.5813,
+      "step": 34364
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.304490932788882e-06,
+      "loss": 0.5903,
+      "step": 34365
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.304035748261163e-06,
+      "loss": 0.6705,
+      "step": 34366
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3035806028388917e-06,
+      "loss": 0.6274,
+      "step": 34367
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.303125496524379e-06,
+      "loss": 0.6417,
+      "step": 34368
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.302670429319942e-06,
+      "loss": 0.5819,
+      "step": 34369
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3022154012278887e-06,
+      "loss": 0.5643,
+      "step": 34370
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.301760412250539e-06,
+      "loss": 0.632,
+      "step": 34371
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.301305462390192e-06,
+      "loss": 0.6039,
+      "step": 34372
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300850551649171e-06,
+      "loss": 0.5723,
+      "step": 34373
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.30039568002978e-06,
+      "loss": 0.671,
+      "step": 34374
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2999408475343353e-06,
+      "loss": 0.6081,
+      "step": 34375
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2994860541651475e-06,
+      "loss": 0.669,
+      "step": 34376
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2990312999245255e-06,
+      "loss": 0.6535,
+      "step": 34377
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298576584814779e-06,
+      "loss": 0.5975,
+      "step": 34378
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2981219088382225e-06,
+      "loss": 0.6118,
+      "step": 34379
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2976672719971636e-06,
+      "loss": 0.6385,
+      "step": 34380
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2972126742939118e-06,
+      "loss": 0.571,
+      "step": 34381
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2967581157307827e-06,
+      "loss": 0.6107,
+      "step": 34382
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2963035963100753e-06,
+      "loss": 0.637,
+      "step": 34383
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.295849116034109e-06,
+      "loss": 0.5559,
+      "step": 34384
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.295394674905187e-06,
+      "loss": 0.5419,
+      "step": 34385
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2949402729256243e-06,
+      "loss": 0.64,
+      "step": 34386
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2944859100977236e-06,
+      "loss": 0.6577,
+      "step": 34387
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2940315864238015e-06,
+      "loss": 0.6376,
+      "step": 34388
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2935773019061568e-06,
+      "loss": 0.5848,
+      "step": 34389
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2931230565471053e-06,
+      "loss": 0.6039,
+      "step": 34390
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.292668850348949e-06,
+      "loss": 0.5883,
+      "step": 34391
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2922146833140034e-06,
+      "loss": 0.637,
+      "step": 34392
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.291760555444572e-06,
+      "loss": 0.5872,
+      "step": 34393
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.291306466742963e-06,
+      "loss": 0.6767,
+      "step": 34394
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.29085241721148e-06,
+      "loss": 0.5482,
+      "step": 34395
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2903984068524367e-06,
+      "loss": 0.6455,
+      "step": 34396
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2899444356681377e-06,
+      "loss": 0.5859,
+      "step": 34397
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2894905036608884e-06,
+      "loss": 0.5721,
+      "step": 34398
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.289036610832993e-06,
+      "loss": 0.5214,
+      "step": 34399
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2885827571867636e-06,
+      "loss": 0.6055,
+      "step": 34400
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2881289427245036e-06,
+      "loss": 0.5756,
+      "step": 34401
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2876751674485166e-06,
+      "loss": 0.5789,
+      "step": 34402
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.287221431361113e-06,
+      "loss": 0.6023,
+      "step": 34403
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2867677344645954e-06,
+      "loss": 0.5418,
+      "step": 34404
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.286314076761271e-06,
+      "loss": 0.6218,
+      "step": 34405
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2858604582534415e-06,
+      "loss": 0.644,
+      "step": 34406
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.285406878943416e-06,
+      "loss": 0.5716,
+      "step": 34407
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2849533388334953e-06,
+      "loss": 0.622,
+      "step": 34408
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.284499837925991e-06,
+      "loss": 0.5452,
+      "step": 34409
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.284046376223198e-06,
+      "loss": 0.6555,
+      "step": 34410
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.283592953727427e-06,
+      "loss": 0.5906,
+      "step": 34411
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2831395704409776e-06,
+      "loss": 0.6334,
+      "step": 34412
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2826862263661588e-06,
+      "loss": 0.6202,
+      "step": 34413
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.282232921505272e-06,
+      "loss": 0.634,
+      "step": 34414
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.281779655860619e-06,
+      "loss": 0.6795,
+      "step": 34415
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2813264294345018e-06,
+      "loss": 0.5881,
+      "step": 34416
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.280873242229228e-06,
+      "loss": 0.6479,
+      "step": 34417
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2804200942470978e-06,
+      "loss": 0.6175,
+      "step": 34418
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2799669854904115e-06,
+      "loss": 0.5721,
+      "step": 34419
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2795139159614765e-06,
+      "loss": 0.6176,
+      "step": 34420
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2790608856625917e-06,
+      "loss": 0.6238,
+      "step": 34421
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.278607894596061e-06,
+      "loss": 0.5932,
+      "step": 34422
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2781549427641816e-06,
+      "loss": 0.6283,
+      "step": 34423
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.277702030169261e-06,
+      "loss": 0.6065,
+      "step": 34424
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2772491568135946e-06,
+      "loss": 0.6577,
+      "step": 34425
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.276796322699493e-06,
+      "loss": 0.5795,
+      "step": 34426
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2763435278292454e-06,
+      "loss": 0.5586,
+      "step": 34427
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2758907722051616e-06,
+      "loss": 0.5909,
+      "step": 34428
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2754380558295366e-06,
+      "loss": 0.6107,
+      "step": 34429
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.274985378704675e-06,
+      "loss": 0.5868,
+      "step": 34430
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2745327408328745e-06,
+      "loss": 0.5873,
+      "step": 34431
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.274080142216436e-06,
+      "loss": 0.6021,
+      "step": 34432
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.273627582857657e-06,
+      "loss": 0.6349,
+      "step": 34433
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.273175062758841e-06,
+      "loss": 0.6,
+      "step": 34434
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2727225819222862e-06,
+      "loss": 0.6066,
+      "step": 34435
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.27227014035029e-06,
+      "loss": 0.6468,
+      "step": 34436
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.27181773804515e-06,
+      "loss": 0.6177,
+      "step": 34437
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2713653750091702e-06,
+      "loss": 0.5815,
+      "step": 34438
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2709130512446466e-06,
+      "loss": 0.5973,
+      "step": 34439
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.270460766753875e-06,
+      "loss": 0.6016,
+      "step": 34440
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.270008521539158e-06,
+      "loss": 0.661,
+      "step": 34441
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2695563156027913e-06,
+      "loss": 0.5586,
+      "step": 34442
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2691041489470733e-06,
+      "loss": 0.6184,
+      "step": 34443
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2686520215742992e-06,
+      "loss": 0.6161,
+      "step": 34444
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2681999334867697e-06,
+      "loss": 0.5821,
+      "step": 34445
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2677478846867785e-06,
+      "loss": 0.633,
+      "step": 34446
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2672958751766303e-06,
+      "loss": 0.5801,
+      "step": 34447
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.266843904958611e-06,
+      "loss": 0.5811,
+      "step": 34448
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.266391974035026e-06,
+      "loss": 0.5493,
+      "step": 34449
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2659400824081645e-06,
+      "loss": 0.6254,
+      "step": 34450
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2654882300803305e-06,
+      "loss": 0.5833,
+      "step": 34451
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2650364170538143e-06,
+      "loss": 0.6198,
+      "step": 34452
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2645846433309148e-06,
+      "loss": 0.6175,
+      "step": 34453
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.264132908913922e-06,
+      "loss": 0.6164,
+      "step": 34454
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.263681213805139e-06,
+      "loss": 0.6291,
+      "step": 34455
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2632295580068576e-06,
+      "loss": 0.5872,
+      "step": 34456
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.26277794152137e-06,
+      "loss": 0.6394,
+      "step": 34457
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.262326364350975e-06,
+      "loss": 0.6628,
+      "step": 34458
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.261874826497967e-06,
+      "loss": 0.5786,
+      "step": 34459
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2614233279646393e-06,
+      "loss": 0.6634,
+      "step": 34460
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2609718687532834e-06,
+      "loss": 0.5626,
+      "step": 34461
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.260520448866198e-06,
+      "loss": 0.6119,
+      "step": 34462
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2600690683056758e-06,
+      "loss": 0.5935,
+      "step": 34463
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.259617727074008e-06,
+      "loss": 0.657,
+      "step": 34464
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.259166425173488e-06,
+      "loss": 0.5887,
+      "step": 34465
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.258715162606412e-06,
+      "loss": 0.5566,
+      "step": 34466
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2582639393750694e-06,
+      "loss": 0.5461,
+      "step": 34467
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2578127554817576e-06,
+      "loss": 0.6023,
+      "step": 34468
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2573616109287666e-06,
+      "loss": 0.5922,
+      "step": 34469
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2569105057183884e-06,
+      "loss": 0.5854,
+      "step": 34470
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2564594398529137e-06,
+      "loss": 0.5811,
+      "step": 34471
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.256008413334638e-06,
+      "loss": 0.5798,
+      "step": 34472
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.255557426165852e-06,
+      "loss": 0.5848,
+      "step": 34473
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2551064783488474e-06,
+      "loss": 0.5906,
+      "step": 34474
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.254655569885912e-06,
+      "loss": 0.645,
+      "step": 34475
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2542047007793424e-06,
+      "loss": 0.5318,
+      "step": 34476
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2537538710314266e-06,
+      "loss": 0.5827,
+      "step": 34477
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2533030806444534e-06,
+      "loss": 0.6358,
+      "step": 34478
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2528523296207195e-06,
+      "loss": 0.6264,
+      "step": 34479
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.252401617962512e-06,
+      "loss": 0.635,
+      "step": 34480
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.25195094567212e-06,
+      "loss": 0.556,
+      "step": 34481
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.251500312751832e-06,
+      "loss": 0.6109,
+      "step": 34482
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.251049719203943e-06,
+      "loss": 0.6011,
+      "step": 34483
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.250599165030736e-06,
+      "loss": 0.574,
+      "step": 34484
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2501486502345103e-06,
+      "loss": 0.5975,
+      "step": 34485
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.249698174817544e-06,
+      "loss": 0.613,
+      "step": 34486
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.249247738782133e-06,
+      "loss": 0.5373,
+      "step": 34487
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.248797342130562e-06,
+      "loss": 0.5842,
+      "step": 34488
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.248346984865125e-06,
+      "loss": 0.6093,
+      "step": 34489
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2478966669881064e-06,
+      "loss": 0.5922,
+      "step": 34490
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.247446388501795e-06,
+      "loss": 0.6451,
+      "step": 34491
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2469961494084767e-06,
+      "loss": 0.6351,
+      "step": 34492
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2465459497104435e-06,
+      "loss": 0.598,
+      "step": 34493
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.246095789409982e-06,
+      "loss": 0.6098,
+      "step": 34494
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2456456685093753e-06,
+      "loss": 0.6313,
+      "step": 34495
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2451955870109164e-06,
+      "loss": 0.5797,
+      "step": 34496
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.244745544916891e-06,
+      "loss": 0.7052,
+      "step": 34497
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.244295542229583e-06,
+      "loss": 0.6105,
+      "step": 34498
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2438455789512794e-06,
+      "loss": 0.5948,
+      "step": 34499
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2433956550842705e-06,
+      "loss": 0.6557,
+      "step": 34500
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2429457706308388e-06,
+      "loss": 0.6531,
+      "step": 34501
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2424959255932734e-06,
+      "loss": 0.5959,
+      "step": 34502
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.242046119973853e-06,
+      "loss": 0.6116,
+      "step": 34503
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.241596353774873e-06,
+      "loss": 0.6311,
+      "step": 34504
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2411466269986103e-06,
+      "loss": 0.626,
+      "step": 34505
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2406969396473564e-06,
+      "loss": 0.5901,
+      "step": 34506
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.240247291723394e-06,
+      "loss": 0.6203,
+      "step": 34507
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.239797683229008e-06,
+      "loss": 0.5723,
+      "step": 34508
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.23934811416648e-06,
+      "loss": 0.5681,
+      "step": 34509
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2388985845380995e-06,
+      "loss": 0.5819,
+      "step": 34510
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2384490943461456e-06,
+      "loss": 0.5794,
+      "step": 34511
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2379996435929096e-06,
+      "loss": 0.5663,
+      "step": 34512
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2375502322806663e-06,
+      "loss": 0.5912,
+      "step": 34513
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.237100860411706e-06,
+      "loss": 0.5633,
+      "step": 34514
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2366515279883093e-06,
+      "loss": 0.6189,
+      "step": 34515
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2362022350127576e-06,
+      "loss": 0.5819,
+      "step": 34516
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.235752981487339e-06,
+      "loss": 0.5473,
+      "step": 34517
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.235303767414333e-06,
+      "loss": 0.6061,
+      "step": 34518
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.234854592796022e-06,
+      "loss": 0.602,
+      "step": 34519
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2344054576346864e-06,
+      "loss": 0.5784,
+      "step": 34520
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2339563619326143e-06,
+      "loss": 0.6384,
+      "step": 34521
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.233507305692081e-06,
+      "loss": 0.6135,
+      "step": 34522
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2330582889153772e-06,
+      "loss": 0.5602,
+      "step": 34523
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2326093116047743e-06,
+      "loss": 0.5935,
+      "step": 34524
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2321603737625598e-06,
+      "loss": 0.5741,
+      "step": 34525
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2317114753910107e-06,
+      "loss": 0.5767,
+      "step": 34526
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2312626164924146e-06,
+      "loss": 0.6556,
+      "step": 34527
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.230813797069047e-06,
+      "loss": 0.6469,
+      "step": 34528
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2303650171231907e-06,
+      "loss": 0.5908,
+      "step": 34529
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2299162766571216e-06,
+      "loss": 0.5729,
+      "step": 34530
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2294675756731267e-06,
+      "loss": 0.6159,
+      "step": 34531
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2290189141734795e-06,
+      "loss": 0.5982,
+      "step": 34532
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.228570292160467e-06,
+      "loss": 0.5929,
+      "step": 34533
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2281217096363638e-06,
+      "loss": 0.583,
+      "step": 34534
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.22767316660345e-06,
+      "loss": 0.654,
+      "step": 34535
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.227224663064006e-06,
+      "loss": 0.6036,
+      "step": 34536
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2267761990203074e-06,
+      "loss": 0.564,
+      "step": 34537
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.226327774474637e-06,
+      "loss": 0.6312,
+      "step": 34538
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.225879389429273e-06,
+      "loss": 0.569,
+      "step": 34539
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.225431043886491e-06,
+      "loss": 0.5376,
+      "step": 34540
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.224982737848569e-06,
+      "loss": 0.6406,
+      "step": 34541
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2245344713177897e-06,
+      "loss": 0.642,
+      "step": 34542
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.224086244296425e-06,
+      "loss": 0.5343,
+      "step": 34543
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.223638056786758e-06,
+      "loss": 0.6248,
+      "step": 34544
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2231899087910623e-06,
+      "loss": 0.5839,
+      "step": 34545
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2227418003116174e-06,
+      "loss": 0.5804,
+      "step": 34546
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2222937313506963e-06,
+      "loss": 0.588,
+      "step": 34547
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.221845701910582e-06,
+      "loss": 0.6181,
+      "step": 34548
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2213977119935437e-06,
+      "loss": 0.568,
+      "step": 34549
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2209497616018683e-06,
+      "loss": 0.6735,
+      "step": 34550
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2205018507378185e-06,
+      "loss": 0.6378,
+      "step": 34551
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2200539794036813e-06,
+      "loss": 0.6119,
+      "step": 34552
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2196061476017274e-06,
+      "loss": 0.6189,
+      "step": 34553
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.21915835533423e-06,
+      "loss": 0.647,
+      "step": 34554
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.218710602603471e-06,
+      "loss": 0.6348,
+      "step": 34555
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.218262889411723e-06,
+      "loss": 0.5936,
+      "step": 34556
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.21781521576126e-06,
+      "loss": 0.5505,
+      "step": 34557
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.217367581654354e-06,
+      "loss": 0.6263,
+      "step": 34558
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2169199870932847e-06,
+      "loss": 0.5593,
+      "step": 34559
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2164724320803223e-06,
+      "loss": 0.6203,
+      "step": 34560
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.216024916617748e-06,
+      "loss": 0.641,
+      "step": 34561
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2155774407078255e-06,
+      "loss": 0.5614,
+      "step": 34562
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.215130004352837e-06,
+      "loss": 0.5745,
+      "step": 34563
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.21468260755505e-06,
+      "loss": 0.6035,
+      "step": 34564
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.214235250316743e-06,
+      "loss": 0.6353,
+      "step": 34565
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.213787932640187e-06,
+      "loss": 0.6059,
+      "step": 34566
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.213340654527655e-06,
+      "loss": 0.6337,
+      "step": 34567
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2128934159814164e-06,
+      "loss": 0.5139,
+      "step": 34568
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.21244621700375e-06,
+      "loss": 0.6318,
+      "step": 34569
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.211999057596922e-06,
+      "loss": 0.5864,
+      "step": 34570
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.21155193776321e-06,
+      "loss": 0.6159,
+      "step": 34571
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.211104857504884e-06,
+      "loss": 0.6266,
+      "step": 34572
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2106578168242153e-06,
+      "loss": 0.5586,
+      "step": 34573
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.210210815723476e-06,
+      "loss": 0.6467,
+      "step": 34574
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2097638542049337e-06,
+      "loss": 0.5997,
+      "step": 34575
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2093169322708652e-06,
+      "loss": 0.5742,
+      "step": 34576
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2088700499235393e-06,
+      "loss": 0.6279,
+      "step": 34577
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.208423207165227e-06,
+      "loss": 0.5982,
+      "step": 34578
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2079764039981944e-06,
+      "loss": 0.6097,
+      "step": 34579
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2075296404247192e-06,
+      "loss": 0.6386,
+      "step": 34580
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.207082916447065e-06,
+      "loss": 0.6293,
+      "step": 34581
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2066362320675105e-06,
+      "loss": 0.6065,
+      "step": 34582
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2061895872883145e-06,
+      "loss": 0.5816,
+      "step": 34583
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.205742982111754e-06,
+      "loss": 0.559,
+      "step": 34584
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2052964165400924e-06,
+      "loss": 0.6275,
+      "step": 34585
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2048498905756067e-06,
+      "loss": 0.6283,
+      "step": 34586
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2044034042205577e-06,
+      "loss": 0.5796,
+      "step": 34587
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2039569574772247e-06,
+      "loss": 0.6186,
+      "step": 34588
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2035105503478626e-06,
+      "loss": 0.6352,
+      "step": 34589
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.203064182834751e-06,
+      "loss": 0.5826,
+      "step": 34590
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.20261785494015e-06,
+      "loss": 0.6552,
+      "step": 34591
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2021715666663344e-06,
+      "loss": 0.6106,
+      "step": 34592
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2017253180155686e-06,
+      "loss": 0.5389,
+      "step": 34593
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.201279108990121e-06,
+      "loss": 0.605,
+      "step": 34594
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.200832939592258e-06,
+      "loss": 0.5836,
+      "step": 34595
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.200386809824243e-06,
+      "loss": 0.6265,
+      "step": 34596
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.1999407196883505e-06,
+      "loss": 0.6023,
+      "step": 34597
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.19949466918684e-06,
+      "loss": 0.581,
+      "step": 34598
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.1990486583219883e-06,
+      "loss": 0.5932,
+      "step": 34599
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.1986026870960486e-06,
+      "loss": 0.6253,
+      "step": 34600
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.198156755511297e-06,
+      "loss": 0.6432,
+      "step": 34601
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.197710863569993e-06,
+      "loss": 0.6023,
+      "step": 34602
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.1972650112744064e-06,
+      "loss": 0.6447,
+      "step": 34603
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.196819198626803e-06,
+      "loss": 0.5907,
+      "step": 34604
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1963734256294458e-06,
+      "loss": 0.6016,
+      "step": 34605
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1959276922845972e-06,
+      "loss": 0.6414,
+      "step": 34606
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1954819985945296e-06,
+      "loss": 0.5071,
+      "step": 34607
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.195036344561501e-06,
+      "loss": 0.6094,
+      "step": 34608
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.19459073018778e-06,
+      "loss": 0.5864,
+      "step": 34609
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.19414515547563e-06,
+      "loss": 0.6697,
+      "step": 34610
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1936996204273143e-06,
+      "loss": 0.5862,
+      "step": 34611
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1932541250450945e-06,
+      "loss": 0.6022,
+      "step": 34612
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.19280866933124e-06,
+      "loss": 0.6525,
+      "step": 34613
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1923632532880102e-06,
+      "loss": 0.5943,
+      "step": 34614
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1919178769176698e-06,
+      "loss": 0.6324,
+      "step": 34615
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1914725402224813e-06,
+      "loss": 0.5409,
+      "step": 34616
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.191027243204705e-06,
+      "loss": 0.628,
+      "step": 34617
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1905819858666087e-06,
+      "loss": 0.582,
+      "step": 34618
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1901367682104504e-06,
+      "loss": 0.5472,
+      "step": 34619
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1896915902384997e-06,
+      "loss": 0.5866,
+      "step": 34620
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1892464519530076e-06,
+      "loss": 0.6498,
+      "step": 34621
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1888013533562437e-06,
+      "loss": 0.5621,
+      "step": 34622
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1883562944504667e-06,
+      "loss": 0.5801,
+      "step": 34623
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1879112752379406e-06,
+      "loss": 0.619,
+      "step": 34624
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1874662957209225e-06,
+      "loss": 0.6032,
+      "step": 34625
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.187021355901682e-06,
+      "loss": 0.6172,
+      "step": 34626
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1865764557824685e-06,
+      "loss": 0.5648,
+      "step": 34627
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1861315953655515e-06,
+      "loss": 0.6307,
+      "step": 34628
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.185686774653185e-06,
+      "loss": 0.6484,
+      "step": 34629
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1852419936476355e-06,
+      "loss": 0.5481,
+      "step": 34630
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1847972523511595e-06,
+      "loss": 0.5983,
+      "step": 34631
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.184352550766018e-06,
+      "loss": 0.6166,
+      "step": 34632
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.18390788889447e-06,
+      "loss": 0.612,
+      "step": 34633
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1834632667387723e-06,
+      "loss": 0.5996,
+      "step": 34634
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1830186843011902e-06,
+      "loss": 0.6151,
+      "step": 34635
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1825741415839765e-06,
+      "loss": 0.6529,
+      "step": 34636
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1821296385893977e-06,
+      "loss": 0.6271,
+      "step": 34637
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1816851753197023e-06,
+      "loss": 0.5258,
+      "step": 34638
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1812407517771583e-06,
+      "loss": 0.5911,
+      "step": 34639
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1807963679640154e-06,
+      "loss": 0.5739,
+      "step": 34640
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1803520238825406e-06,
+      "loss": 0.5376,
+      "step": 34641
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.179907719534986e-06,
+      "loss": 0.6771,
+      "step": 34642
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1794634549236118e-06,
+      "loss": 0.6147,
+      "step": 34643
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.17901923005067e-06,
+      "loss": 0.6814,
+      "step": 34644
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.178575044918425e-06,
+      "loss": 0.5831,
+      "step": 34645
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1781308995291283e-06,
+      "loss": 0.6213,
+      "step": 34646
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1776867938850454e-06,
+      "loss": 0.5782,
+      "step": 34647
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.17724272798842e-06,
+      "loss": 0.5782,
+      "step": 34648
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1767987018415195e-06,
+      "loss": 0.5755,
+      "step": 34649
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1763547154465915e-06,
+      "loss": 0.5922,
+      "step": 34650
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1759107688059e-06,
+      "loss": 0.5622,
+      "step": 34651
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1754668619216978e-06,
+      "loss": 0.6199,
+      "step": 34652
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.17502299479624e-06,
+      "loss": 0.6149,
+      "step": 34653
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.174579167431781e-06,
+      "loss": 0.5825,
+      "step": 34654
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.174135379830574e-06,
+      "loss": 0.5301,
+      "step": 34655
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1736916319948807e-06,
+      "loss": 0.5327,
+      "step": 34656
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1732479239269498e-06,
+      "loss": 0.6105,
+      "step": 34657
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.172804255629044e-06,
+      "loss": 0.5741,
+      "step": 34658
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.172360627103406e-06,
+      "loss": 0.6125,
+      "step": 34659
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1719170383522993e-06,
+      "loss": 0.6182,
+      "step": 34660
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.171473489377971e-06,
+      "loss": 0.6378,
+      "step": 34661
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1710299801826817e-06,
+      "loss": 0.628,
+      "step": 34662
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.17058651076868e-06,
+      "loss": 0.5676,
+      "step": 34663
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1701430811382253e-06,
+      "loss": 0.6479,
+      "step": 34664
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.169699691293562e-06,
+      "loss": 0.5905,
+      "step": 34665
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1692563412369503e-06,
+      "loss": 0.5726,
+      "step": 34666
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.168813030970638e-06,
+      "loss": 0.6093,
+      "step": 34667
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1683697604968834e-06,
+      "loss": 0.593,
+      "step": 34668
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.167926529817934e-06,
+      "loss": 0.5516,
+      "step": 34669
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.167483338936045e-06,
+      "loss": 0.5803,
+      "step": 34670
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.167040187853464e-06,
+      "loss": 0.6131,
+      "step": 34671
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1665970765724487e-06,
+      "loss": 0.6015,
+      "step": 34672
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.166154005095248e-06,
+      "loss": 0.6431,
+      "step": 34673
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.165710973424112e-06,
+      "loss": 0.6867,
+      "step": 34674
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.165267981561293e-06,
+      "loss": 0.6295,
+      "step": 34675
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.16482502950904e-06,
+      "loss": 0.6071,
+      "step": 34676
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.164382117269608e-06,
+      "loss": 0.6129,
+      "step": 34677
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.163939244845242e-06,
+      "loss": 0.6407,
+      "step": 34678
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1634964122381995e-06,
+      "loss": 0.5797,
+      "step": 34679
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.163053619450726e-06,
+      "loss": 0.5642,
+      "step": 34680
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.162610866485072e-06,
+      "loss": 0.6221,
+      "step": 34681
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1621681533434856e-06,
+      "loss": 0.6169,
+      "step": 34682
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1617254800282207e-06,
+      "loss": 0.5992,
+      "step": 34683
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1612828465415202e-06,
+      "loss": 0.6366,
+      "step": 34684
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.160840252885644e-06,
+      "loss": 0.6095,
+      "step": 34685
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1603976990628286e-06,
+      "loss": 0.5798,
+      "step": 34686
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.159955185075331e-06,
+      "loss": 0.611,
+      "step": 34687
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1595127109253945e-06,
+      "loss": 0.6557,
+      "step": 34688
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.159070276615273e-06,
+      "loss": 0.6009,
+      "step": 34689
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.158627882147212e-06,
+      "loss": 0.5888,
+      "step": 34690
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.158185527523459e-06,
+      "loss": 0.5467,
+      "step": 34691
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.157743212746258e-06,
+      "loss": 0.6329,
+      "step": 34692
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1573009378178645e-06,
+      "loss": 0.6597,
+      "step": 34693
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.156858702740522e-06,
+      "loss": 0.6453,
+      "step": 34694
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1564165075164733e-06,
+      "loss": 0.5552,
+      "step": 34695
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1559743521479757e-06,
+      "loss": 0.6471,
+      "step": 34696
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1555322366372644e-06,
+      "loss": 0.5841,
+      "step": 34697
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1550901609865937e-06,
+      "loss": 0.5856,
+      "step": 34698
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1546481251982045e-06,
+      "loss": 0.6002,
+      "step": 34699
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1542061292743486e-06,
+      "loss": 0.6355,
+      "step": 34700
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1537641732172665e-06,
+      "loss": 0.6592,
+      "step": 34701
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.153322257029211e-06,
+      "loss": 0.634,
+      "step": 34702
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.152880380712419e-06,
+      "loss": 0.5969,
+      "step": 34703
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.152438544269142e-06,
+      "loss": 0.6146,
+      "step": 34704
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.151996747701621e-06,
+      "loss": 0.6538,
+      "step": 34705
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1515549910121047e-06,
+      "loss": 0.5935,
+      "step": 34706
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.151113274202836e-06,
+      "loss": 0.5771,
+      "step": 34707
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1506715972760605e-06,
+      "loss": 0.6588,
+      "step": 34708
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1502299602340183e-06,
+      "loss": 0.6537,
+      "step": 34709
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1497883630789583e-06,
+      "loss": 0.5742,
+      "step": 34710
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.149346805813124e-06,
+      "loss": 0.5425,
+      "step": 34711
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.148905288438757e-06,
+      "loss": 0.6435,
+      "step": 34712
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.148463810958099e-06,
+      "loss": 0.6136,
+      "step": 34713
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1480223733733986e-06,
+      "loss": 0.6377,
+      "step": 34714
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.147580975686896e-06,
+      "loss": 0.6482,
+      "step": 34715
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1471396179008307e-06,
+      "loss": 0.5728,
+      "step": 34716
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1466983000174526e-06,
+      "loss": 0.6288,
+      "step": 34717
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.146257022039e-06,
+      "loss": 0.6008,
+      "step": 34718
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1458157839677153e-06,
+      "loss": 0.5849,
+      "step": 34719
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.145374585805838e-06,
+      "loss": 0.579,
+      "step": 34720
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1449334275556166e-06,
+      "loss": 0.6142,
+      "step": 34721
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.144492309219285e-06,
+      "loss": 0.5566,
+      "step": 34722
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1440512307990945e-06,
+      "loss": 0.5858,
+      "step": 34723
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1436101922972753e-06,
+      "loss": 0.6317,
+      "step": 34724
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1431691937160758e-06,
+      "loss": 0.5999,
+      "step": 34725
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1427282350577315e-06,
+      "loss": 0.5198,
+      "step": 34726
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1422873163244894e-06,
+      "loss": 0.6045,
+      "step": 34727
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1418464375185864e-06,
+      "loss": 0.659,
+      "step": 34728
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1414055986422633e-06,
+      "loss": 0.5938,
+      "step": 34729
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.140964799697758e-06,
+      "loss": 0.5752,
+      "step": 34730
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.140524040687313e-06,
+      "loss": 0.6461,
+      "step": 34731
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.140083321613169e-06,
+      "loss": 0.5936,
+      "step": 34732
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1396426424775594e-06,
+      "loss": 0.5128,
+      "step": 34733
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.139202003282733e-06,
+      "loss": 0.5363,
+      "step": 34734
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.138761404030919e-06,
+      "loss": 0.5707,
+      "step": 34735
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.138320844724362e-06,
+      "loss": 0.6043,
+      "step": 34736
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.137880325365298e-06,
+      "loss": 0.5849,
+      "step": 34737
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1374398459559674e-06,
+      "loss": 0.6364,
+      "step": 34738
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1369994064986087e-06,
+      "loss": 0.5847,
+      "step": 34739
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.136559006995459e-06,
+      "loss": 0.6519,
+      "step": 34740
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1361186474487527e-06,
+      "loss": 0.5827,
+      "step": 34741
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.135678327860734e-06,
+      "loss": 0.6342,
+      "step": 34742
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1352380482336344e-06,
+      "loss": 0.5602,
+      "step": 34743
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1347978085696953e-06,
+      "loss": 0.6125,
+      "step": 34744
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1343576088711524e-06,
+      "loss": 0.6283,
+      "step": 34745
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1339174491402417e-06,
+      "loss": 0.6142,
+      "step": 34746
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.133477329379198e-06,
+      "loss": 0.6141,
+      "step": 34747
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1330372495902618e-06,
+      "loss": 0.5917,
+      "step": 34748
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1325972097756675e-06,
+      "loss": 0.548,
+      "step": 34749
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.132157209937651e-06,
+      "loss": 0.5493,
+      "step": 34750
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.1317172500784454e-06,
+      "loss": 0.565,
+      "step": 34751
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.131277330200291e-06,
+      "loss": 0.6226,
+      "step": 34752
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.130837450305422e-06,
+      "loss": 0.619,
+      "step": 34753
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1303976103960688e-06,
+      "loss": 0.6509,
+      "step": 34754
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1299578104744735e-06,
+      "loss": 0.6034,
+      "step": 34755
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1295180505428657e-06,
+      "loss": 0.5942,
+      "step": 34756
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1290783306034835e-06,
+      "loss": 0.56,
+      "step": 34757
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1286386506585567e-06,
+      "loss": 0.697,
+      "step": 34758
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1281990107103233e-06,
+      "loss": 0.6243,
+      "step": 34759
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1277594107610143e-06,
+      "loss": 0.5834,
+      "step": 34760
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.127319850812871e-06,
+      "loss": 0.5921,
+      "step": 34761
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.126880330868115e-06,
+      "loss": 0.6315,
+      "step": 34762
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.126440850928989e-06,
+      "loss": 0.612,
+      "step": 34763
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1260014109977213e-06,
+      "loss": 0.5585,
+      "step": 34764
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.125562011076547e-06,
+      "loss": 0.5642,
+      "step": 34765
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1251226511676993e-06,
+      "loss": 0.5933,
+      "step": 34766
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1246833312734095e-06,
+      "loss": 0.5451,
+      "step": 34767
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1242440513959084e-06,
+      "loss": 0.5914,
+      "step": 34768
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.123804811537431e-06,
+      "loss": 0.552,
+      "step": 34769
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.123365611700209e-06,
+      "loss": 0.5674,
+      "step": 34770
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.12292645188647e-06,
+      "loss": 0.6184,
+      "step": 34771
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.122487332098451e-06,
+      "loss": 0.6411,
+      "step": 34772
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.122048252338381e-06,
+      "loss": 0.5682,
+      "step": 34773
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.121609212608492e-06,
+      "loss": 0.5822,
+      "step": 34774
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1211702129110103e-06,
+      "loss": 0.5209,
+      "step": 34775
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1207312532481727e-06,
+      "loss": 0.5981,
+      "step": 34776
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.120292333622207e-06,
+      "loss": 0.5792,
+      "step": 34777
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1198534540353443e-06,
+      "loss": 0.5425,
+      "step": 34778
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.119414614489811e-06,
+      "loss": 0.6057,
+      "step": 34779
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1189758149878425e-06,
+      "loss": 0.6385,
+      "step": 34780
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1185370555316643e-06,
+      "loss": 0.5936,
+      "step": 34781
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1180983361235084e-06,
+      "loss": 0.5716,
+      "step": 34782
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.117659656765605e-06,
+      "loss": 0.6081,
+      "step": 34783
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1172210174601804e-06,
+      "loss": 0.5382,
+      "step": 34784
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.116782418209462e-06,
+      "loss": 0.648,
+      "step": 34785
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1163438590156827e-06,
+      "loss": 0.6145,
+      "step": 34786
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1159053398810705e-06,
+      "loss": 0.6177,
+      "step": 34787
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.115466860807851e-06,
+      "loss": 0.6288,
+      "step": 34788
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1150284217982508e-06,
+      "loss": 0.5352,
+      "step": 34789
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1145900228545034e-06,
+      "loss": 0.5827,
+      "step": 34790
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.114151663978834e-06,
+      "loss": 0.6088,
+      "step": 34791
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.113713345173467e-06,
+      "loss": 0.5675,
+      "step": 34792
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1132750664406343e-06,
+      "loss": 0.6271,
+      "step": 34793
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.112836827782561e-06,
+      "loss": 0.5698,
+      "step": 34794
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.112398629201472e-06,
+      "loss": 0.585,
+      "step": 34795
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1119604706995944e-06,
+      "loss": 0.5796,
+      "step": 34796
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1115223522791586e-06,
+      "loss": 0.5614,
+      "step": 34797
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1110842739423843e-06,
+      "loss": 0.6333,
+      "step": 34798
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.110646235691507e-06,
+      "loss": 0.6031,
+      "step": 34799
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1102082375287415e-06,
+      "loss": 0.6311,
+      "step": 34800
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.109770279456321e-06,
+      "loss": 0.538,
+      "step": 34801
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.109332361476466e-06,
+      "loss": 0.5684,
+      "step": 34802
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.108894483591407e-06,
+      "loss": 0.6109,
+      "step": 34803
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1084566458033663e-06,
+      "loss": 0.6115,
+      "step": 34804
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.108018848114568e-06,
+      "loss": 0.5242,
+      "step": 34805
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.107581090527234e-06,
+      "loss": 0.6097,
+      "step": 34806
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1071433730435953e-06,
+      "loss": 0.6149,
+      "step": 34807
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1067056956658726e-06,
+      "loss": 0.6415,
+      "step": 34808
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.106268058396288e-06,
+      "loss": 0.5464,
+      "step": 34809
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1058304612370685e-06,
+      "loss": 0.6114,
+      "step": 34810
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1053929041904374e-06,
+      "loss": 0.6606,
+      "step": 34811
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.104955387258617e-06,
+      "loss": 0.579,
+      "step": 34812
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.104517910443826e-06,
+      "loss": 0.6279,
+      "step": 34813
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1040804737482965e-06,
+      "loss": 0.634,
+      "step": 34814
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1036430771742456e-06,
+      "loss": 0.6504,
+      "step": 34815
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1032057207238977e-06,
+      "loss": 0.5834,
+      "step": 34816
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1027684043994702e-06,
+      "loss": 0.5825,
+      "step": 34817
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1023311282031935e-06,
+      "loss": 0.5838,
+      "step": 34818
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.101893892137281e-06,
+      "loss": 0.6431,
+      "step": 34819
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.101456696203962e-06,
+      "loss": 0.5483,
+      "step": 34820
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1010195404054544e-06,
+      "loss": 0.6339,
+      "step": 34821
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.100582424743981e-06,
+      "loss": 0.6016,
+      "step": 34822
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.1001453492217584e-06,
+      "loss": 0.6329,
+      "step": 34823
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0997083138410135e-06,
+      "loss": 0.5841,
+      "step": 34824
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0992713186039636e-06,
+      "loss": 0.6292,
+      "step": 34825
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0988343635128305e-06,
+      "loss": 0.5993,
+      "step": 34826
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0983974485698323e-06,
+      "loss": 0.6123,
+      "step": 34827
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0979605737771926e-06,
+      "loss": 0.6029,
+      "step": 34828
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0975237391371296e-06,
+      "loss": 0.5394,
+      "step": 34829
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0970869446518595e-06,
+      "loss": 0.5622,
+      "step": 34830
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0966501903236082e-06,
+      "loss": 0.5821,
+      "step": 34831
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0962134761545917e-06,
+      "loss": 0.6066,
+      "step": 34832
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.095776802147029e-06,
+      "loss": 0.5476,
+      "step": 34833
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0953401683031384e-06,
+      "loss": 0.6129,
+      "step": 34834
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.09490357462514e-06,
+      "loss": 0.612,
+      "step": 34835
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0944670211152496e-06,
+      "loss": 0.5901,
+      "step": 34836
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.094030507775693e-06,
+      "loss": 0.5976,
+      "step": 34837
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.093594034608678e-06,
+      "loss": 0.6107,
+      "step": 34838
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.093157601616429e-06,
+      "loss": 0.6198,
+      "step": 34839
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0927212088011595e-06,
+      "loss": 0.5953,
+      "step": 34840
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.092284856165092e-06,
+      "loss": 0.6148,
+      "step": 34841
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.091848543710441e-06,
+      "loss": 0.5503,
+      "step": 34842
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0914122714394226e-06,
+      "loss": 0.5961,
+      "step": 34843
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0909760393542533e-06,
+      "loss": 0.5911,
+      "step": 34844
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.090539847457154e-06,
+      "loss": 0.6083,
+      "step": 34845
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.090103695750336e-06,
+      "loss": 0.5951,
+      "step": 34846
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.089667584236017e-06,
+      "loss": 0.6122,
+      "step": 34847
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0892315129164153e-06,
+      "loss": 0.5632,
+      "step": 34848
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.088795481793745e-06,
+      "loss": 0.6375,
+      "step": 34849
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.088359490870222e-06,
+      "loss": 0.6239,
+      "step": 34850
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.087923540148059e-06,
+      "loss": 0.5818,
+      "step": 34851
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0874876296294766e-06,
+      "loss": 0.6069,
+      "step": 34852
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.087051759316686e-06,
+      "loss": 0.6114,
+      "step": 34853
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0866159292119036e-06,
+      "loss": 0.67,
+      "step": 34854
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.086180139317341e-06,
+      "loss": 0.6059,
+      "step": 34855
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.085744389635216e-06,
+      "loss": 0.5844,
+      "step": 34856
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0853086801677404e-06,
+      "loss": 0.6034,
+      "step": 34857
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0848730109171333e-06,
+      "loss": 0.6261,
+      "step": 34858
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0844373818856e-06,
+      "loss": 0.646,
+      "step": 34859
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.084001793075361e-06,
+      "loss": 0.603,
+      "step": 34860
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.083566244488624e-06,
+      "loss": 0.6233,
+      "step": 34861
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0831307361276086e-06,
+      "loss": 0.5963,
+      "step": 34862
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.082695267994521e-06,
+      "loss": 0.5559,
+      "step": 34863
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0822598400915837e-06,
+      "loss": 0.6093,
+      "step": 34864
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0818244524209973e-06,
+      "loss": 0.5926,
+      "step": 34865
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.081389104984982e-06,
+      "loss": 0.6184,
+      "step": 34866
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0809537977857485e-06,
+      "loss": 0.6073,
+      "step": 34867
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.080518530825505e-06,
+      "loss": 0.57,
+      "step": 34868
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0800833041064693e-06,
+      "loss": 0.5418,
+      "step": 34869
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0796481176308493e-06,
+      "loss": 0.5832,
+      "step": 34870
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0792129714008567e-06,
+      "loss": 0.5391,
+      "step": 34871
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0787778654187006e-06,
+      "loss": 0.5913,
+      "step": 34872
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0783427996865956e-06,
+      "loss": 0.6412,
+      "step": 34873
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.07790777420675e-06,
+      "loss": 0.5784,
+      "step": 34874
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0774727889813783e-06,
+      "loss": 0.683,
+      "step": 34875
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0770378440126837e-06,
+      "loss": 0.5578,
+      "step": 34876
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.076602939302882e-06,
+      "loss": 0.5678,
+      "step": 34877
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.076168074854179e-06,
+      "loss": 0.574,
+      "step": 34878
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.075733250668789e-06,
+      "loss": 0.5992,
+      "step": 34879
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0752984667489194e-06,
+      "loss": 0.6469,
+      "step": 34880
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0748637230967795e-06,
+      "loss": 0.6326,
+      "step": 34881
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0744290197145743e-06,
+      "loss": 0.6066,
+      "step": 34882
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0739943566045196e-06,
+      "loss": 0.6166,
+      "step": 34883
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0735597337688186e-06,
+      "loss": 0.5663,
+      "step": 34884
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.073125151209684e-06,
+      "loss": 0.5805,
+      "step": 34885
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.072690608929322e-06,
+      "loss": 0.5725,
+      "step": 34886
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.072256106929941e-06,
+      "loss": 0.611,
+      "step": 34887
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0718216452137474e-06,
+      "loss": 0.6518,
+      "step": 34888
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.071387223782948e-06,
+      "loss": 0.5754,
+      "step": 34889
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.070952842639755e-06,
+      "loss": 0.5574,
+      "step": 34890
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0705185017863727e-06,
+      "loss": 0.6253,
+      "step": 34891
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0700842012250076e-06,
+      "loss": 0.5498,
+      "step": 34892
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.069649940957865e-06,
+      "loss": 0.5705,
+      "step": 34893
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0692157209871566e-06,
+      "loss": 0.6114,
+      "step": 34894
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0687815413150825e-06,
+      "loss": 0.5793,
+      "step": 34895
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0683474019438578e-06,
+      "loss": 0.5709,
+      "step": 34896
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.067913302875677e-06,
+      "loss": 0.6016,
+      "step": 34897
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.0674792441127543e-06,
+      "loss": 0.5898,
+      "step": 34898
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0670452256572903e-06,
+      "loss": 0.6055,
+      "step": 34899
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.066611247511496e-06,
+      "loss": 0.5681,
+      "step": 34900
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0661773096775704e-06,
+      "loss": 0.6492,
+      "step": 34901
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.065743412157726e-06,
+      "loss": 0.6181,
+      "step": 34902
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.065309554954158e-06,
+      "loss": 0.6118,
+      "step": 34903
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.06487573806908e-06,
+      "loss": 0.627,
+      "step": 34904
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0644419615046874e-06,
+      "loss": 0.6258,
+      "step": 34905
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.064008225263193e-06,
+      "loss": 0.6066,
+      "step": 34906
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0635745293467967e-06,
+      "loss": 0.6478,
+      "step": 34907
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.063140873757703e-06,
+      "loss": 0.6194,
+      "step": 34908
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0627072584981143e-06,
+      "loss": 0.6055,
+      "step": 34909
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.062273683570233e-06,
+      "loss": 0.6938,
+      "step": 34910
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.061840148976265e-06,
+      "loss": 0.5788,
+      "step": 34911
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0614066547184106e-06,
+      "loss": 0.6322,
+      "step": 34912
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.060973200798878e-06,
+      "loss": 0.5887,
+      "step": 34913
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.060539787219862e-06,
+      "loss": 0.6187,
+      "step": 34914
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.060106413983569e-06,
+      "loss": 0.6257,
+      "step": 34915
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0596730810921996e-06,
+      "loss": 0.6404,
+      "step": 34916
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0592397885479588e-06,
+      "loss": 0.6125,
+      "step": 34917
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0588065363530464e-06,
+      "loss": 0.5712,
+      "step": 34918
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.058373324509664e-06,
+      "loss": 0.6236,
+      "step": 34919
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0579401530200096e-06,
+      "loss": 0.5993,
+      "step": 34920
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0575070218862893e-06,
+      "loss": 0.5461,
+      "step": 34921
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.057073931110699e-06,
+      "loss": 0.6405,
+      "step": 34922
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0566408806954483e-06,
+      "loss": 0.6091,
+      "step": 34923
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0562078706427268e-06,
+      "loss": 0.6272,
+      "step": 34924
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.055774900954741e-06,
+      "loss": 0.6195,
+      "step": 34925
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.05534197163369e-06,
+      "loss": 0.6322,
+      "step": 34926
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0549090826817706e-06,
+      "loss": 0.6369,
+      "step": 34927
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.054476234101188e-06,
+      "loss": 0.6595,
+      "step": 34928
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0540434258941377e-06,
+      "loss": 0.5506,
+      "step": 34929
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.053610658062821e-06,
+      "loss": 0.6185,
+      "step": 34930
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0531779306094325e-06,
+      "loss": 0.5639,
+      "step": 34931
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0527452435361763e-06,
+      "loss": 0.5666,
+      "step": 34932
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.052312596845247e-06,
+      "loss": 0.5613,
+      "step": 34933
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.051879990538851e-06,
+      "loss": 0.578,
+      "step": 34934
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.051447424619174e-06,
+      "loss": 0.5994,
+      "step": 34935
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0510148990884238e-06,
+      "loss": 0.5819,
+      "step": 34936
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0505824139487916e-06,
+      "loss": 0.6844,
+      "step": 34937
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0501499692024806e-06,
+      "loss": 0.5586,
+      "step": 34938
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0497175648516842e-06,
+      "loss": 0.6115,
+      "step": 34939
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0492852008986064e-06,
+      "loss": 0.5353,
+      "step": 34940
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.048852877345433e-06,
+      "loss": 0.6144,
+      "step": 34941
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.04842059419437e-06,
+      "loss": 0.6134,
+      "step": 34942
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0479883514476073e-06,
+      "loss": 0.5539,
+      "step": 34943
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.047556149107348e-06,
+      "loss": 0.6119,
+      "step": 34944
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0471239871757853e-06,
+      "loss": 0.6465,
+      "step": 34945
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0466918656551138e-06,
+      "loss": 0.6202,
+      "step": 34946
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0462597845475306e-06,
+      "loss": 0.6073,
+      "step": 34947
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0458277438552275e-06,
+      "loss": 0.6513,
+      "step": 34948
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0453957435804063e-06,
+      "loss": 0.6394,
+      "step": 34949
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0449637837252588e-06,
+      "loss": 0.6077,
+      "step": 34950
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0445318642919808e-06,
+      "loss": 0.6378,
+      "step": 34951
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0440999852827624e-06,
+      "loss": 0.6027,
+      "step": 34952
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0436681466998054e-06,
+      "loss": 0.6094,
+      "step": 34953
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0432363485452977e-06,
+      "loss": 0.5732,
+      "step": 34954
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.04280459082144e-06,
+      "loss": 0.6498,
+      "step": 34955
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0423728735304216e-06,
+      "loss": 0.592,
+      "step": 34956
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0419411966744375e-06,
+      "loss": 0.618,
+      "step": 34957
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0415095602556767e-06,
+      "loss": 0.6338,
+      "step": 34958
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.041077964276341e-06,
+      "loss": 0.5487,
+      "step": 34959
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.040646408738616e-06,
+      "loss": 0.6357,
+      "step": 34960
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.040214893644702e-06,
+      "loss": 0.63,
+      "step": 34961
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0397834189967826e-06,
+      "loss": 0.6053,
+      "step": 34962
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.039351984797058e-06,
+      "loss": 0.6286,
+      "step": 34963
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0389205910477137e-06,
+      "loss": 0.6058,
+      "step": 34964
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.038489237750949e-06,
+      "loss": 0.5758,
+      "step": 34965
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0380579249089518e-06,
+      "loss": 0.5985,
+      "step": 34966
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.037626652523914e-06,
+      "loss": 0.6555,
+      "step": 34967
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.037195420598027e-06,
+      "loss": 0.5909,
+      "step": 34968
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.03676422913348e-06,
+      "loss": 0.5728,
+      "step": 34969
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0363330781324686e-06,
+      "loss": 0.6281,
+      "step": 34970
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.035901967597178e-06,
+      "loss": 0.6314,
+      "step": 34971
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0354708975298064e-06,
+      "loss": 0.5873,
+      "step": 34972
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0350398679325357e-06,
+      "loss": 0.6155,
+      "step": 34973
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0346088788075623e-06,
+      "loss": 0.5744,
+      "step": 34974
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.034177930157072e-06,
+      "loss": 0.633,
+      "step": 34975
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0337470219832578e-06,
+      "loss": 0.6273,
+      "step": 34976
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0333161542883084e-06,
+      "loss": 0.5851,
+      "step": 34977
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.032885327074413e-06,
+      "loss": 0.5422,
+      "step": 34978
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.032454540343759e-06,
+      "loss": 0.5977,
+      "step": 34979
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0320237940985378e-06,
+      "loss": 0.5799,
+      "step": 34980
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.031593088340935e-06,
+      "loss": 0.5727,
+      "step": 34981
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0311624230731454e-06,
+      "loss": 0.5996,
+      "step": 34982
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0307317982973516e-06,
+      "loss": 0.5979,
+      "step": 34983
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.030301214015744e-06,
+      "loss": 0.5805,
+      "step": 34984
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0298706702305073e-06,
+      "loss": 0.5965,
+      "step": 34985
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0294401669438346e-06,
+      "loss": 0.5854,
+      "step": 34986
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0290097041579103e-06,
+      "loss": 0.5715,
+      "step": 34987
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.028579281874923e-06,
+      "loss": 0.6156,
+      "step": 34988
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.028148900097059e-06,
+      "loss": 0.6753,
+      "step": 34989
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0277185588265016e-06,
+      "loss": 0.5857,
+      "step": 34990
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.027288258065444e-06,
+      "loss": 0.5955,
+      "step": 34991
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0268579978160663e-06,
+      "loss": 0.6132,
+      "step": 34992
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.026427778080562e-06,
+      "loss": 0.5923,
+      "step": 34993
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0259975988611125e-06,
+      "loss": 0.5873,
+      "step": 34994
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0255674601599053e-06,
+      "loss": 0.6557,
+      "step": 34995
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0251373619791216e-06,
+      "loss": 0.6501,
+      "step": 34996
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.024707304320954e-06,
+      "loss": 0.622,
+      "step": 34997
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0242772871875802e-06,
+      "loss": 0.5615,
+      "step": 34998
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.023847310581196e-06,
+      "loss": 0.5818,
+      "step": 34999
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.023417374503974e-06,
+      "loss": 0.6156,
+      "step": 35000
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0229874789581074e-06,
+      "loss": 0.6415,
+      "step": 35001
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0225576239457746e-06,
+      "loss": 0.6264,
+      "step": 35002
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0221278094691653e-06,
+      "loss": 0.6595,
+      "step": 35003
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0216980355304605e-06,
+      "loss": 0.6333,
+      "step": 35004
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.021268302131846e-06,
+      "loss": 0.6279,
+      "step": 35005
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0208386092754995e-06,
+      "loss": 0.6147,
+      "step": 35006
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.020408956963613e-06,
+      "loss": 0.6232,
+      "step": 35007
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.019979345198364e-06,
+      "loss": 0.5884,
+      "step": 35008
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.019549773981936e-06,
+      "loss": 0.6015,
+      "step": 35009
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.019120243316517e-06,
+      "loss": 0.5789,
+      "step": 35010
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.018690753204281e-06,
+      "loss": 0.6054,
+      "step": 35011
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0182613036474162e-06,
+      "loss": 0.615,
+      "step": 35012
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.017831894648101e-06,
+      "loss": 0.6412,
+      "step": 35013
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.017402526208523e-06,
+      "loss": 0.6381,
+      "step": 35014
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0169731983308594e-06,
+      "loss": 0.5487,
+      "step": 35015
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0165439110172937e-06,
+      "loss": 0.5741,
+      "step": 35016
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0161146642700035e-06,
+      "loss": 0.5783,
+      "step": 35017
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0156854580911755e-06,
+      "loss": 0.5774,
+      "step": 35018
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0152562924829857e-06,
+      "loss": 0.6069,
+      "step": 35019
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0148271674476194e-06,
+      "loss": 0.5832,
+      "step": 35020
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0143980829872556e-06,
+      "loss": 0.5624,
+      "step": 35021
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0139690391040734e-06,
+      "loss": 0.6308,
+      "step": 35022
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.01354003580025e-06,
+      "loss": 0.6794,
+      "step": 35023
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0131110730779724e-06,
+      "loss": 0.6208,
+      "step": 35024
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0126821509394155e-06,
+      "loss": 0.6113,
+      "step": 35025
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0122532693867603e-06,
+      "loss": 0.5772,
+      "step": 35026
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0118244284221854e-06,
+      "loss": 0.5762,
+      "step": 35027
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0113956280478677e-06,
+      "loss": 0.608,
+      "step": 35028
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.010966868265991e-06,
+      "loss": 0.5762,
+      "step": 35029
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0105381490787278e-06,
+      "loss": 0.5834,
+      "step": 35030
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.010109470488264e-06,
+      "loss": 0.5402,
+      "step": 35031
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0096808324967733e-06,
+      "loss": 0.6203,
+      "step": 35032
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0092522351064335e-06,
+      "loss": 0.5487,
+      "step": 35033
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0088236783194215e-06,
+      "loss": 0.6023,
+      "step": 35034
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0083951621379196e-06,
+      "loss": 0.6138,
+      "step": 35035
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0079666865641e-06,
+      "loss": 0.5658,
+      "step": 35036
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.007538251600147e-06,
+      "loss": 0.574,
+      "step": 35037
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.007109857248227e-06,
+      "loss": 0.5648,
+      "step": 35038
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.006681503510526e-06,
+      "loss": 0.6117,
+      "step": 35039
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0062531903892146e-06,
+      "loss": 0.6277,
+      "step": 35040
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.005824917886474e-06,
+      "loss": 0.6411,
+      "step": 35041
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0053966860044795e-06,
+      "loss": 0.6348,
+      "step": 35042
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.004968494745404e-06,
+      "loss": 0.614,
+      "step": 35043
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.0045403441114232e-06,
+      "loss": 0.5386,
+      "step": 35044
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.004112234104717e-06,
+      "loss": 0.6175,
+      "step": 35045
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0036841647274587e-06,
+      "loss": 0.5678,
+      "step": 35046
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0032561359818194e-06,
+      "loss": 0.6038,
+      "step": 35047
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.002828147869984e-06,
+      "loss": 0.551,
+      "step": 35048
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0024002003941147e-06,
+      "loss": 0.6023,
+      "step": 35049
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.001972293556396e-06,
+      "loss": 0.597,
+      "step": 35050
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0015444273589947e-06,
+      "loss": 0.5764,
+      "step": 35051
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0011166018040915e-06,
+      "loss": 0.6282,
+      "step": 35052
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.0006888168938576e-06,
+      "loss": 0.593,
+      "step": 35053
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.000261072630466e-06,
+      "loss": 0.6054,
+      "step": 35054
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9998333690160887e-06,
+      "loss": 0.5975,
+      "step": 35055
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9994057060529027e-06,
+      "loss": 0.6367,
+      "step": 35056
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9989780837430783e-06,
+      "loss": 0.5498,
+      "step": 35057
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9985505020887907e-06,
+      "loss": 0.6301,
+      "step": 35058
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9981229610922115e-06,
+      "loss": 0.6581,
+      "step": 35059
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9976954607555144e-06,
+      "loss": 0.6377,
+      "step": 35060
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9972680010808664e-06,
+      "loss": 0.5914,
+      "step": 35061
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.996840582070446e-06,
+      "loss": 0.6051,
+      "step": 35062
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.996413203726423e-06,
+      "loss": 0.5598,
+      "step": 35063
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.995985866050968e-06,
+      "loss": 0.6657,
+      "step": 35064
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9955585690462496e-06,
+      "loss": 0.6,
+      "step": 35065
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9951313127144467e-06,
+      "loss": 0.5772,
+      "step": 35066
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9947040970577245e-06,
+      "loss": 0.5941,
+      "step": 35067
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.994276922078253e-06,
+      "loss": 0.6067,
+      "step": 35068
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9938497877782104e-06,
+      "loss": 0.5777,
+      "step": 35069
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9934226941597566e-06,
+      "loss": 0.6336,
+      "step": 35070
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9929956412250693e-06,
+      "loss": 0.5943,
+      "step": 35071
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9925686289763134e-06,
+      "loss": 0.6024,
+      "step": 35072
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.992141657415665e-06,
+      "loss": 0.6112,
+      "step": 35073
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.991714726545286e-06,
+      "loss": 0.642,
+      "step": 35074
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.991287836367356e-06,
+      "loss": 0.5924,
+      "step": 35075
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.990860986884032e-06,
+      "loss": 0.6212,
+      "step": 35076
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9904341780974924e-06,
+      "loss": 0.6385,
+      "step": 35077
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9900074100098997e-06,
+      "loss": 0.5554,
+      "step": 35078
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9895806826234277e-06,
+      "loss": 0.5613,
+      "step": 35079
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9891539959402418e-06,
+      "loss": 0.6019,
+      "step": 35080
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9887273499625103e-06,
+      "loss": 0.6198,
+      "step": 35081
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9883007446923997e-06,
+      "loss": 0.6145,
+      "step": 35082
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9878741801320812e-06,
+      "loss": 0.5638,
+      "step": 35083
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9874476562837208e-06,
+      "loss": 0.6086,
+      "step": 35084
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9870211731494825e-06,
+      "loss": 0.6056,
+      "step": 35085
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.986594730731539e-06,
+      "loss": 0.597,
+      "step": 35086
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.986168329032054e-06,
+      "loss": 0.5808,
+      "step": 35087
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9857419680531963e-06,
+      "loss": 0.6272,
+      "step": 35088
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.985315647797127e-06,
+      "loss": 0.6181,
+      "step": 35089
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9848893682660186e-06,
+      "loss": 0.6431,
+      "step": 35090
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9844631294620355e-06,
+      "loss": 0.5694,
+      "step": 35091
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.984036931387342e-06,
+      "loss": 0.6159,
+      "step": 35092
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9836107740441015e-06,
+      "loss": 0.6437,
+      "step": 35093
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.983184657434486e-06,
+      "loss": 0.6073,
+      "step": 35094
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.982758581560653e-06,
+      "loss": 0.6185,
+      "step": 35095
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9823325464247758e-06,
+      "loss": 0.5634,
+      "step": 35096
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9819065520290148e-06,
+      "loss": 0.5943,
+      "step": 35097
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9814805983755335e-06,
+      "loss": 0.6364,
+      "step": 35098
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.981054685466496e-06,
+      "loss": 0.5698,
+      "step": 35099
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.980628813304071e-06,
+      "loss": 0.6193,
+      "step": 35100
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.980202981890419e-06,
+      "loss": 0.6484,
+      "step": 35101
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.979777191227704e-06,
+      "loss": 0.6024,
+      "step": 35102
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.979351441318088e-06,
+      "loss": 0.5695,
+      "step": 35103
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.978925732163739e-06,
+      "loss": 0.5839,
+      "step": 35104
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9785000637668173e-06,
+      "loss": 0.5773,
+      "step": 35105
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9780744361294836e-06,
+      "loss": 0.5547,
+      "step": 35106
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.977648849253907e-06,
+      "loss": 0.6131,
+      "step": 35107
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9772233031422416e-06,
+      "loss": 0.5931,
+      "step": 35108
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.976797797796658e-06,
+      "loss": 0.6354,
+      "step": 35109
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9763723332193097e-06,
+      "loss": 0.6245,
+      "step": 35110
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.975946909412367e-06,
+      "loss": 0.5935,
+      "step": 35111
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.975521526377986e-06,
+      "loss": 0.6273,
+      "step": 35112
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.975096184118335e-06,
+      "loss": 0.544,
+      "step": 35113
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.974670882635565e-06,
+      "loss": 0.591,
+      "step": 35114
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.974245621931845e-06,
+      "loss": 0.6204,
+      "step": 35115
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.973820402009332e-06,
+      "loss": 0.6245,
+      "step": 35116
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.973395222870189e-06,
+      "loss": 0.5776,
+      "step": 35117
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9729700845165765e-06,
+      "loss": 0.6605,
+      "step": 35118
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.972544986950654e-06,
+      "loss": 0.5988,
+      "step": 35119
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9721199301745784e-06,
+      "loss": 0.6164,
+      "step": 35120
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.971694914190516e-06,
+      "loss": 0.5783,
+      "step": 35121
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9712699390006217e-06,
+      "loss": 0.5583,
+      "step": 35122
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.970845004607055e-06,
+      "loss": 0.6143,
+      "step": 35123
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.970420111011977e-06,
+      "loss": 0.6595,
+      "step": 35124
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9699952582175474e-06,
+      "loss": 0.5608,
+      "step": 35125
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9695704462259235e-06,
+      "loss": 0.6037,
+      "step": 35126
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.96914567503926e-06,
+      "loss": 0.6046,
+      "step": 35127
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.968720944659723e-06,
+      "loss": 0.6178,
+      "step": 35128
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.968296255089467e-06,
+      "loss": 0.5809,
+      "step": 35129
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9678716063306503e-06,
+      "loss": 0.5827,
+      "step": 35130
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.967446998385427e-06,
+      "loss": 0.6082,
+      "step": 35131
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9670224312559604e-06,
+      "loss": 0.6338,
+      "step": 35132
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9665979049444027e-06,
+      "loss": 0.6264,
+      "step": 35133
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.96617341945292e-06,
+      "loss": 0.5387,
+      "step": 35134
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9657489747836567e-06,
+      "loss": 0.6252,
+      "step": 35135
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9653245709387794e-06,
+      "loss": 0.5495,
+      "step": 35136
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9649002079204373e-06,
+      "loss": 0.5865,
+      "step": 35137
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9644758857307934e-06,
+      "loss": 0.6367,
+      "step": 35138
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9640516043720014e-06,
+      "loss": 0.5786,
+      "step": 35139
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.963627363846217e-06,
+      "loss": 0.5862,
+      "step": 35140
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9632031641555916e-06,
+      "loss": 0.6381,
+      "step": 35141
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.962779005302288e-06,
+      "loss": 0.6375,
+      "step": 35142
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.962354887288458e-06,
+      "loss": 0.5846,
+      "step": 35143
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9619308101162535e-06,
+      "loss": 0.5848,
+      "step": 35144
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9615067737878356e-06,
+      "loss": 0.5766,
+      "step": 35145
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9610827783053564e-06,
+      "loss": 0.6046,
+      "step": 35146
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.960658823670969e-06,
+      "loss": 0.5687,
+      "step": 35147
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9602349098868268e-06,
+      "loss": 0.6434,
+      "step": 35148
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.959811036955088e-06,
+      "loss": 0.6378,
+      "step": 35149
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9593872048779005e-06,
+      "loss": 0.5448,
+      "step": 35150
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9589634136574275e-06,
+      "loss": 0.6334,
+      "step": 35151
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.958539663295812e-06,
+      "loss": 0.5923,
+      "step": 35152
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9581159537952133e-06,
+      "loss": 0.6138,
+      "step": 35153
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9576922851577796e-06,
+      "loss": 0.6048,
+      "step": 35154
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9572686573856693e-06,
+      "loss": 0.5855,
+      "step": 35155
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.956845070481033e-06,
+      "loss": 0.5436,
+      "step": 35156
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9564215244460215e-06,
+      "loss": 0.6334,
+      "step": 35157
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9559980192827867e-06,
+      "loss": 0.6409,
+      "step": 35158
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.955574554993482e-06,
+      "loss": 0.6079,
+      "step": 35159
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.955151131580262e-06,
+      "loss": 0.5765,
+      "step": 35160
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9547277490452734e-06,
+      "loss": 0.5952,
+      "step": 35161
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.954304407390667e-06,
+      "loss": 0.6134,
+      "step": 35162
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9538811066185982e-06,
+      "loss": 0.598,
+      "step": 35163
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.953457846731216e-06,
+      "loss": 0.5994,
+      "step": 35164
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.953034627730669e-06,
+      "loss": 0.5787,
+      "step": 35165
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9526114496191117e-06,
+      "loss": 0.5965,
+      "step": 35166
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.952188312398693e-06,
+      "loss": 0.626,
+      "step": 35167
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.951765216071563e-06,
+      "loss": 0.6171,
+      "step": 35168
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9513421606398676e-06,
+      "loss": 0.6126,
+      "step": 35169
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.950919146105762e-06,
+      "loss": 0.615,
+      "step": 35170
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.950496172471391e-06,
+      "loss": 0.6234,
+      "step": 35171
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9500732397389124e-06,
+      "loss": 0.6067,
+      "step": 35172
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9496503479104636e-06,
+      "loss": 0.5986,
+      "step": 35173
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.949227496988201e-06,
+      "loss": 0.5692,
+      "step": 35174
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9488046869742693e-06,
+      "loss": 0.5856,
+      "step": 35175
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.948381917870822e-06,
+      "loss": 0.6101,
+      "step": 35176
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.947959189680001e-06,
+      "loss": 0.637,
+      "step": 35177
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9475365024039628e-06,
+      "loss": 0.673,
+      "step": 35178
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9471138560448443e-06,
+      "loss": 0.6176,
+      "step": 35179
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.946691250604802e-06,
+      "loss": 0.6019,
+      "step": 35180
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9462686860859794e-06,
+      "loss": 0.6,
+      "step": 35181
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.945846162490521e-06,
+      "loss": 0.6371,
+      "step": 35182
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9454236798205806e-06,
+      "loss": 0.5022,
+      "step": 35183
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.945001238078301e-06,
+      "loss": 0.6097,
+      "step": 35184
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9445788372658305e-06,
+      "loss": 0.5786,
+      "step": 35185
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9441564773853093e-06,
+      "loss": 0.6135,
+      "step": 35186
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9437341584388924e-06,
+      "loss": 0.6321,
+      "step": 35187
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.943311880428719e-06,
+      "loss": 0.6419,
+      "step": 35188
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9428896433569412e-06,
+      "loss": 0.6434,
+      "step": 35189
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9424674472256964e-06,
+      "loss": 0.6473,
+      "step": 35190
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.942045292037137e-06,
+      "loss": 0.6385,
+      "step": 35191
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.9416231777934038e-06,
+      "loss": 0.5608,
+      "step": 35192
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9412011044966437e-06,
+      "loss": 0.5864,
+      "step": 35193
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.940779072149003e-06,
+      "loss": 0.5968,
+      "step": 35194
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.940357080752623e-06,
+      "loss": 0.6029,
+      "step": 35195
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.939935130309646e-06,
+      "loss": 0.529,
+      "step": 35196
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9395132208222224e-06,
+      "loss": 0.5864,
+      "step": 35197
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9390913522924904e-06,
+      "loss": 0.5384,
+      "step": 35198
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9386695247226006e-06,
+      "loss": 0.6175,
+      "step": 35199
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.938247738114687e-06,
+      "loss": 0.5893,
+      "step": 35200
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9378259924708997e-06,
+      "loss": 0.566,
+      "step": 35201
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9374042877933796e-06,
+      "loss": 0.6134,
+      "step": 35202
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.936982624084267e-06,
+      "loss": 0.64,
+      "step": 35203
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9365610013457103e-06,
+      "loss": 0.6361,
+      "step": 35204
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.936139419579849e-06,
+      "loss": 0.6317,
+      "step": 35205
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.935717878788824e-06,
+      "loss": 0.6562,
+      "step": 35206
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.935296378974776e-06,
+      "loss": 0.5976,
+      "step": 35207
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.934874920139851e-06,
+      "loss": 0.5916,
+      "step": 35208
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9344535022861866e-06,
+      "loss": 0.6178,
+      "step": 35209
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9340321254159313e-06,
+      "loss": 0.65,
+      "step": 35210
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.933610789531215e-06,
+      "loss": 0.6178,
+      "step": 35211
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9331894946341877e-06,
+      "loss": 0.5787,
+      "step": 35212
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.932768240726984e-06,
+      "loss": 0.6014,
+      "step": 35213
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9323470278117508e-06,
+      "loss": 0.6179,
+      "step": 35214
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9319258558906216e-06,
+      "loss": 0.6242,
+      "step": 35215
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9315047249657457e-06,
+      "loss": 0.6028,
+      "step": 35216
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9310836350392524e-06,
+      "loss": 0.5796,
+      "step": 35217
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.930662586113289e-06,
+      "loss": 0.5885,
+      "step": 35218
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.930241578189992e-06,
+      "loss": 0.5858,
+      "step": 35219
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.929820611271498e-06,
+      "loss": 0.5632,
+      "step": 35220
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.929399685359952e-06,
+      "loss": 0.6354,
+      "step": 35221
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9289788004574883e-06,
+      "loss": 0.6158,
+      "step": 35222
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9285579565662483e-06,
+      "loss": 0.6104,
+      "step": 35223
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9281371536883665e-06,
+      "loss": 0.616,
+      "step": 35224
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9277163918259868e-06,
+      "loss": 0.6073,
+      "step": 35225
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.927295670981243e-06,
+      "loss": 0.6166,
+      "step": 35226
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9268749911562743e-06,
+      "loss": 0.6773,
+      "step": 35227
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9264543523532162e-06,
+      "loss": 0.6261,
+      "step": 35228
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.92603375457421e-06,
+      "loss": 0.5792,
+      "step": 35229
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9256131978213888e-06,
+      "loss": 0.5847,
+      "step": 35230
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9251926820968948e-06,
+      "loss": 0.5664,
+      "step": 35231
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9247722074028606e-06,
+      "loss": 0.592,
+      "step": 35232
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.924351773741424e-06,
+      "loss": 0.6048,
+      "step": 35233
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9239313811147176e-06,
+      "loss": 0.5837,
+      "step": 35234
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9235110295248847e-06,
+      "loss": 0.5758,
+      "step": 35235
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9230907189740556e-06,
+      "loss": 0.5844,
+      "step": 35236
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9226704494643724e-06,
+      "loss": 0.635,
+      "step": 35237
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.92225022099796e-06,
+      "loss": 0.6177,
+      "step": 35238
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9218300335769637e-06,
+      "loss": 0.6513,
+      "step": 35239
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.921409887203515e-06,
+      "loss": 0.6525,
+      "step": 35240
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9209897818797454e-06,
+      "loss": 0.5919,
+      "step": 35241
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9205697176077953e-06,
+      "loss": 0.5337,
+      "step": 35242
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9201496943897966e-06,
+      "loss": 0.6786,
+      "step": 35243
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9197297122278847e-06,
+      "loss": 0.5763,
+      "step": 35244
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.919309771124188e-06,
+      "loss": 0.5867,
+      "step": 35245
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.918889871080849e-06,
+      "loss": 0.6258,
+      "step": 35246
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9184700120999943e-06,
+      "loss": 0.6286,
+      "step": 35247
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.918050194183765e-06,
+      "loss": 0.6082,
+      "step": 35248
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.917630417334285e-06,
+      "loss": 0.581,
+      "step": 35249
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.917210681553694e-06,
+      "loss": 0.6236,
+      "step": 35250
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9167909868441206e-06,
+      "loss": 0.5422,
+      "step": 35251
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9163713332077014e-06,
+      "loss": 0.6398,
+      "step": 35252
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.915951720646567e-06,
+      "loss": 0.5846,
+      "step": 35253
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.91553214916285e-06,
+      "loss": 0.567,
+      "step": 35254
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.915112618758679e-06,
+      "loss": 0.6147,
+      "step": 35255
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9146931294361915e-06,
+      "loss": 0.5854,
+      "step": 35256
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9142736811975126e-06,
+      "loss": 0.5192,
+      "step": 35257
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9138542740447796e-06,
+      "loss": 0.6494,
+      "step": 35258
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9134349079801216e-06,
+      "loss": 0.5876,
+      "step": 35259
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9130155830056696e-06,
+      "loss": 0.6348,
+      "step": 35260
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9125962991235537e-06,
+      "loss": 0.6041,
+      "step": 35261
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9121770563359022e-06,
+      "loss": 0.6154,
+      "step": 35262
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.911757854644849e-06,
+      "loss": 0.6056,
+      "step": 35263
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9113386940525234e-06,
+      "loss": 0.5853,
+      "step": 35264
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9109195745610554e-06,
+      "loss": 0.5901,
+      "step": 35265
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9105004961725714e-06,
+      "loss": 0.616,
+      "step": 35266
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.910081458889206e-06,
+      "loss": 0.5946,
+      "step": 35267
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.909662462713082e-06,
+      "loss": 0.5967,
+      "step": 35268
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9092435076463368e-06,
+      "loss": 0.5665,
+      "step": 35269
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.908824593691094e-06,
+      "loss": 0.6013,
+      "step": 35270
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9084057208494832e-06,
+      "loss": 0.5831,
+      "step": 35271
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.90798688912363e-06,
+      "loss": 0.6038,
+      "step": 35272
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9075680985156676e-06,
+      "loss": 0.5949,
+      "step": 35273
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9071493490277193e-06,
+      "loss": 0.6087,
+      "step": 35274
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9067306406619212e-06,
+      "loss": 0.6108,
+      "step": 35275
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9063119734203884e-06,
+      "loss": 0.5544,
+      "step": 35276
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9058933473052588e-06,
+      "loss": 0.59,
+      "step": 35277
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.905474762318652e-06,
+      "loss": 0.5647,
+      "step": 35278
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9050562184627008e-06,
+      "loss": 0.5606,
+      "step": 35279
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9046377157395301e-06,
+      "loss": 0.6575,
+      "step": 35280
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.904219254151266e-06,
+      "loss": 0.5898,
+      "step": 35281
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9038008337000357e-06,
+      "loss": 0.5927,
+      "step": 35282
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.903382454387961e-06,
+      "loss": 0.5707,
+      "step": 35283
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.902964116217174e-06,
+      "loss": 0.5453,
+      "step": 35284
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.902545819189795e-06,
+      "loss": 0.5609,
+      "step": 35285
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9021275633079571e-06,
+      "loss": 0.5862,
+      "step": 35286
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9017093485737758e-06,
+      "loss": 0.5577,
+      "step": 35287
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9012911749893837e-06,
+      "loss": 0.5973,
+      "step": 35288
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9008730425569e-06,
+      "loss": 0.6526,
+      "step": 35289
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9004549512784553e-06,
+      "loss": 0.6296,
+      "step": 35290
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.9000369011561704e-06,
+      "loss": 0.59,
+      "step": 35291
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.899618892192171e-06,
+      "loss": 0.5077,
+      "step": 35292
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8992009243885767e-06,
+      "loss": 0.6394,
+      "step": 35293
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.898782997747518e-06,
+      "loss": 0.5796,
+      "step": 35294
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8983651122711134e-06,
+      "loss": 0.6331,
+      "step": 35295
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8979472679614907e-06,
+      "loss": 0.6071,
+      "step": 35296
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8975294648207709e-06,
+      "loss": 0.633,
+      "step": 35297
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8971117028510755e-06,
+      "loss": 0.5747,
+      "step": 35298
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8966939820545305e-06,
+      "loss": 0.6168,
+      "step": 35299
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.896276302433252e-06,
+      "loss": 0.5988,
+      "step": 35300
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8958586639893705e-06,
+      "loss": 0.6001,
+      "step": 35301
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8954410667250045e-06,
+      "loss": 0.636,
+      "step": 35302
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8950235106422765e-06,
+      "loss": 0.6125,
+      "step": 35303
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.894605995743305e-06,
+      "loss": 0.5572,
+      "step": 35304
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8941885220302159e-06,
+      "loss": 0.624,
+      "step": 35305
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8937710895051264e-06,
+      "loss": 0.629,
+      "step": 35306
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8933536981701627e-06,
+      "loss": 0.6149,
+      "step": 35307
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8929363480274433e-06,
+      "loss": 0.613,
+      "step": 35308
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8925190390790894e-06,
+      "loss": 0.6055,
+      "step": 35309
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.892101771327216e-06,
+      "loss": 0.5749,
+      "step": 35310
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8916845447739519e-06,
+      "loss": 0.6654,
+      "step": 35311
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8912673594214115e-06,
+      "loss": 0.6427,
+      "step": 35312
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.89085021527172e-06,
+      "loss": 0.6135,
+      "step": 35313
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8904331123269892e-06,
+      "loss": 0.6051,
+      "step": 35314
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.890016050589345e-06,
+      "loss": 0.5775,
+      "step": 35315
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8895990300609012e-06,
+      "loss": 0.6192,
+      "step": 35316
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.889182050743783e-06,
+      "loss": 0.6376,
+      "step": 35317
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8887651126401063e-06,
+      "loss": 0.5912,
+      "step": 35318
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8883482157519894e-06,
+      "loss": 0.6243,
+      "step": 35319
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.887931360081552e-06,
+      "loss": 0.5708,
+      "step": 35320
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8875145456309062e-06,
+      "loss": 0.6355,
+      "step": 35321
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.887097772402179e-06,
+      "loss": 0.5944,
+      "step": 35322
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8866810403974811e-06,
+      "loss": 0.6506,
+      "step": 35323
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8862643496189382e-06,
+      "loss": 0.6396,
+      "step": 35324
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.885847700068657e-06,
+      "loss": 0.5575,
+      "step": 35325
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8854310917487627e-06,
+      "loss": 0.6381,
+      "step": 35326
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.885014524661367e-06,
+      "loss": 0.5402,
+      "step": 35327
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8845979988085917e-06,
+      "loss": 0.6053,
+      "step": 35328
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.88418151419255e-06,
+      "loss": 0.606,
+      "step": 35329
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8837650708153598e-06,
+      "loss": 0.5184,
+      "step": 35330
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8833486686791336e-06,
+      "loss": 0.6517,
+      "step": 35331
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8829323077859917e-06,
+      "loss": 0.5956,
+      "step": 35332
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.882515988138045e-06,
+      "loss": 0.6735,
+      "step": 35333
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8820997097374161e-06,
+      "loss": 0.5842,
+      "step": 35334
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8816834725862142e-06,
+      "loss": 0.5927,
+      "step": 35335
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.881267276686557e-06,
+      "loss": 0.6098,
+      "step": 35336
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8808511220405568e-06,
+      "loss": 0.5937,
+      "step": 35337
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.880435008650331e-06,
+      "loss": 0.5732,
+      "step": 35338
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.8800189365179933e-06,
+      "loss": 0.5828,
+      "step": 35339
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8796029056456577e-06,
+      "loss": 0.6535,
+      "step": 35340
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.879186916035438e-06,
+      "loss": 0.62,
+      "step": 35341
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.878770967689445e-06,
+      "loss": 0.5834,
+      "step": 35342
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8783550606097978e-06,
+      "loss": 0.5769,
+      "step": 35343
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8779391947986036e-06,
+      "loss": 0.6382,
+      "step": 35344
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8775233702579854e-06,
+      "loss": 0.5916,
+      "step": 35345
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8771075869900434e-06,
+      "loss": 0.6008,
+      "step": 35346
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8766918449969006e-06,
+      "loss": 0.6074,
+      "step": 35347
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.876276144280662e-06,
+      "loss": 0.6169,
+      "step": 35348
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.875860484843447e-06,
+      "loss": 0.5598,
+      "step": 35349
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8754448666873604e-06,
+      "loss": 0.5604,
+      "step": 35350
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.875029289814524e-06,
+      "loss": 0.5842,
+      "step": 35351
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8746137542270372e-06,
+      "loss": 0.5796,
+      "step": 35352
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8741982599270203e-06,
+      "loss": 0.6246,
+      "step": 35353
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8737828069165797e-06,
+      "loss": 0.5949,
+      "step": 35354
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8733673951978304e-06,
+      "loss": 0.6011,
+      "step": 35355
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8729520247728828e-06,
+      "loss": 0.5507,
+      "step": 35356
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.872536695643845e-06,
+      "loss": 0.6653,
+      "step": 35357
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8721214078128258e-06,
+      "loss": 0.6357,
+      "step": 35358
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8717061612819409e-06,
+      "loss": 0.5912,
+      "step": 35359
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8712909560532976e-06,
+      "loss": 0.6569,
+      "step": 35360
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.870875792129002e-06,
+      "loss": 0.5962,
+      "step": 35361
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8704606695111739e-06,
+      "loss": 0.5866,
+      "step": 35362
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8700455882019097e-06,
+      "loss": 0.6192,
+      "step": 35363
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8696305482033283e-06,
+      "loss": 0.6276,
+      "step": 35364
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8692155495175323e-06,
+      "loss": 0.5682,
+      "step": 35365
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8688005921466356e-06,
+      "loss": 0.561,
+      "step": 35366
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8683856760927443e-06,
+      "loss": 0.6027,
+      "step": 35367
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8679708013579667e-06,
+      "loss": 0.6219,
+      "step": 35368
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8675559679444089e-06,
+      "loss": 0.5511,
+      "step": 35369
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8671411758541834e-06,
+      "loss": 0.5963,
+      "step": 35370
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8667264250893934e-06,
+      "loss": 0.5922,
+      "step": 35371
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8663117156521504e-06,
+      "loss": 0.611,
+      "step": 35372
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8658970475445592e-06,
+      "loss": 0.5522,
+      "step": 35373
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8654824207687283e-06,
+      "loss": 0.6468,
+      "step": 35374
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8650678353267604e-06,
+      "loss": 0.6014,
+      "step": 35375
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8646532912207681e-06,
+      "loss": 0.5476,
+      "step": 35376
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8642387884528545e-06,
+      "loss": 0.6347,
+      "step": 35377
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8638243270251267e-06,
+      "loss": 0.6378,
+      "step": 35378
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.863409906939687e-06,
+      "loss": 0.5612,
+      "step": 35379
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8629955281986477e-06,
+      "loss": 0.6011,
+      "step": 35380
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8625811908041114e-06,
+      "loss": 0.5354,
+      "step": 35381
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8621668947581795e-06,
+      "loss": 0.5586,
+      "step": 35382
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.861752640062967e-06,
+      "loss": 0.641,
+      "step": 35383
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.861338426720567e-06,
+      "loss": 0.6612,
+      "step": 35384
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.860924254733093e-06,
+      "loss": 0.6511,
+      "step": 35385
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8605101241026446e-06,
+      "loss": 0.5739,
+      "step": 35386
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.860096034831329e-06,
+      "loss": 0.6296,
+      "step": 35387
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.859681986921248e-06,
+      "loss": 0.5425,
+      "step": 35388
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8592679803745116e-06,
+      "loss": 0.589,
+      "step": 35389
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8588540151932134e-06,
+      "loss": 0.5819,
+      "step": 35390
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8584400913794654e-06,
+      "loss": 0.6128,
+      "step": 35391
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.858026208935364e-06,
+      "loss": 0.5789,
+      "step": 35392
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8576123678630187e-06,
+      "loss": 0.5469,
+      "step": 35393
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8571985681645299e-06,
+      "loss": 0.5871,
+      "step": 35394
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8567848098420005e-06,
+      "loss": 0.5956,
+      "step": 35395
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8563710928975287e-06,
+      "loss": 0.6484,
+      "step": 35396
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.855957417333224e-06,
+      "loss": 0.6648,
+      "step": 35397
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8555437831511835e-06,
+      "loss": 0.6247,
+      "step": 35398
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.855130190353509e-06,
+      "loss": 0.56,
+      "step": 35399
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8547166389423076e-06,
+      "loss": 0.5405,
+      "step": 35400
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8543031289196712e-06,
+      "loss": 0.5709,
+      "step": 35401
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.85388966028771e-06,
+      "loss": 0.6421,
+      "step": 35402
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8534762330485168e-06,
+      "loss": 0.6334,
+      "step": 35403
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8530628472042e-06,
+      "loss": 0.5943,
+      "step": 35404
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.852649502756857e-06,
+      "loss": 0.6523,
+      "step": 35405
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8522361997085881e-06,
+      "loss": 0.6129,
+      "step": 35406
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8518229380614906e-06,
+      "loss": 0.5525,
+      "step": 35407
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8514097178176692e-06,
+      "loss": 0.6264,
+      "step": 35408
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8509965389792195e-06,
+      "loss": 0.6175,
+      "step": 35409
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.850583401548247e-06,
+      "loss": 0.6161,
+      "step": 35410
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8501703055268427e-06,
+      "loss": 0.5745,
+      "step": 35411
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8497572509171124e-06,
+      "loss": 0.6088,
+      "step": 35412
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8493442377211502e-06,
+      "loss": 0.6476,
+      "step": 35413
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8489312659410585e-06,
+      "loss": 0.5741,
+      "step": 35414
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8485183355789348e-06,
+      "loss": 0.5964,
+      "step": 35415
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8481054466368764e-06,
+      "loss": 0.5404,
+      "step": 35416
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8476925991169791e-06,
+      "loss": 0.6366,
+      "step": 35417
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.847279793021346e-06,
+      "loss": 0.5289,
+      "step": 35418
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8468670283520717e-06,
+      "loss": 0.5766,
+      "step": 35419
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8464543051112516e-06,
+      "loss": 0.6333,
+      "step": 35420
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8460416233009892e-06,
+      "loss": 0.6971,
+      "step": 35421
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8456289829233731e-06,
+      "loss": 0.5564,
+      "step": 35422
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.845216383980507e-06,
+      "loss": 0.605,
+      "step": 35423
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8448038264744816e-06,
+      "loss": 0.6166,
+      "step": 35424
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8443913104073984e-06,
+      "loss": 0.6661,
+      "step": 35425
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8439788357813504e-06,
+      "loss": 0.581,
+      "step": 35426
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.843566402598439e-06,
+      "loss": 0.6326,
+      "step": 35427
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8431540108607494e-06,
+      "loss": 0.5222,
+      "step": 35428
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8427416605703864e-06,
+      "loss": 0.574,
+      "step": 35429
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8423293517294393e-06,
+      "loss": 0.6149,
+      "step": 35430
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.841917084340008e-06,
+      "loss": 0.6147,
+      "step": 35431
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8415048584041862e-06,
+      "loss": 0.5961,
+      "step": 35432
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8410926739240675e-06,
+      "loss": 0.6373,
+      "step": 35433
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.840680530901743e-06,
+      "loss": 0.617,
+      "step": 35434
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8402684293393137e-06,
+      "loss": 0.5948,
+      "step": 35435
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8398563692388694e-06,
+      "loss": 0.553,
+      "step": 35436
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.839444350602505e-06,
+      "loss": 0.6229,
+      "step": 35437
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8390323734323113e-06,
+      "loss": 0.6017,
+      "step": 35438
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8386204377303863e-06,
+      "loss": 0.6453,
+      "step": 35439
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8382085434988207e-06,
+      "loss": 0.6534,
+      "step": 35440
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8377966907397061e-06,
+      "loss": 0.6307,
+      "step": 35441
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8373848794551397e-06,
+      "loss": 0.5874,
+      "step": 35442
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.836973109647211e-06,
+      "loss": 0.597,
+      "step": 35443
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8365613813180116e-06,
+      "loss": 0.5413,
+      "step": 35444
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8361496944696333e-06,
+      "loss": 0.6266,
+      "step": 35445
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.835738049104171e-06,
+      "loss": 0.5931,
+      "step": 35446
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.835326445223713e-06,
+      "loss": 0.6283,
+      "step": 35447
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8349148828303565e-06,
+      "loss": 0.6107,
+      "step": 35448
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8345033619261843e-06,
+      "loss": 0.6056,
+      "step": 35449
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8340918825132947e-06,
+      "loss": 0.577,
+      "step": 35450
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8336804445937728e-06,
+      "loss": 0.6491,
+      "step": 35451
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8332690481697145e-06,
+      "loss": 0.6606,
+      "step": 35452
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8328576932432086e-06,
+      "loss": 0.5575,
+      "step": 35453
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.832446379816344e-06,
+      "loss": 0.6166,
+      "step": 35454
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.832035107891209e-06,
+      "loss": 0.5681,
+      "step": 35455
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.831623877469898e-06,
+      "loss": 0.6353,
+      "step": 35456
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.831212688554499e-06,
+      "loss": 0.622,
+      "step": 35457
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8308015411470969e-06,
+      "loss": 0.5873,
+      "step": 35458
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8303904352497881e-06,
+      "loss": 0.5778,
+      "step": 35459
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8299793708646584e-06,
+      "loss": 0.5917,
+      "step": 35460
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8295683479937953e-06,
+      "loss": 0.6433,
+      "step": 35461
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.829157366639287e-06,
+      "loss": 0.6376,
+      "step": 35462
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.828746426803224e-06,
+      "loss": 0.6207,
+      "step": 35463
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8283355284876924e-06,
+      "loss": 0.6309,
+      "step": 35464
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8279246716947853e-06,
+      "loss": 0.6726,
+      "step": 35465
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8275138564265816e-06,
+      "loss": 0.6262,
+      "step": 35466
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8271030826851755e-06,
+      "loss": 0.6065,
+      "step": 35467
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8266923504726497e-06,
+      "loss": 0.5714,
+      "step": 35468
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.826281659791097e-06,
+      "loss": 0.5069,
+      "step": 35469
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8258710106426003e-06,
+      "loss": 0.6541,
+      "step": 35470
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8254604030292467e-06,
+      "loss": 0.662,
+      "step": 35471
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8250498369531211e-06,
+      "loss": 0.5517,
+      "step": 35472
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.824639312416313e-06,
+      "loss": 0.5737,
+      "step": 35473
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8242288294209075e-06,
+      "loss": 0.5617,
+      "step": 35474
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8238183879689897e-06,
+      "loss": 0.623,
+      "step": 35475
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.823407988062642e-06,
+      "loss": 0.6213,
+      "step": 35476
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8229976297039552e-06,
+      "loss": 0.6178,
+      "step": 35477
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8225873128950134e-06,
+      "loss": 0.5589,
+      "step": 35478
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8221770376378977e-06,
+      "loss": 0.6931,
+      "step": 35479
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8217668039346969e-06,
+      "loss": 0.6438,
+      "step": 35480
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8213566117874949e-06,
+      "loss": 0.5046,
+      "step": 35481
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8209464611983752e-06,
+      "loss": 0.6361,
+      "step": 35482
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8205363521694196e-06,
+      "loss": 0.5863,
+      "step": 35483
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8201262847027157e-06,
+      "loss": 0.5591,
+      "step": 35484
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8197162588003447e-06,
+      "loss": 0.6167,
+      "step": 35485
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.8193062744643953e-06,
+      "loss": 0.6131,
+      "step": 35486
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.818896331696941e-06,
+      "loss": 0.5575,
+      "step": 35487
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8184864305000738e-06,
+      "loss": 0.574,
+      "step": 35488
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8180765708758697e-06,
+      "loss": 0.6255,
+      "step": 35489
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.817666752826418e-06,
+      "loss": 0.5688,
+      "step": 35490
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8172569763537973e-06,
+      "loss": 0.6102,
+      "step": 35491
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8168472414600912e-06,
+      "loss": 0.6028,
+      "step": 35492
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8164375481473773e-06,
+      "loss": 0.6533,
+      "step": 35493
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8160278964177446e-06,
+      "loss": 0.6215,
+      "step": 35494
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8156182862732696e-06,
+      "loss": 0.5607,
+      "step": 35495
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8152087177160338e-06,
+      "loss": 0.6426,
+      "step": 35496
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.814799190748121e-06,
+      "loss": 0.5406,
+      "step": 35497
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.814389705371611e-06,
+      "loss": 0.6618,
+      "step": 35498
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8139802615885837e-06,
+      "loss": 0.5813,
+      "step": 35499
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8135708594011181e-06,
+      "loss": 0.5082,
+      "step": 35500
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.813161498811299e-06,
+      "loss": 0.5879,
+      "step": 35501
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8127521798212045e-06,
+      "loss": 0.6734,
+      "step": 35502
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.812342902432913e-06,
+      "loss": 0.5956,
+      "step": 35503
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.811933666648502e-06,
+      "loss": 0.6142,
+      "step": 35504
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8115244724700575e-06,
+      "loss": 0.5769,
+      "step": 35505
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8111153198996535e-06,
+      "loss": 0.5864,
+      "step": 35506
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8107062089393724e-06,
+      "loss": 0.6563,
+      "step": 35507
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8102971395912916e-06,
+      "loss": 0.5781,
+      "step": 35508
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8098881118574885e-06,
+      "loss": 0.6134,
+      "step": 35509
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8094791257400412e-06,
+      "loss": 0.614,
+      "step": 35510
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8090701812410316e-06,
+      "loss": 0.6485,
+      "step": 35511
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8086612783625346e-06,
+      "loss": 0.6404,
+      "step": 35512
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8082524171066285e-06,
+      "loss": 0.5959,
+      "step": 35513
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8078435974753893e-06,
+      "loss": 0.6702,
+      "step": 35514
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8074348194708978e-06,
+      "loss": 0.6344,
+      "step": 35515
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.807026083095229e-06,
+      "loss": 0.6228,
+      "step": 35516
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.806617388350459e-06,
+      "loss": 0.5809,
+      "step": 35517
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.806208735238667e-06,
+      "loss": 0.5993,
+      "step": 35518
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8058001237619282e-06,
+      "loss": 0.5559,
+      "step": 35519
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8053915539223188e-06,
+      "loss": 0.5752,
+      "step": 35520
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8049830257219124e-06,
+      "loss": 0.6204,
+      "step": 35521
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8045745391627901e-06,
+      "loss": 0.63,
+      "step": 35522
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8041660942470207e-06,
+      "loss": 0.6621,
+      "step": 35523
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8037576909766906e-06,
+      "loss": 0.5764,
+      "step": 35524
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8033493293538618e-06,
+      "loss": 0.6091,
+      "step": 35525
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8029410093806188e-06,
+      "loss": 0.5775,
+      "step": 35526
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.80253273105903e-06,
+      "loss": 0.5865,
+      "step": 35527
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8021244943911763e-06,
+      "loss": 0.5981,
+      "step": 35528
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8017162993791281e-06,
+      "loss": 0.613,
+      "step": 35529
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8013081460249605e-06,
+      "loss": 0.6594,
+      "step": 35530
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.800900034330746e-06,
+      "loss": 0.6009,
+      "step": 35531
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8004919642985607e-06,
+      "loss": 0.6195,
+      "step": 35532
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.8000839359304778e-06,
+      "loss": 0.5905,
+      "step": 35533
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7996759492285676e-06,
+      "loss": 0.6034,
+      "step": 35534
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7992680041949074e-06,
+      "loss": 0.6274,
+      "step": 35535
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7988601008315675e-06,
+      "loss": 0.6128,
+      "step": 35536
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7984522391406224e-06,
+      "loss": 0.6502,
+      "step": 35537
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7980444191241408e-06,
+      "loss": 0.5833,
+      "step": 35538
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7976366407841995e-06,
+      "loss": 0.5819,
+      "step": 35539
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7972289041228697e-06,
+      "loss": 0.6236,
+      "step": 35540
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.796821209142221e-06,
+      "loss": 0.5719,
+      "step": 35541
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7964135558443241e-06,
+      "loss": 0.6028,
+      "step": 35542
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.796005944231255e-06,
+      "loss": 0.5972,
+      "step": 35543
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7955983743050797e-06,
+      "loss": 0.6371,
+      "step": 35544
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7951908460678747e-06,
+      "loss": 0.5382,
+      "step": 35545
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7947833595217067e-06,
+      "loss": 0.6282,
+      "step": 35546
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.794375914668649e-06,
+      "loss": 0.6001,
+      "step": 35547
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7939685115107676e-06,
+      "loss": 0.5807,
+      "step": 35548
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7935611500501371e-06,
+      "loss": 0.6753,
+      "step": 35549
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7931538302888241e-06,
+      "loss": 0.5249,
+      "step": 35550
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7927465522289044e-06,
+      "loss": 0.6141,
+      "step": 35551
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7923393158724378e-06,
+      "loss": 0.5899,
+      "step": 35552
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7919321212215024e-06,
+      "loss": 0.6543,
+      "step": 35553
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7915249682781631e-06,
+      "loss": 0.6325,
+      "step": 35554
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7911178570444866e-06,
+      "loss": 0.5777,
+      "step": 35555
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.790710787522547e-06,
+      "loss": 0.5819,
+      "step": 35556
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7903037597144103e-06,
+      "loss": 0.5885,
+      "step": 35557
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7898967736221451e-06,
+      "loss": 0.5678,
+      "step": 35558
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7894898292478158e-06,
+      "loss": 0.5882,
+      "step": 35559
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.789082926593495e-06,
+      "loss": 0.6198,
+      "step": 35560
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7886760656612467e-06,
+      "loss": 0.5408,
+      "step": 35561
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7882692464531458e-06,
+      "loss": 0.589,
+      "step": 35562
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7878624689712477e-06,
+      "loss": 0.5846,
+      "step": 35563
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.787455733217629e-06,
+      "loss": 0.6077,
+      "step": 35564
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7870490391943495e-06,
+      "loss": 0.5866,
+      "step": 35565
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7866423869034822e-06,
+      "loss": 0.6117,
+      "step": 35566
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7862357763470894e-06,
+      "loss": 0.616,
+      "step": 35567
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7858292075272387e-06,
+      "loss": 0.6029,
+      "step": 35568
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.785422680445993e-06,
+      "loss": 0.5796,
+      "step": 35569
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7850161951054235e-06,
+      "loss": 0.5938,
+      "step": 35570
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7846097515075889e-06,
+      "loss": 0.5984,
+      "step": 35571
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7842033496545608e-06,
+      "loss": 0.5792,
+      "step": 35572
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.783796989548402e-06,
+      "loss": 0.6069,
+      "step": 35573
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7833906711911775e-06,
+      "loss": 0.5943,
+      "step": 35574
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.78298439458495e-06,
+      "loss": 0.662,
+      "step": 35575
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7825781597317837e-06,
+      "loss": 0.6284,
+      "step": 35576
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7821719666337467e-06,
+      "loss": 0.589,
+      "step": 35577
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7817658152929006e-06,
+      "loss": 0.6192,
+      "step": 35578
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7813597057113097e-06,
+      "loss": 0.5751,
+      "step": 35579
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.780953637891033e-06,
+      "loss": 0.5684,
+      "step": 35580
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7805476118341425e-06,
+      "loss": 0.6559,
+      "step": 35581
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.780141627542693e-06,
+      "loss": 0.6721,
+      "step": 35582
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7797356850187542e-06,
+      "loss": 0.5946,
+      "step": 35583
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7793297842643853e-06,
+      "loss": 0.6104,
+      "step": 35584
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7789239252816494e-06,
+      "loss": 0.5926,
+      "step": 35585
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7785181080726066e-06,
+      "loss": 0.6448,
+      "step": 35586
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7781123326393235e-06,
+      "loss": 0.5907,
+      "step": 35587
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.777706598983856e-06,
+      "loss": 0.6095,
+      "step": 35588
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.777300907108276e-06,
+      "loss": 0.6092,
+      "step": 35589
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7768952570146314e-06,
+      "loss": 0.5926,
+      "step": 35590
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7764896487049942e-06,
+      "loss": 0.5886,
+      "step": 35591
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7760840821814208e-06,
+      "loss": 0.5753,
+      "step": 35592
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.77567855744597e-06,
+      "loss": 0.6192,
+      "step": 35593
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7752730745007086e-06,
+      "loss": 0.628,
+      "step": 35594
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7748676333476921e-06,
+      "loss": 0.5773,
+      "step": 35595
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7744622339889828e-06,
+      "loss": 0.5933,
+      "step": 35596
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7740568764266365e-06,
+      "loss": 0.5726,
+      "step": 35597
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7736515606627202e-06,
+      "loss": 0.6049,
+      "step": 35598
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.773246286699286e-06,
+      "loss": 0.6563,
+      "step": 35599
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7728410545384024e-06,
+      "loss": 0.649,
+      "step": 35600
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7724358641821172e-06,
+      "loss": 0.6192,
+      "step": 35601
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7720307156324967e-06,
+      "loss": 0.61,
+      "step": 35602
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7716256088915973e-06,
+      "loss": 0.6609,
+      "step": 35603
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7712205439614793e-06,
+      "loss": 0.5526,
+      "step": 35604
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.770815520844199e-06,
+      "loss": 0.5801,
+      "step": 35605
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7704105395418159e-06,
+      "loss": 0.6043,
+      "step": 35606
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7700056000563848e-06,
+      "loss": 0.5411,
+      "step": 35607
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7696007023899676e-06,
+      "loss": 0.5916,
+      "step": 35608
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7691958465446169e-06,
+      "loss": 0.6703,
+      "step": 35609
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7687910325223957e-06,
+      "loss": 0.5975,
+      "step": 35610
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7683862603253576e-06,
+      "loss": 0.5836,
+      "step": 35611
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7679815299555591e-06,
+      "loss": 0.6107,
+      "step": 35612
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7675768414150573e-06,
+      "loss": 0.563,
+      "step": 35613
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7671721947059072e-06,
+      "loss": 0.62,
+      "step": 35614
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7667675898301683e-06,
+      "loss": 0.5438,
+      "step": 35615
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7663630267898945e-06,
+      "loss": 0.6401,
+      "step": 35616
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7659585055871408e-06,
+      "loss": 0.5834,
+      "step": 35617
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7655540262239612e-06,
+      "loss": 0.5394,
+      "step": 35618
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7651495887024162e-06,
+      "loss": 0.5768,
+      "step": 35619
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7647451930245552e-06,
+      "loss": 0.6203,
+      "step": 35620
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.76434083919244e-06,
+      "loss": 0.5753,
+      "step": 35621
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7639365272081166e-06,
+      "loss": 0.5917,
+      "step": 35622
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7635322570736456e-06,
+      "loss": 0.5937,
+      "step": 35623
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7631280287910779e-06,
+      "loss": 0.6552,
+      "step": 35624
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7627238423624703e-06,
+      "loss": 0.6558,
+      "step": 35625
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7623196977898737e-06,
+      "loss": 0.6389,
+      "step": 35626
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7619155950753476e-06,
+      "loss": 0.6334,
+      "step": 35627
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7615115342209366e-06,
+      "loss": 0.569,
+      "step": 35628
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7611075152286993e-06,
+      "loss": 0.6308,
+      "step": 35629
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7607035381006864e-06,
+      "loss": 0.572,
+      "step": 35630
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.760299602838954e-06,
+      "loss": 0.6289,
+      "step": 35631
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.7598957094455528e-06,
+      "loss": 0.6219,
+      "step": 35632
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.759491857922534e-06,
+      "loss": 0.5709,
+      "step": 35633
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7590880482719508e-06,
+      "loss": 0.577,
+      "step": 35634
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7586842804958526e-06,
+      "loss": 0.5937,
+      "step": 35635
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7582805545962955e-06,
+      "loss": 0.6102,
+      "step": 35636
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7578768705753258e-06,
+      "loss": 0.5725,
+      "step": 35637
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7574732284350026e-06,
+      "loss": 0.5908,
+      "step": 35638
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.757069628177367e-06,
+      "loss": 0.6149,
+      "step": 35639
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7566660698044768e-06,
+      "loss": 0.6565,
+      "step": 35640
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7562625533183785e-06,
+      "loss": 0.6327,
+      "step": 35641
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.755859078721126e-06,
+      "loss": 0.5669,
+      "step": 35642
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7554556460147687e-06,
+      "loss": 0.6322,
+      "step": 35643
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.755052255201356e-06,
+      "loss": 0.5679,
+      "step": 35644
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7546489062829343e-06,
+      "loss": 0.6046,
+      "step": 35645
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7542455992615593e-06,
+      "loss": 0.63,
+      "step": 35646
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7538423341392752e-06,
+      "loss": 0.5366,
+      "step": 35647
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7534391109181348e-06,
+      "loss": 0.5492,
+      "step": 35648
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7530359296001853e-06,
+      "loss": 0.6363,
+      "step": 35649
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7526327901874762e-06,
+      "loss": 0.6514,
+      "step": 35650
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7522296926820526e-06,
+      "loss": 0.5865,
+      "step": 35651
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7518266370859671e-06,
+      "loss": 0.611,
+      "step": 35652
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7514236234012672e-06,
+      "loss": 0.6143,
+      "step": 35653
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7510206516299989e-06,
+      "loss": 0.5389,
+      "step": 35654
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7506177217742104e-06,
+      "loss": 0.5732,
+      "step": 35655
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.750214833835947e-06,
+      "loss": 0.5417,
+      "step": 35656
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7498119878172603e-06,
+      "loss": 0.5962,
+      "step": 35657
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7494091837201933e-06,
+      "loss": 0.6301,
+      "step": 35658
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7490064215467983e-06,
+      "loss": 0.5888,
+      "step": 35659
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.748603701299114e-06,
+      "loss": 0.5609,
+      "step": 35660
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7482010229791935e-06,
+      "loss": 0.5821,
+      "step": 35661
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.747798386589078e-06,
+      "loss": 0.5632,
+      "step": 35662
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.747395792130817e-06,
+      "loss": 0.5935,
+      "step": 35663
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7469932396064526e-06,
+      "loss": 0.6252,
+      "step": 35664
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7465907290180385e-06,
+      "loss": 0.5566,
+      "step": 35665
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7461882603676084e-06,
+      "loss": 0.5967,
+      "step": 35666
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7457858336572152e-06,
+      "loss": 0.6071,
+      "step": 35667
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7453834488888998e-06,
+      "loss": 0.6036,
+      "step": 35668
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7449811060647104e-06,
+      "loss": 0.6208,
+      "step": 35669
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7445788051866897e-06,
+      "loss": 0.561,
+      "step": 35670
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7441765462568816e-06,
+      "loss": 0.6516,
+      "step": 35671
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.743774329277328e-06,
+      "loss": 0.6033,
+      "step": 35672
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7433721542500781e-06,
+      "loss": 0.6339,
+      "step": 35673
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7429700211771716e-06,
+      "loss": 0.5856,
+      "step": 35674
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.74256793006065e-06,
+      "loss": 0.6217,
+      "step": 35675
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7421658809025643e-06,
+      "loss": 0.5037,
+      "step": 35676
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.741763873704948e-06,
+      "loss": 0.5821,
+      "step": 35677
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7413619084698497e-06,
+      "loss": 0.5758,
+      "step": 35678
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7409599851993076e-06,
+      "loss": 0.6264,
+      "step": 35679
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.740558103895369e-06,
+      "loss": 0.5818,
+      "step": 35680
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7401562645600745e-06,
+      "loss": 0.6014,
+      "step": 35681
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7397544671954636e-06,
+      "loss": 0.6499,
+      "step": 35682
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.739352711803578e-06,
+      "loss": 0.6075,
+      "step": 35683
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7389509983864627e-06,
+      "loss": 0.605,
+      "step": 35684
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.738549326946154e-06,
+      "loss": 0.5618,
+      "step": 35685
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.738147697484701e-06,
+      "loss": 0.5985,
+      "step": 35686
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7377461100041338e-06,
+      "loss": 0.5937,
+      "step": 35687
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7373445645065013e-06,
+      "loss": 0.564,
+      "step": 35688
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7369430609938375e-06,
+      "loss": 0.582,
+      "step": 35689
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7365415994681888e-06,
+      "loss": 0.6075,
+      "step": 35690
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7361401799315935e-06,
+      "loss": 0.6373,
+      "step": 35691
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7357388023860887e-06,
+      "loss": 0.6198,
+      "step": 35692
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7353374668337164e-06,
+      "loss": 0.5886,
+      "step": 35693
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7349361732765113e-06,
+      "loss": 0.559,
+      "step": 35694
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7345349217165185e-06,
+      "loss": 0.5608,
+      "step": 35695
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7341337121557733e-06,
+      "loss": 0.5806,
+      "step": 35696
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7337325445963194e-06,
+      "loss": 0.584,
+      "step": 35697
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7333314190401861e-06,
+      "loss": 0.5858,
+      "step": 35698
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.73293033548942e-06,
+      "loss": 0.5532,
+      "step": 35699
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7325292939460536e-06,
+      "loss": 0.6134,
+      "step": 35700
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7321282944121288e-06,
+      "loss": 0.5829,
+      "step": 35701
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7317273368896803e-06,
+      "loss": 0.6331,
+      "step": 35702
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.73132642138075e-06,
+      "loss": 0.592,
+      "step": 35703
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7309255478873676e-06,
+      "loss": 0.612,
+      "step": 35704
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7305247164115768e-06,
+      "loss": 0.6178,
+      "step": 35705
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7301239269554093e-06,
+      "loss": 0.5367,
+      "step": 35706
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7297231795209069e-06,
+      "loss": 0.6127,
+      "step": 35707
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7293224741101023e-06,
+      "loss": 0.6311,
+      "step": 35708
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7289218107250327e-06,
+      "loss": 0.5852,
+      "step": 35709
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7285211893677312e-06,
+      "loss": 0.6232,
+      "step": 35710
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7281206100402381e-06,
+      "loss": 0.5629,
+      "step": 35711
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7277200727445874e-06,
+      "loss": 0.5251,
+      "step": 35712
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7273195774828134e-06,
+      "loss": 0.5333,
+      "step": 35713
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7269191242569515e-06,
+      "loss": 0.618,
+      "step": 35714
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.726518713069033e-06,
+      "loss": 0.605,
+      "step": 35715
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7261183439210994e-06,
+      "loss": 0.6438,
+      "step": 35716
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7257180168151788e-06,
+      "loss": 0.6307,
+      "step": 35717
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.725317731753311e-06,
+      "loss": 0.6489,
+      "step": 35718
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7249174887375263e-06,
+      "loss": 0.6021,
+      "step": 35719
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.72451728776986e-06,
+      "loss": 0.5856,
+      "step": 35720
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7241171288523428e-06,
+      "loss": 0.6347,
+      "step": 35721
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7237170119870129e-06,
+      "loss": 0.5803,
+      "step": 35722
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7233169371758974e-06,
+      "loss": 0.6087,
+      "step": 35723
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.722916904421037e-06,
+      "loss": 0.5909,
+      "step": 35724
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7225169137244557e-06,
+      "loss": 0.5957,
+      "step": 35725
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.722116965088192e-06,
+      "loss": 0.609,
+      "step": 35726
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.721717058514274e-06,
+      "loss": 0.615,
+      "step": 35727
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.721317194004739e-06,
+      "loss": 0.6385,
+      "step": 35728
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7209173715616156e-06,
+      "loss": 0.6353,
+      "step": 35729
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7205175911869354e-06,
+      "loss": 0.5764,
+      "step": 35730
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7201178528827278e-06,
+      "loss": 0.6102,
+      "step": 35731
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7197181566510291e-06,
+      "loss": 0.5648,
+      "step": 35732
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7193185024938674e-06,
+      "loss": 0.6226,
+      "step": 35733
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7189188904132702e-06,
+      "loss": 0.5639,
+      "step": 35734
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7185193204112772e-06,
+      "loss": 0.6076,
+      "step": 35735
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7181197924899073e-06,
+      "loss": 0.6401,
+      "step": 35736
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7177203066511994e-06,
+      "loss": 0.6034,
+      "step": 35737
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7173208628971772e-06,
+      "loss": 0.5471,
+      "step": 35738
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7169214612298757e-06,
+      "loss": 0.587,
+      "step": 35739
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.71652210165132e-06,
+      "loss": 0.6565,
+      "step": 35740
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7161227841635464e-06,
+      "loss": 0.5975,
+      "step": 35741
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7157235087685743e-06,
+      "loss": 0.6429,
+      "step": 35742
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7153242754684384e-06,
+      "loss": 0.5809,
+      "step": 35743
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7149250842651643e-06,
+      "loss": 0.625,
+      "step": 35744
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7145259351607857e-06,
+      "loss": 0.5627,
+      "step": 35745
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7141268281573264e-06,
+      "loss": 0.5874,
+      "step": 35746
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.713727763256815e-06,
+      "loss": 0.5685,
+      "step": 35747
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7133287404612775e-06,
+      "loss": 0.6013,
+      "step": 35748
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7129297597727469e-06,
+      "loss": 0.5908,
+      "step": 35749
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7125308211932457e-06,
+      "loss": 0.6048,
+      "step": 35750
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7121319247248036e-06,
+      "loss": 0.5875,
+      "step": 35751
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7117330703694435e-06,
+      "loss": 0.5651,
+      "step": 35752
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7113342581291969e-06,
+      "loss": 0.5696,
+      "step": 35753
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.710935488006089e-06,
+      "loss": 0.6263,
+      "step": 35754
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7105367600021427e-06,
+      "loss": 0.6003,
+      "step": 35755
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7101380741193895e-06,
+      "loss": 0.6308,
+      "step": 35756
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7097394303598535e-06,
+      "loss": 0.6048,
+      "step": 35757
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7093408287255576e-06,
+      "loss": 0.5604,
+      "step": 35758
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7089422692185275e-06,
+      "loss": 0.6236,
+      "step": 35759
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.708543751840792e-06,
+      "loss": 0.6356,
+      "step": 35760
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7081452765943718e-06,
+      "loss": 0.5738,
+      "step": 35761
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7077468434812993e-06,
+      "loss": 0.6061,
+      "step": 35762
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7073484525035877e-06,
+      "loss": 0.5789,
+      "step": 35763
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7069501036632697e-06,
+      "loss": 0.6306,
+      "step": 35764
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7065517969623647e-06,
+      "loss": 0.6122,
+      "step": 35765
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.706153532402901e-06,
+      "loss": 0.5735,
+      "step": 35766
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7057553099869006e-06,
+      "loss": 0.5936,
+      "step": 35767
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7053571297163873e-06,
+      "loss": 0.6091,
+      "step": 35768
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7049589915933807e-06,
+      "loss": 0.5953,
+      "step": 35769
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.704560895619909e-06,
+      "loss": 0.5941,
+      "step": 35770
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7041628417979916e-06,
+      "loss": 0.5877,
+      "step": 35771
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7037648301296517e-06,
+      "loss": 0.5736,
+      "step": 35772
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7033668606169163e-06,
+      "loss": 0.5677,
+      "step": 35773
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7029689332617994e-06,
+      "loss": 0.5814,
+      "step": 35774
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7025710480663293e-06,
+      "loss": 0.5575,
+      "step": 35775
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7021732050325234e-06,
+      "loss": 0.646,
+      "step": 35776
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7017754041624079e-06,
+      "loss": 0.6533,
+      "step": 35777
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7013776454580022e-06,
+      "loss": 0.6293,
+      "step": 35778
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7009799289213257e-06,
+      "loss": 0.6257,
+      "step": 35779
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.7005822545544005e-06,
+      "loss": 0.569,
+      "step": 35780
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.7001846223592478e-06,
+      "loss": 0.6052,
+      "step": 35781
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6997870323378863e-06,
+      "loss": 0.6879,
+      "step": 35782
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6993894844923398e-06,
+      "loss": 0.6044,
+      "step": 35783
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6989919788246267e-06,
+      "loss": 0.6052,
+      "step": 35784
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6985945153367666e-06,
+      "loss": 0.6167,
+      "step": 35785
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6981970940307756e-06,
+      "loss": 0.5814,
+      "step": 35786
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.69779971490868e-06,
+      "loss": 0.6817,
+      "step": 35787
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6974023779724947e-06,
+      "loss": 0.6383,
+      "step": 35788
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.697005083224239e-06,
+      "loss": 0.5943,
+      "step": 35789
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6966078306659305e-06,
+      "loss": 0.6764,
+      "step": 35790
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.696210620299592e-06,
+      "loss": 0.6284,
+      "step": 35791
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6958134521272386e-06,
+      "loss": 0.665,
+      "step": 35792
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.695416326150887e-06,
+      "loss": 0.5854,
+      "step": 35793
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6950192423725598e-06,
+      "loss": 0.6174,
+      "step": 35794
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6946222007942714e-06,
+      "loss": 0.603,
+      "step": 35795
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6942252014180405e-06,
+      "loss": 0.6055,
+      "step": 35796
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.693828244245881e-06,
+      "loss": 0.6498,
+      "step": 35797
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6934313292798154e-06,
+      "loss": 0.6069,
+      "step": 35798
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6930344565218549e-06,
+      "loss": 0.5693,
+      "step": 35799
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.692637625974024e-06,
+      "loss": 0.5405,
+      "step": 35800
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6922408376383303e-06,
+      "loss": 0.6194,
+      "step": 35801
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6918440915167956e-06,
+      "loss": 0.5779,
+      "step": 35802
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6914473876114323e-06,
+      "loss": 0.6582,
+      "step": 35803
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6910507259242604e-06,
+      "loss": 0.6441,
+      "step": 35804
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6906541064572935e-06,
+      "loss": 0.6368,
+      "step": 35805
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6902575292125457e-06,
+      "loss": 0.6244,
+      "step": 35806
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6898609941920318e-06,
+      "loss": 0.5897,
+      "step": 35807
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6894645013977696e-06,
+      "loss": 0.6393,
+      "step": 35808
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6890680508317713e-06,
+      "loss": 0.6082,
+      "step": 35809
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6886716424960514e-06,
+      "loss": 0.5621,
+      "step": 35810
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.688275276392627e-06,
+      "loss": 0.6146,
+      "step": 35811
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6878789525235096e-06,
+      "loss": 0.5993,
+      "step": 35812
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6874826708907133e-06,
+      "loss": 0.6968,
+      "step": 35813
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6870864314962498e-06,
+      "loss": 0.6258,
+      "step": 35814
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6866902343421376e-06,
+      "loss": 0.5662,
+      "step": 35815
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6862940794303862e-06,
+      "loss": 0.6104,
+      "step": 35816
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6858979667630105e-06,
+      "loss": 0.5882,
+      "step": 35817
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6855018963420189e-06,
+      "loss": 0.5844,
+      "step": 35818
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.68510586816943e-06,
+      "loss": 0.635,
+      "step": 35819
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6847098822472496e-06,
+      "loss": 0.5973,
+      "step": 35820
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6843139385774976e-06,
+      "loss": 0.5697,
+      "step": 35821
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.683918037162181e-06,
+      "loss": 0.638,
+      "step": 35822
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6835221780033128e-06,
+      "loss": 0.6155,
+      "step": 35823
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6831263611029014e-06,
+      "loss": 0.6088,
+      "step": 35824
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6827305864629628e-06,
+      "loss": 0.6673,
+      "step": 35825
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6823348540855067e-06,
+      "loss": 0.6215,
+      "step": 35826
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6819391639725425e-06,
+      "loss": 0.6274,
+      "step": 35827
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6815435161260785e-06,
+      "loss": 0.6228,
+      "step": 35828
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.681147910548131e-06,
+      "loss": 0.6593,
+      "step": 35829
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6807523472407084e-06,
+      "loss": 0.5792,
+      "step": 35830
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6803568262058156e-06,
+      "loss": 0.5889,
+      "step": 35831
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6799613474454691e-06,
+      "loss": 0.6304,
+      "step": 35832
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6795659109616757e-06,
+      "loss": 0.5845,
+      "step": 35833
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6791705167564442e-06,
+      "loss": 0.5885,
+      "step": 35834
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.678775164831783e-06,
+      "loss": 0.6043,
+      "step": 35835
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6783798551897034e-06,
+      "loss": 0.6023,
+      "step": 35836
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6779845878322109e-06,
+      "loss": 0.592,
+      "step": 35837
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6775893627613204e-06,
+      "loss": 0.5098,
+      "step": 35838
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6771941799790313e-06,
+      "loss": 0.5428,
+      "step": 35839
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.676799039487359e-06,
+      "loss": 0.5686,
+      "step": 35840
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6764039412883048e-06,
+      "loss": 0.6028,
+      "step": 35841
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.676008885383883e-06,
+      "loss": 0.6038,
+      "step": 35842
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6756138717760972e-06,
+      "loss": 0.6151,
+      "step": 35843
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6752189004669562e-06,
+      "loss": 0.6133,
+      "step": 35844
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6748239714584624e-06,
+      "loss": 0.5687,
+      "step": 35845
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.674429084752629e-06,
+      "loss": 0.5385,
+      "step": 35846
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6740342403514597e-06,
+      "loss": 0.5747,
+      "step": 35847
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6736394382569587e-06,
+      "loss": 0.5936,
+      "step": 35848
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6732446784711364e-06,
+      "loss": 0.5666,
+      "step": 35849
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6728499609959969e-06,
+      "loss": 0.5747,
+      "step": 35850
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6724552858335452e-06,
+      "loss": 0.5801,
+      "step": 35851
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6720606529857842e-06,
+      "loss": 0.6257,
+      "step": 35852
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6716660624547243e-06,
+      "loss": 0.6438,
+      "step": 35853
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6712715142423685e-06,
+      "loss": 0.6408,
+      "step": 35854
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.670877008350722e-06,
+      "loss": 0.5991,
+      "step": 35855
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.670482544781785e-06,
+      "loss": 0.6026,
+      "step": 35856
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6700881235375687e-06,
+      "loss": 0.5764,
+      "step": 35857
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6696937446200722e-06,
+      "loss": 0.5621,
+      "step": 35858
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.669299408031303e-06,
+      "loss": 0.5953,
+      "step": 35859
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6689051137732636e-06,
+      "loss": 0.5582,
+      "step": 35860
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6685108618479574e-06,
+      "loss": 0.6117,
+      "step": 35861
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6681166522573856e-06,
+      "loss": 0.6333,
+      "step": 35862
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6677224850035544e-06,
+      "loss": 0.6301,
+      "step": 35863
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6673283600884638e-06,
+      "loss": 0.5887,
+      "step": 35864
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6669342775141228e-06,
+      "loss": 0.6161,
+      "step": 35865
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6665402372825247e-06,
+      "loss": 0.5953,
+      "step": 35866
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6661462393956774e-06,
+      "loss": 0.6224,
+      "step": 35867
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.665752283855583e-06,
+      "loss": 0.6558,
+      "step": 35868
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6653583706642396e-06,
+      "loss": 0.625,
+      "step": 35869
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6649644998236525e-06,
+      "loss": 0.5959,
+      "step": 35870
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6645706713358222e-06,
+      "loss": 0.6009,
+      "step": 35871
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6641768852027507e-06,
+      "loss": 0.6408,
+      "step": 35872
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6637831414264339e-06,
+      "loss": 0.5994,
+      "step": 35873
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6633894400088801e-06,
+      "loss": 0.5807,
+      "step": 35874
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6629957809520825e-06,
+      "loss": 0.5748,
+      "step": 35875
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6626021642580503e-06,
+      "loss": 0.5388,
+      "step": 35876
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6622085899287732e-06,
+      "loss": 0.6085,
+      "step": 35877
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6618150579662583e-06,
+      "loss": 0.6379,
+      "step": 35878
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.661421568372501e-06,
+      "loss": 0.5921,
+      "step": 35879
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.661028121149505e-06,
+      "loss": 0.5949,
+      "step": 35880
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6606347162992665e-06,
+      "loss": 0.633,
+      "step": 35881
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6602413538237861e-06,
+      "loss": 0.5693,
+      "step": 35882
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.65984803372506e-06,
+      "loss": 0.6605,
+      "step": 35883
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.65945475600509e-06,
+      "loss": 0.6338,
+      "step": 35884
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6590615206658732e-06,
+      "loss": 0.6034,
+      "step": 35885
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6586683277094052e-06,
+      "loss": 0.6039,
+      "step": 35886
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6582751771376894e-06,
+      "loss": 0.6047,
+      "step": 35887
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6578820689527198e-06,
+      "loss": 0.6116,
+      "step": 35888
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6574890031564949e-06,
+      "loss": 0.5273,
+      "step": 35889
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.657095979751009e-06,
+      "loss": 0.6639,
+      "step": 35890
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6567029987382631e-06,
+      "loss": 0.5549,
+      "step": 35891
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6563100601202541e-06,
+      "loss": 0.5683,
+      "step": 35892
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6559171638989768e-06,
+      "loss": 0.5976,
+      "step": 35893
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6555243100764252e-06,
+      "loss": 0.6261,
+      "step": 35894
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6551314986546018e-06,
+      "loss": 0.6083,
+      "step": 35895
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6547387296354956e-06,
+      "loss": 0.5638,
+      "step": 35896
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6543460030211122e-06,
+      "loss": 0.6105,
+      "step": 35897
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.653953318813435e-06,
+      "loss": 0.5636,
+      "step": 35898
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6535606770144674e-06,
+      "loss": 0.581,
+      "step": 35899
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6531680776262005e-06,
+      "loss": 0.5906,
+      "step": 35900
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6527755206506335e-06,
+      "loss": 0.6151,
+      "step": 35901
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.652383006089756e-06,
+      "loss": 0.6022,
+      "step": 35902
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6519905339455711e-06,
+      "loss": 0.5959,
+      "step": 35903
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6515981042200614e-06,
+      "loss": 0.5857,
+      "step": 35904
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6512057169152284e-06,
+      "loss": 0.5808,
+      "step": 35905
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6508133720330654e-06,
+      "loss": 0.6424,
+      "step": 35906
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6504210695755618e-06,
+      "loss": 0.5719,
+      "step": 35907
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.650028809544716e-06,
+      "loss": 0.6062,
+      "step": 35908
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6496365919425195e-06,
+      "loss": 0.6158,
+      "step": 35909
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6492444167709654e-06,
+      "loss": 0.5886,
+      "step": 35910
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6488522840320442e-06,
+      "loss": 0.5952,
+      "step": 35911
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6484601937277511e-06,
+      "loss": 0.6196,
+      "step": 35912
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6480681458600767e-06,
+      "loss": 0.5581,
+      "step": 35913
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6476761404310182e-06,
+      "loss": 0.6113,
+      "step": 35914
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6472841774425586e-06,
+      "loss": 0.6467,
+      "step": 35915
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6468922568966961e-06,
+      "loss": 0.6025,
+      "step": 35916
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6465003787954193e-06,
+      "loss": 0.6283,
+      "step": 35917
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.646108543140722e-06,
+      "loss": 0.6069,
+      "step": 35918
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6457167499345938e-06,
+      "loss": 0.6321,
+      "step": 35919
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6453249991790254e-06,
+      "loss": 0.6039,
+      "step": 35920
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6449332908760053e-06,
+      "loss": 0.6437,
+      "step": 35921
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6445416250275281e-06,
+      "loss": 0.5778,
+      "step": 35922
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6441500016355793e-06,
+      "loss": 0.5853,
+      "step": 35923
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6437584207021551e-06,
+      "loss": 0.6092,
+      "step": 35924
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6433668822292403e-06,
+      "loss": 0.5915,
+      "step": 35925
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.642975386218827e-06,
+      "loss": 0.5765,
+      "step": 35926
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.6425839326729021e-06,
+      "loss": 0.6245,
+      "step": 35927
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.642192521593453e-06,
+      "loss": 0.618,
+      "step": 35928
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.641801152982475e-06,
+      "loss": 0.5703,
+      "step": 35929
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6414098268419532e-06,
+      "loss": 0.5991,
+      "step": 35930
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6410185431738756e-06,
+      "loss": 0.6519,
+      "step": 35931
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6406273019802277e-06,
+      "loss": 0.6457,
+      "step": 35932
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6402361032630043e-06,
+      "loss": 0.5781,
+      "step": 35933
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6398449470241862e-06,
+      "loss": 0.6279,
+      "step": 35934
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6394538332657694e-06,
+      "loss": 0.5991,
+      "step": 35935
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6390627619897316e-06,
+      "loss": 0.6162,
+      "step": 35936
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.638671733198066e-06,
+      "loss": 0.5939,
+      "step": 35937
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6382807468927576e-06,
+      "loss": 0.5539,
+      "step": 35938
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.637889803075794e-06,
+      "loss": 0.6349,
+      "step": 35939
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6374989017491593e-06,
+      "loss": 0.6608,
+      "step": 35940
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6371080429148478e-06,
+      "loss": 0.579,
+      "step": 35941
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6367172265748331e-06,
+      "loss": 0.5438,
+      "step": 35942
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6363264527311107e-06,
+      "loss": 0.6637,
+      "step": 35943
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6359357213856608e-06,
+      "loss": 0.5602,
+      "step": 35944
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6355450325404732e-06,
+      "loss": 0.5638,
+      "step": 35945
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6351543861975306e-06,
+      "loss": 0.5653,
+      "step": 35946
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.634763782358818e-06,
+      "loss": 0.5605,
+      "step": 35947
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6343732210263219e-06,
+      "loss": 0.615,
+      "step": 35948
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6339827022020217e-06,
+      "loss": 0.6456,
+      "step": 35949
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6335922258879088e-06,
+      "loss": 0.622,
+      "step": 35950
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6332017920859612e-06,
+      "loss": 0.5768,
+      "step": 35951
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6328114007981699e-06,
+      "loss": 0.5769,
+      "step": 35952
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6324210520265104e-06,
+      "loss": 0.593,
+      "step": 35953
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6320307457729722e-06,
+      "loss": 0.5667,
+      "step": 35954
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6316404820395336e-06,
+      "loss": 0.6129,
+      "step": 35955
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6312502608281832e-06,
+      "loss": 0.6045,
+      "step": 35956
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6308600821409016e-06,
+      "loss": 0.6074,
+      "step": 35957
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6304699459796703e-06,
+      "loss": 0.5832,
+      "step": 35958
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.63007985234647e-06,
+      "loss": 0.6106,
+      "step": 35959
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6296898012432882e-06,
+      "loss": 0.5761,
+      "step": 35960
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.629299792672101e-06,
+      "loss": 0.6245,
+      "step": 35961
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.628909826634898e-06,
+      "loss": 0.5732,
+      "step": 35962
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6285199031336506e-06,
+      "loss": 0.611,
+      "step": 35963
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.628130022170349e-06,
+      "loss": 0.6965,
+      "step": 35964
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6277401837469686e-06,
+      "loss": 0.5368,
+      "step": 35965
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6273503878654906e-06,
+      "loss": 0.6192,
+      "step": 35966
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.626960634527901e-06,
+      "loss": 0.6249,
+      "step": 35967
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6265709237361749e-06,
+      "loss": 0.5863,
+      "step": 35968
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6261812554922952e-06,
+      "loss": 0.631,
+      "step": 35969
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.625791629798238e-06,
+      "loss": 0.6351,
+      "step": 35970
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6254020466559884e-06,
+      "loss": 0.6028,
+      "step": 35971
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6250125060675214e-06,
+      "loss": 0.6422,
+      "step": 35972
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6246230080348235e-06,
+      "loss": 0.5686,
+      "step": 35973
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.624233552559864e-06,
+      "loss": 0.5889,
+      "step": 35974
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.623844139644629e-06,
+      "loss": 0.5859,
+      "step": 35975
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.623454769291093e-06,
+      "loss": 0.5739,
+      "step": 35976
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6230654415012382e-06,
+      "loss": 0.571,
+      "step": 35977
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6226761562770387e-06,
+      "loss": 0.5657,
+      "step": 35978
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.622286913620481e-06,
+      "loss": 0.5696,
+      "step": 35979
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6218977135335312e-06,
+      "loss": 0.6184,
+      "step": 35980
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6215085560181764e-06,
+      "loss": 0.637,
+      "step": 35981
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6211194410763864e-06,
+      "loss": 0.6408,
+      "step": 35982
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6207303687101461e-06,
+      "loss": 0.6021,
+      "step": 35983
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6203413389214285e-06,
+      "loss": 0.5408,
+      "step": 35984
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6199523517122094e-06,
+      "loss": 0.5668,
+      "step": 35985
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6195634070844679e-06,
+      "loss": 0.5696,
+      "step": 35986
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6191745050401763e-06,
+      "loss": 0.5777,
+      "step": 35987
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6187856455813166e-06,
+      "loss": 0.5839,
+      "step": 35988
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6183968287098606e-06,
+      "loss": 0.5767,
+      "step": 35989
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6180080544277853e-06,
+      "loss": 0.5451,
+      "step": 35990
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.617619322737064e-06,
+      "loss": 0.6219,
+      "step": 35991
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6172306336396748e-06,
+      "loss": 0.5549,
+      "step": 35992
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6168419871375906e-06,
+      "loss": 0.6065,
+      "step": 35993
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.61645338323279e-06,
+      "loss": 0.5529,
+      "step": 35994
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6160648219272434e-06,
+      "loss": 0.5528,
+      "step": 35995
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6156763032229284e-06,
+      "loss": 0.5923,
+      "step": 35996
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6152878271218141e-06,
+      "loss": 0.5666,
+      "step": 35997
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6148993936258818e-06,
+      "loss": 0.6039,
+      "step": 35998
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.614511002737097e-06,
+      "loss": 0.5838,
+      "step": 35999
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.614122654457444e-06,
+      "loss": 0.5764,
+      "step": 36000
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6137343487888845e-06,
+      "loss": 0.5994,
+      "step": 36001
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6133460857333983e-06,
+      "loss": 0.5489,
+      "step": 36002
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6129578652929557e-06,
+      "loss": 0.552,
+      "step": 36003
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6125696874695318e-06,
+      "loss": 0.6399,
+      "step": 36004
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6121815522650984e-06,
+      "loss": 0.5921,
+      "step": 36005
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6117934596816265e-06,
+      "loss": 0.5515,
+      "step": 36006
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6114054097210884e-06,
+      "loss": 0.5965,
+      "step": 36007
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6110174023854552e-06,
+      "loss": 0.5967,
+      "step": 36008
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6106294376767006e-06,
+      "loss": 0.6263,
+      "step": 36009
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.610241515596792e-06,
+      "loss": 0.6131,
+      "step": 36010
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.609853636147709e-06,
+      "loss": 0.5489,
+      "step": 36011
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.609465799331411e-06,
+      "loss": 0.6311,
+      "step": 36012
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6090780051498788e-06,
+      "loss": 0.5863,
+      "step": 36013
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6086902536050742e-06,
+      "loss": 0.5908,
+      "step": 36014
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6083025446989752e-06,
+      "loss": 0.6199,
+      "step": 36015
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6079148784335463e-06,
+      "loss": 0.6057,
+      "step": 36016
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6075272548107645e-06,
+      "loss": 0.6356,
+      "step": 36017
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6071396738325894e-06,
+      "loss": 0.587,
+      "step": 36018
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6067521355009985e-06,
+      "loss": 0.5993,
+      "step": 36019
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6063646398179555e-06,
+      "loss": 0.5848,
+      "step": 36020
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6059771867854346e-06,
+      "loss": 0.537,
+      "step": 36021
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6055897764054017e-06,
+      "loss": 0.5522,
+      "step": 36022
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6052024086798256e-06,
+      "loss": 0.6243,
+      "step": 36023
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6048150836106712e-06,
+      "loss": 0.5936,
+      "step": 36024
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6044278011999137e-06,
+      "loss": 0.6033,
+      "step": 36025
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6040405614495169e-06,
+      "loss": 0.6096,
+      "step": 36026
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6036533643614493e-06,
+      "loss": 0.5876,
+      "step": 36027
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6032662099376772e-06,
+      "loss": 0.6338,
+      "step": 36028
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6028790981801668e-06,
+      "loss": 0.58,
+      "step": 36029
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6024920290908884e-06,
+      "loss": 0.5528,
+      "step": 36030
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6021050026718056e-06,
+      "loss": 0.6377,
+      "step": 36031
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6017180189248892e-06,
+      "loss": 0.6057,
+      "step": 36032
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6013310778521029e-06,
+      "loss": 0.6418,
+      "step": 36033
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6009441794554126e-06,
+      "loss": 0.6273,
+      "step": 36034
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6005573237367832e-06,
+      "loss": 0.6215,
+      "step": 36035
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.6001705106981835e-06,
+      "loss": 0.6135,
+      "step": 36036
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5997837403415762e-06,
+      "loss": 0.5388,
+      "step": 36037
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5993970126689318e-06,
+      "loss": 0.62,
+      "step": 36038
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5990103276822066e-06,
+      "loss": 0.584,
+      "step": 36039
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5986236853833725e-06,
+      "loss": 0.6451,
+      "step": 36040
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.59823708577439e-06,
+      "loss": 0.6086,
+      "step": 36041
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5978505288572267e-06,
+      "loss": 0.6283,
+      "step": 36042
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.597464014633847e-06,
+      "loss": 0.6228,
+      "step": 36043
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5970775431062125e-06,
+      "loss": 0.6393,
+      "step": 36044
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5966911142762863e-06,
+      "loss": 0.5742,
+      "step": 36045
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5963047281460353e-06,
+      "loss": 0.6134,
+      "step": 36046
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.595918384717422e-06,
+      "loss": 0.6236,
+      "step": 36047
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5955320839924048e-06,
+      "loss": 0.6397,
+      "step": 36048
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5951458259729568e-06,
+      "loss": 0.5789,
+      "step": 36049
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5947596106610285e-06,
+      "loss": 0.62,
+      "step": 36050
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5943734380585907e-06,
+      "loss": 0.5858,
+      "step": 36051
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5939873081676005e-06,
+      "loss": 0.6149,
+      "step": 36052
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5936012209900264e-06,
+      "loss": 0.6117,
+      "step": 36053
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5932151765278248e-06,
+      "loss": 0.6226,
+      "step": 36054
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5928291747829593e-06,
+      "loss": 0.5447,
+      "step": 36055
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.59244321575739e-06,
+      "loss": 0.6136,
+      "step": 36056
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5920572994530792e-06,
+      "loss": 0.5939,
+      "step": 36057
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5916714258719867e-06,
+      "loss": 0.6722,
+      "step": 36058
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5912855950160765e-06,
+      "loss": 0.519,
+      "step": 36059
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.590899806887306e-06,
+      "loss": 0.5969,
+      "step": 36060
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5905140614876368e-06,
+      "loss": 0.562,
+      "step": 36061
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5901283588190264e-06,
+      "loss": 0.6294,
+      "step": 36062
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5897426988834385e-06,
+      "loss": 0.5672,
+      "step": 36063
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5893570816828308e-06,
+      "loss": 0.6252,
+      "step": 36064
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5889715072191637e-06,
+      "loss": 0.5835,
+      "step": 36065
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5885859754943955e-06,
+      "loss": 0.6358,
+      "step": 36066
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5882004865104828e-06,
+      "loss": 0.5872,
+      "step": 36067
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5878150402693893e-06,
+      "loss": 0.6243,
+      "step": 36068
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5874296367730692e-06,
+      "loss": 0.5592,
+      "step": 36069
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.587044276023485e-06,
+      "loss": 0.5998,
+      "step": 36070
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5866589580225922e-06,
+      "loss": 0.6319,
+      "step": 36071
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5862736827723503e-06,
+      "loss": 0.5595,
+      "step": 36072
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5858884502747118e-06,
+      "loss": 0.6016,
+      "step": 36073
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5855032605316422e-06,
+      "loss": 0.6151,
+      "step": 36074
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5851181135450922e-06,
+      "loss": 0.6371,
+      "step": 36075
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5847330093170254e-06,
+      "loss": 0.5858,
+      "step": 36076
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5843479478493907e-06,
+      "loss": 0.6422,
+      "step": 36077
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5839629291441505e-06,
+      "loss": 0.5668,
+      "step": 36078
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.583577953203258e-06,
+      "loss": 0.6139,
+      "step": 36079
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5831930200286727e-06,
+      "loss": 0.6212,
+      "step": 36080
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5828081296223486e-06,
+      "loss": 0.6231,
+      "step": 36081
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5824232819862406e-06,
+      "loss": 0.6278,
+      "step": 36082
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5820384771223041e-06,
+      "loss": 0.6697,
+      "step": 36083
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5816537150324972e-06,
+      "loss": 0.607,
+      "step": 36084
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5812689957187732e-06,
+      "loss": 0.6007,
+      "step": 36085
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5808843191830846e-06,
+      "loss": 0.6435,
+      "step": 36086
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5804996854273935e-06,
+      "loss": 0.6058,
+      "step": 36087
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5801150944536436e-06,
+      "loss": 0.5584,
+      "step": 36088
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.579730546263798e-06,
+      "loss": 0.5923,
+      "step": 36089
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.579346040859805e-06,
+      "loss": 0.6483,
+      "step": 36090
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.578961578243623e-06,
+      "loss": 0.5743,
+      "step": 36091
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.578577158417204e-06,
+      "loss": 0.6062,
+      "step": 36092
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5781927813824993e-06,
+      "loss": 0.6367,
+      "step": 36093
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5778084471414623e-06,
+      "loss": 0.6194,
+      "step": 36094
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5774241556960502e-06,
+      "loss": 0.6201,
+      "step": 36095
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.577039907048209e-06,
+      "loss": 0.5768,
+      "step": 36096
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5766557011998972e-06,
+      "loss": 0.6104,
+      "step": 36097
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5762715381530658e-06,
+      "loss": 0.5702,
+      "step": 36098
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5758874179096651e-06,
+      "loss": 0.5975,
+      "step": 36099
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5755033404716447e-06,
+      "loss": 0.601,
+      "step": 36100
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.575119305840962e-06,
+      "loss": 0.5517,
+      "step": 36101
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5747353140195653e-06,
+      "loss": 0.6747,
+      "step": 36102
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5743513650094057e-06,
+      "loss": 0.5906,
+      "step": 36103
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.573967458812432e-06,
+      "loss": 0.6145,
+      "step": 36104
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5735835954306e-06,
+      "loss": 0.6152,
+      "step": 36105
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5731997748658579e-06,
+      "loss": 0.6474,
+      "step": 36106
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5728159971201528e-06,
+      "loss": 0.6032,
+      "step": 36107
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5724322621954403e-06,
+      "loss": 0.5962,
+      "step": 36108
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5720485700936672e-06,
+      "loss": 0.5992,
+      "step": 36109
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5716649208167834e-06,
+      "loss": 0.6241,
+      "step": 36110
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5712813143667361e-06,
+      "loss": 0.5812,
+      "step": 36111
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5708977507454804e-06,
+      "loss": 0.5697,
+      "step": 36112
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.570514229954958e-06,
+      "loss": 0.5928,
+      "step": 36113
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5701307519971264e-06,
+      "loss": 0.5576,
+      "step": 36114
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5697473168739253e-06,
+      "loss": 0.6224,
+      "step": 36115
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5693639245873093e-06,
+      "loss": 0.6273,
+      "step": 36116
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5689805751392218e-06,
+      "loss": 0.6474,
+      "step": 36117
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5685972685316154e-06,
+      "loss": 0.5893,
+      "step": 36118
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.568214004766435e-06,
+      "loss": 0.6003,
+      "step": 36119
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5678307838456286e-06,
+      "loss": 0.6156,
+      "step": 36120
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5674476057711418e-06,
+      "loss": 0.6277,
+      "step": 36121
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5670644705449256e-06,
+      "loss": 0.6237,
+      "step": 36122
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5666813781689237e-06,
+      "loss": 0.5553,
+      "step": 36123
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5662983286450817e-06,
+      "loss": 0.5105,
+      "step": 36124
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.565915321975351e-06,
+      "loss": 0.555,
+      "step": 36125
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5655323581616733e-06,
+      "loss": 0.611,
+      "step": 36126
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5651494372059973e-06,
+      "loss": 0.6338,
+      "step": 36127
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5647665591102635e-06,
+      "loss": 0.6223,
+      "step": 36128
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5643837238764237e-06,
+      "loss": 0.6242,
+      "step": 36129
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.564000931506421e-06,
+      "loss": 0.6111,
+      "step": 36130
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5636181820022e-06,
+      "loss": 0.5777,
+      "step": 36131
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.563235475365704e-06,
+      "loss": 0.6202,
+      "step": 36132
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5628528115988817e-06,
+      "loss": 0.5994,
+      "step": 36133
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5624701907036721e-06,
+      "loss": 0.6355,
+      "step": 36134
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.562087612682025e-06,
+      "loss": 0.6071,
+      "step": 36135
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5617050775358823e-06,
+      "loss": 0.5799,
+      "step": 36136
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5613225852671864e-06,
+      "loss": 0.6075,
+      "step": 36137
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5609401358778808e-06,
+      "loss": 0.5388,
+      "step": 36138
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5605577293699114e-06,
+      "loss": 0.6606,
+      "step": 36139
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.56017536574522e-06,
+      "loss": 0.5678,
+      "step": 36140
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5597930450057496e-06,
+      "loss": 0.5749,
+      "step": 36141
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5594107671534398e-06,
+      "loss": 0.5721,
+      "step": 36142
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5590285321902387e-06,
+      "loss": 0.6086,
+      "step": 36143
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.558646340118085e-06,
+      "loss": 0.6381,
+      "step": 36144
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5582641909389196e-06,
+      "loss": 0.5865,
+      "step": 36145
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5578820846546882e-06,
+      "loss": 0.5432,
+      "step": 36146
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5575000212673296e-06,
+      "loss": 0.5999,
+      "step": 36147
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5571180007787868e-06,
+      "loss": 0.5328,
+      "step": 36148
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5567360231909979e-06,
+      "loss": 0.6554,
+      "step": 36149
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5563540885059082e-06,
+      "loss": 0.6576,
+      "step": 36150
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5559721967254525e-06,
+      "loss": 0.6261,
+      "step": 36151
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.555590347851581e-06,
+      "loss": 0.5771,
+      "step": 36152
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.555208541886224e-06,
+      "loss": 0.5957,
+      "step": 36153
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5548267788313265e-06,
+      "loss": 0.5893,
+      "step": 36154
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.554445058688826e-06,
+      "loss": 0.6512,
+      "step": 36155
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5540633814606653e-06,
+      "loss": 0.5584,
+      "step": 36156
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.553681747148782e-06,
+      "loss": 0.6449,
+      "step": 36157
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5533001557551164e-06,
+      "loss": 0.5975,
+      "step": 36158
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5529186072816027e-06,
+      "loss": 0.6008,
+      "step": 36159
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.552537101730187e-06,
+      "loss": 0.6296,
+      "step": 36160
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5521556391028036e-06,
+      "loss": 0.5715,
+      "step": 36161
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5517742194013885e-06,
+      "loss": 0.5704,
+      "step": 36162
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5513928426278858e-06,
+      "loss": 0.5879,
+      "step": 36163
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5510115087842304e-06,
+      "loss": 0.6579,
+      "step": 36164
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5506302178723597e-06,
+      "loss": 0.5831,
+      "step": 36165
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5502489698942092e-06,
+      "loss": 0.6123,
+      "step": 36166
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5498677648517212e-06,
+      "loss": 0.5669,
+      "step": 36167
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5494866027468292e-06,
+      "loss": 0.5802,
+      "step": 36168
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5491054835814701e-06,
+      "loss": 0.6847,
+      "step": 36169
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5487244073575803e-06,
+      "loss": 0.6614,
+      "step": 36170
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.548343374077097e-06,
+      "loss": 0.581,
+      "step": 36171
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5479623837419556e-06,
+      "loss": 0.6809,
+      "step": 36172
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5475814363540964e-06,
+      "loss": 0.6062,
+      "step": 36173
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.547200531915447e-06,
+      "loss": 0.5816,
+      "step": 36174
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5468196704279492e-06,
+      "loss": 0.5346,
+      "step": 36175
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5464388518935348e-06,
+      "loss": 0.6034,
+      "step": 36176
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.546058076314142e-06,
+      "loss": 0.6491,
+      "step": 36177
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5456773436917038e-06,
+      "loss": 0.6157,
+      "step": 36178
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5452966540281556e-06,
+      "loss": 0.5711,
+      "step": 36179
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5449160073254287e-06,
+      "loss": 0.5886,
+      "step": 36180
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.544535403585462e-06,
+      "loss": 0.5927,
+      "step": 36181
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.544154842810187e-06,
+      "loss": 0.5172,
+      "step": 36182
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5437743250015347e-06,
+      "loss": 0.5967,
+      "step": 36183
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5433938501614454e-06,
+      "loss": 0.6297,
+      "step": 36184
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5430134182918476e-06,
+      "loss": 0.5102,
+      "step": 36185
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5426330293946756e-06,
+      "loss": 0.6266,
+      "step": 36186
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5422526834718598e-06,
+      "loss": 0.5858,
+      "step": 36187
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5418723805253366e-06,
+      "loss": 0.6155,
+      "step": 36188
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5414921205570355e-06,
+      "loss": 0.5745,
+      "step": 36189
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5411119035688949e-06,
+      "loss": 0.5866,
+      "step": 36190
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5407317295628366e-06,
+      "loss": 0.5869,
+      "step": 36191
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5403515985408014e-06,
+      "loss": 0.5961,
+      "step": 36192
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5399715105047131e-06,
+      "loss": 0.6018,
+      "step": 36193
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5395914654565102e-06,
+      "loss": 0.6025,
+      "step": 36194
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5392114633981214e-06,
+      "loss": 0.5486,
+      "step": 36195
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5388315043314772e-06,
+      "loss": 0.6148,
+      "step": 36196
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.538451588258505e-06,
+      "loss": 0.5974,
+      "step": 36197
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5380717151811409e-06,
+      "loss": 0.592,
+      "step": 36198
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5376918851013123e-06,
+      "loss": 0.5677,
+      "step": 36199
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5373120980209477e-06,
+      "loss": 0.6447,
+      "step": 36200
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.536932353941981e-06,
+      "loss": 0.5998,
+      "step": 36201
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.53655265286634e-06,
+      "loss": 0.638,
+      "step": 36202
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5361729947959525e-06,
+      "loss": 0.6186,
+      "step": 36203
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5357933797327472e-06,
+      "loss": 0.6096,
+      "step": 36204
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5354138076786574e-06,
+      "loss": 0.6666,
+      "step": 36205
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5350342786356076e-06,
+      "loss": 0.6236,
+      "step": 36206
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5346547926055288e-06,
+      "loss": 0.5866,
+      "step": 36207
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.534275349590345e-06,
+      "loss": 0.6335,
+      "step": 36208
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5338959495919903e-06,
+      "loss": 0.5837,
+      "step": 36209
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5335165926123873e-06,
+      "loss": 0.6637,
+      "step": 36210
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5331372786534714e-06,
+      "loss": 0.6017,
+      "step": 36211
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5327580077171589e-06,
+      "loss": 0.5825,
+      "step": 36212
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5323787798053857e-06,
+      "loss": 0.608,
+      "step": 36213
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5319995949200727e-06,
+      "loss": 0.5914,
+      "step": 36214
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.531620453063153e-06,
+      "loss": 0.627,
+      "step": 36215
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5312413542365468e-06,
+      "loss": 0.6359,
+      "step": 36216
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5308622984421883e-06,
+      "loss": 0.6496,
+      "step": 36217
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.530483285681994e-06,
+      "loss": 0.5739,
+      "step": 36218
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5301043159578978e-06,
+      "loss": 0.5898,
+      "step": 36219
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5297253892718212e-06,
+      "loss": 0.6078,
+      "step": 36220
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.5293465056256884e-06,
+      "loss": 0.6278,
+      "step": 36221
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.528967665021429e-06,
+      "loss": 0.6051,
+      "step": 36222
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5285888674609662e-06,
+      "loss": 0.5906,
+      "step": 36223
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5282101129462236e-06,
+      "loss": 0.5922,
+      "step": 36224
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.527831401479124e-06,
+      "loss": 0.6669,
+      "step": 36225
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5274527330615975e-06,
+      "loss": 0.5802,
+      "step": 36226
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.527074107695562e-06,
+      "loss": 0.5459,
+      "step": 36227
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5266955253829496e-06,
+      "loss": 0.5435,
+      "step": 36228
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5263169861256733e-06,
+      "loss": 0.6062,
+      "step": 36229
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5259384899256645e-06,
+      "loss": 0.5492,
+      "step": 36230
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5255600367848422e-06,
+      "loss": 0.608,
+      "step": 36231
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5251816267051334e-06,
+      "loss": 0.6169,
+      "step": 36232
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5248032596884587e-06,
+      "loss": 0.5463,
+      "step": 36233
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5244249357367403e-06,
+      "loss": 0.5411,
+      "step": 36234
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5240466548518996e-06,
+      "loss": 0.5636,
+      "step": 36235
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5236684170358617e-06,
+      "loss": 0.6125,
+      "step": 36236
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5232902222905444e-06,
+      "loss": 0.6313,
+      "step": 36237
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5229120706178769e-06,
+      "loss": 0.5465,
+      "step": 36238
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.522533962019771e-06,
+      "loss": 0.6767,
+      "step": 36239
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5221558964981565e-06,
+      "loss": 0.6473,
+      "step": 36240
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5217778740549493e-06,
+      "loss": 0.5364,
+      "step": 36241
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5213998946920693e-06,
+      "loss": 0.6588,
+      "step": 36242
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5210219584114428e-06,
+      "loss": 0.6114,
+      "step": 36243
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5206440652149856e-06,
+      "loss": 0.6398,
+      "step": 36244
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5202662151046198e-06,
+      "loss": 0.5905,
+      "step": 36245
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.519888408082263e-06,
+      "loss": 0.5864,
+      "step": 36246
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5195106441498385e-06,
+      "loss": 0.644,
+      "step": 36247
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5191329233092622e-06,
+      "loss": 0.531,
+      "step": 36248
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5187552455624599e-06,
+      "loss": 0.6218,
+      "step": 36249
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5183776109113401e-06,
+      "loss": 0.6144,
+      "step": 36250
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5180000193578315e-06,
+      "loss": 0.5614,
+      "step": 36251
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5176224709038456e-06,
+      "loss": 0.5861,
+      "step": 36252
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5172449655513067e-06,
+      "loss": 0.6016,
+      "step": 36253
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5168675033021286e-06,
+      "loss": 0.6361,
+      "step": 36254
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5164900841582353e-06,
+      "loss": 0.5763,
+      "step": 36255
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5161127081215354e-06,
+      "loss": 0.5712,
+      "step": 36256
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.515735375193954e-06,
+      "loss": 0.6309,
+      "step": 36257
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5153580853774064e-06,
+      "loss": 0.5673,
+      "step": 36258
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5149808386738063e-06,
+      "loss": 0.5516,
+      "step": 36259
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5146036350850756e-06,
+      "loss": 0.5699,
+      "step": 36260
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5142264746131296e-06,
+      "loss": 0.6377,
+      "step": 36261
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5138493572598845e-06,
+      "loss": 0.5967,
+      "step": 36262
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.513472283027253e-06,
+      "loss": 0.5666,
+      "step": 36263
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.513095251917156e-06,
+      "loss": 0.5766,
+      "step": 36264
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5127182639315086e-06,
+      "loss": 0.5895,
+      "step": 36265
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5123413190722248e-06,
+      "loss": 0.5706,
+      "step": 36266
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5119644173412184e-06,
+      "loss": 0.5976,
+      "step": 36267
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5115875587404083e-06,
+      "loss": 0.6175,
+      "step": 36268
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.511210743271706e-06,
+      "loss": 0.6224,
+      "step": 36269
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5108339709370302e-06,
+      "loss": 0.5601,
+      "step": 36270
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5104572417382934e-06,
+      "loss": 0.5783,
+      "step": 36271
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.510080555677409e-06,
+      "loss": 0.5876,
+      "step": 36272
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5097039127562907e-06,
+      "loss": 0.5762,
+      "step": 36273
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5093273129768537e-06,
+      "loss": 0.568,
+      "step": 36274
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.50895075634101e-06,
+      "loss": 0.6045,
+      "step": 36275
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5085742428506788e-06,
+      "loss": 0.6031,
+      "step": 36276
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5081977725077645e-06,
+      "loss": 0.5763,
+      "step": 36277
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5078213453141866e-06,
+      "loss": 0.6158,
+      "step": 36278
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5074449612718545e-06,
+      "loss": 0.5822,
+      "step": 36279
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.50706862038268e-06,
+      "loss": 0.5833,
+      "step": 36280
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5066923226485797e-06,
+      "loss": 0.6313,
+      "step": 36281
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5063160680714628e-06,
+      "loss": 0.6242,
+      "step": 36282
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5059398566532413e-06,
+      "loss": 0.6033,
+      "step": 36283
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5055636883958247e-06,
+      "loss": 0.6337,
+      "step": 36284
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5051875633011292e-06,
+      "loss": 0.6099,
+      "step": 36285
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5048114813710612e-06,
+      "loss": 0.6699,
+      "step": 36286
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.504435442607538e-06,
+      "loss": 0.5315,
+      "step": 36287
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5040594470124625e-06,
+      "loss": 0.5612,
+      "step": 36288
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.503683494587751e-06,
+      "loss": 0.5763,
+      "step": 36289
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5033075853353096e-06,
+      "loss": 0.574,
+      "step": 36290
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5029317192570535e-06,
+      "loss": 0.6003,
+      "step": 36291
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.502555896354887e-06,
+      "loss": 0.6266,
+      "step": 36292
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.502180116630727e-06,
+      "loss": 0.6015,
+      "step": 36293
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5018043800864746e-06,
+      "loss": 0.5714,
+      "step": 36294
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5014286867240457e-06,
+      "loss": 0.5986,
+      "step": 36295
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5010530365453435e-06,
+      "loss": 0.5793,
+      "step": 36296
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5006774295522818e-06,
+      "loss": 0.5934,
+      "step": 36297
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.5003018657467673e-06,
+      "loss": 0.6326,
+      "step": 36298
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4999263451307088e-06,
+      "loss": 0.5942,
+      "step": 36299
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4995508677060134e-06,
+      "loss": 0.592,
+      "step": 36300
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4991754334745878e-06,
+      "loss": 0.6346,
+      "step": 36301
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.498800042438343e-06,
+      "loss": 0.5309,
+      "step": 36302
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4984246945991853e-06,
+      "loss": 0.5823,
+      "step": 36303
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4980493899590221e-06,
+      "loss": 0.5978,
+      "step": 36304
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.497674128519757e-06,
+      "loss": 0.6078,
+      "step": 36305
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4972989102833024e-06,
+      "loss": 0.5916,
+      "step": 36306
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4969237352515587e-06,
+      "loss": 0.6595,
+      "step": 36307
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4965486034264387e-06,
+      "loss": 0.6354,
+      "step": 36308
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4961735148098455e-06,
+      "loss": 0.5841,
+      "step": 36309
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4957984694036853e-06,
+      "loss": 0.5844,
+      "step": 36310
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4954234672098622e-06,
+      "loss": 0.6428,
+      "step": 36311
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4950485082302846e-06,
+      "loss": 0.5739,
+      "step": 36312
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4946735924668532e-06,
+      "loss": 0.5888,
+      "step": 36313
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.494298719921482e-06,
+      "loss": 0.6065,
+      "step": 36314
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4939238905960652e-06,
+      "loss": 0.5665,
+      "step": 36315
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4935491044925143e-06,
+      "loss": 0.6205,
+      "step": 36316
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4931743616127292e-06,
+      "loss": 0.6049,
+      "step": 36317
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4927996619586194e-06,
+      "loss": 0.5824,
+      "step": 36318
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4924250055320854e-06,
+      "loss": 0.5362,
+      "step": 36319
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4920503923350315e-06,
+      "loss": 0.6436,
+      "step": 36320
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4916758223693616e-06,
+      "loss": 0.6171,
+      "step": 36321
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4913012956369755e-06,
+      "loss": 0.5746,
+      "step": 36322
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4909268121397825e-06,
+      "loss": 0.5948,
+      "step": 36323
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.49055237187968e-06,
+      "loss": 0.5683,
+      "step": 36324
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4901779748585777e-06,
+      "loss": 0.6159,
+      "step": 36325
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4898036210783684e-06,
+      "loss": 0.5698,
+      "step": 36326
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.489429310540962e-06,
+      "loss": 0.586,
+      "step": 36327
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4890550432482543e-06,
+      "loss": 0.6123,
+      "step": 36328
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.488680819202154e-06,
+      "loss": 0.6607,
+      "step": 36329
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4883066384045586e-06,
+      "loss": 0.5957,
+      "step": 36330
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4879325008573696e-06,
+      "loss": 0.6017,
+      "step": 36331
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4875584065624871e-06,
+      "loss": 0.5763,
+      "step": 36332
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4871843555218147e-06,
+      "loss": 0.6218,
+      "step": 36333
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4868103477372497e-06,
+      "loss": 0.6058,
+      "step": 36334
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4864363832106965e-06,
+      "loss": 0.6233,
+      "step": 36335
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4860624619440544e-06,
+      "loss": 0.5814,
+      "step": 36336
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.485688583939221e-06,
+      "loss": 0.6162,
+      "step": 36337
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4853147491980956e-06,
+      "loss": 0.5792,
+      "step": 36338
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4849409577225826e-06,
+      "loss": 0.5523,
+      "step": 36339
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.484567209514578e-06,
+      "loss": 0.5658,
+      "step": 36340
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4841935045759803e-06,
+      "loss": 0.6368,
+      "step": 36341
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4838198429086902e-06,
+      "loss": 0.5771,
+      "step": 36342
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.483446224514603e-06,
+      "loss": 0.6372,
+      "step": 36343
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4830726493956227e-06,
+      "loss": 0.6197,
+      "step": 36344
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.482699117553642e-06,
+      "loss": 0.5432,
+      "step": 36345
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4823256289905641e-06,
+      "loss": 0.6107,
+      "step": 36346
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4819521837082828e-06,
+      "loss": 0.5705,
+      "step": 36347
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.481578781708698e-06,
+      "loss": 0.6242,
+      "step": 36348
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4812054229937044e-06,
+      "loss": 0.6757,
+      "step": 36349
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4808321075652022e-06,
+      "loss": 0.5711,
+      "step": 36350
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4804588354250849e-06,
+      "loss": 0.6093,
+      "step": 36351
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4800856065752566e-06,
+      "loss": 0.659,
+      "step": 36352
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4797124210176029e-06,
+      "loss": 0.5925,
+      "step": 36353
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4793392787540295e-06,
+      "loss": 0.5893,
+      "step": 36354
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4789661797864252e-06,
+      "loss": 0.588,
+      "step": 36355
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4785931241166917e-06,
+      "loss": 0.6177,
+      "step": 36356
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4782201117467221e-06,
+      "loss": 0.5974,
+      "step": 36357
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4778471426784113e-06,
+      "loss": 0.629,
+      "step": 36358
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4774742169136558e-06,
+      "loss": 0.5786,
+      "step": 36359
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4771013344543472e-06,
+      "loss": 0.5692,
+      "step": 36360
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4767284953023854e-06,
+      "loss": 0.5788,
+      "step": 36361
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4763556994596595e-06,
+      "loss": 0.5679,
+      "step": 36362
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4759829469280706e-06,
+      "loss": 0.5925,
+      "step": 36363
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.475610237709505e-06,
+      "loss": 0.5635,
+      "step": 36364
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.475237571805862e-06,
+      "loss": 0.5931,
+      "step": 36365
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4748649492190326e-06,
+      "loss": 0.5654,
+      "step": 36366
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4744923699509118e-06,
+      "loss": 0.6469,
+      "step": 36367
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.4741198340033925e-06,
+      "loss": 0.6112,
+      "step": 36368
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4737473413783676e-06,
+      "loss": 0.5604,
+      "step": 36369
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4733748920777268e-06,
+      "loss": 0.6017,
+      "step": 36370
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4730024861033677e-06,
+      "loss": 0.5519,
+      "step": 36371
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4726301234571782e-06,
+      "loss": 0.6074,
+      "step": 36372
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.472257804141054e-06,
+      "loss": 0.627,
+      "step": 36373
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4718855281568867e-06,
+      "loss": 0.6113,
+      "step": 36374
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4715132955065658e-06,
+      "loss": 0.5902,
+      "step": 36375
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.471141106191981e-06,
+      "loss": 0.5686,
+      "step": 36376
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4707689602150289e-06,
+      "loss": 0.523,
+      "step": 36377
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4703968575775963e-06,
+      "loss": 0.555,
+      "step": 36378
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4700247982815763e-06,
+      "loss": 0.6186,
+      "step": 36379
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4696527823288588e-06,
+      "loss": 0.5882,
+      "step": 36380
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4692808097213307e-06,
+      "loss": 0.6478,
+      "step": 36381
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4689088804608876e-06,
+      "loss": 0.7124,
+      "step": 36382
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4685369945494143e-06,
+      "loss": 0.6295,
+      "step": 36383
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4681651519888062e-06,
+      "loss": 0.628,
+      "step": 36384
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4677933527809495e-06,
+      "loss": 0.6218,
+      "step": 36385
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4674215969277329e-06,
+      "loss": 0.6182,
+      "step": 36386
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4670498844310443e-06,
+      "loss": 0.6541,
+      "step": 36387
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.466678215292776e-06,
+      "loss": 0.6237,
+      "step": 36388
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4663065895148132e-06,
+      "loss": 0.6648,
+      "step": 36389
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4659350070990508e-06,
+      "loss": 0.5547,
+      "step": 36390
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4655634680473663e-06,
+      "loss": 0.6094,
+      "step": 36391
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4651919723616558e-06,
+      "loss": 0.6349,
+      "step": 36392
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4648205200438027e-06,
+      "loss": 0.6075,
+      "step": 36393
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.464449111095697e-06,
+      "loss": 0.6187,
+      "step": 36394
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4640777455192267e-06,
+      "loss": 0.5765,
+      "step": 36395
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4637064233162768e-06,
+      "loss": 0.5655,
+      "step": 36396
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.463335144488731e-06,
+      "loss": 0.6437,
+      "step": 36397
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.462963909038484e-06,
+      "loss": 0.5999,
+      "step": 36398
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4625927169674158e-06,
+      "loss": 0.5621,
+      "step": 36399
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4622215682774122e-06,
+      "loss": 0.6626,
+      "step": 36400
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4618504629703667e-06,
+      "loss": 0.5766,
+      "step": 36401
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4614794010481536e-06,
+      "loss": 0.6126,
+      "step": 36402
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4611083825126682e-06,
+      "loss": 0.5696,
+      "step": 36403
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4607374073657887e-06,
+      "loss": 0.5794,
+      "step": 36404
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.460366475609405e-06,
+      "loss": 0.5804,
+      "step": 36405
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4599955872454009e-06,
+      "loss": 0.6213,
+      "step": 36406
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4596247422756605e-06,
+      "loss": 0.5645,
+      "step": 36407
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4592539407020645e-06,
+      "loss": 0.5773,
+      "step": 36408
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4588831825265037e-06,
+      "loss": 0.5926,
+      "step": 36409
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4585124677508567e-06,
+      "loss": 0.6208,
+      "step": 36410
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4581417963770117e-06,
+      "loss": 0.6088,
+      "step": 36411
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4577711684068497e-06,
+      "loss": 0.5365,
+      "step": 36412
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4574005838422533e-06,
+      "loss": 0.6273,
+      "step": 36413
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4570300426851047e-06,
+      "loss": 0.5878,
+      "step": 36414
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.456659544937291e-06,
+      "loss": 0.6576,
+      "step": 36415
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.456289090600691e-06,
+      "loss": 0.537,
+      "step": 36416
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4559186796771895e-06,
+      "loss": 0.6146,
+      "step": 36417
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4555483121686642e-06,
+      "loss": 0.6439,
+      "step": 36418
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4551779880770034e-06,
+      "loss": 0.6109,
+      "step": 36419
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4548077074040857e-06,
+      "loss": 0.5954,
+      "step": 36420
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4544374701517904e-06,
+      "loss": 0.6183,
+      "step": 36421
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4540672763220043e-06,
+      "loss": 0.6084,
+      "step": 36422
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.453697125916602e-06,
+      "loss": 0.5728,
+      "step": 36423
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4533270189374693e-06,
+      "loss": 0.6115,
+      "step": 36424
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.452956955386483e-06,
+      "loss": 0.6312,
+      "step": 36425
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4525869352655275e-06,
+      "loss": 0.6345,
+      "step": 36426
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.452216958576479e-06,
+      "loss": 0.5564,
+      "step": 36427
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4518470253212235e-06,
+      "loss": 0.5965,
+      "step": 36428
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4514771355016321e-06,
+      "loss": 0.6384,
+      "step": 36429
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.451107289119592e-06,
+      "loss": 0.554,
+      "step": 36430
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4507374861769763e-06,
+      "loss": 0.5435,
+      "step": 36431
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4503677266756699e-06,
+      "loss": 0.6558,
+      "step": 36432
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4499980106175493e-06,
+      "loss": 0.6047,
+      "step": 36433
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4496283380044918e-06,
+      "loss": 0.6162,
+      "step": 36434
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4492587088383758e-06,
+      "loss": 0.5523,
+      "step": 36435
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.448889123121081e-06,
+      "loss": 0.6025,
+      "step": 36436
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.448519580854486e-06,
+      "loss": 0.5492,
+      "step": 36437
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4481500820404659e-06,
+      "loss": 0.5863,
+      "step": 36438
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4477806266809035e-06,
+      "loss": 0.6011,
+      "step": 36439
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4474112147776676e-06,
+      "loss": 0.6061,
+      "step": 36440
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.447041846332643e-06,
+      "loss": 0.6475,
+      "step": 36441
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4466725213477007e-06,
+      "loss": 0.6429,
+      "step": 36442
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4463032398247223e-06,
+      "loss": 0.5745,
+      "step": 36443
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.445934001765583e-06,
+      "loss": 0.5789,
+      "step": 36444
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4455648071721585e-06,
+      "loss": 0.5518,
+      "step": 36445
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.445195656046321e-06,
+      "loss": 0.608,
+      "step": 36446
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4448265483899527e-06,
+      "loss": 0.568,
+      "step": 36447
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4444574842049242e-06,
+      "loss": 0.6052,
+      "step": 36448
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4440884634931174e-06,
+      "loss": 0.6291,
+      "step": 36449
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4437194862563975e-06,
+      "loss": 0.6095,
+      "step": 36450
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4433505524966474e-06,
+      "loss": 0.5843,
+      "step": 36451
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4429816622157367e-06,
+      "loss": 0.5719,
+      "step": 36452
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.442612815415545e-06,
+      "loss": 0.5911,
+      "step": 36453
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.442244012097943e-06,
+      "loss": 0.5856,
+      "step": 36454
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4418752522648061e-06,
+      "loss": 0.5619,
+      "step": 36455
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4415065359180048e-06,
+      "loss": 0.5614,
+      "step": 36456
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4411378630594175e-06,
+      "loss": 0.6121,
+      "step": 36457
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4407692336909162e-06,
+      "loss": 0.5941,
+      "step": 36458
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4404006478143696e-06,
+      "loss": 0.5973,
+      "step": 36459
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.440032105431659e-06,
+      "loss": 0.6013,
+      "step": 36460
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.439663606544648e-06,
+      "loss": 0.5885,
+      "step": 36461
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4392951511552144e-06,
+      "loss": 0.6104,
+      "step": 36462
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4389267392652285e-06,
+      "loss": 0.5923,
+      "step": 36463
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.438558370876565e-06,
+      "loss": 0.5717,
+      "step": 36464
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4381900459910902e-06,
+      "loss": 0.6074,
+      "step": 36465
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4378217646106851e-06,
+      "loss": 0.6074,
+      "step": 36466
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4374535267372102e-06,
+      "loss": 0.6657,
+      "step": 36467
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4370853323725443e-06,
+      "loss": 0.6005,
+      "step": 36468
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4367171815185522e-06,
+      "loss": 0.671,
+      "step": 36469
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4363490741771113e-06,
+      "loss": 0.6076,
+      "step": 36470
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4359810103500883e-06,
+      "loss": 0.6135,
+      "step": 36471
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4356129900393533e-06,
+      "loss": 0.6227,
+      "step": 36472
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4352450132467754e-06,
+      "loss": 0.6255,
+      "step": 36473
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4348770799742285e-06,
+      "loss": 0.6407,
+      "step": 36474
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4345091902235787e-06,
+      "loss": 0.6304,
+      "step": 36475
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4341413439966945e-06,
+      "loss": 0.6007,
+      "step": 36476
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4337735412954489e-06,
+      "loss": 0.6514,
+      "step": 36477
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4334057821217084e-06,
+      "loss": 0.5854,
+      "step": 36478
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4330380664773423e-06,
+      "loss": 0.653,
+      "step": 36479
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4326703943642162e-06,
+      "loss": 0.5518,
+      "step": 36480
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4323027657842036e-06,
+      "loss": 0.5787,
+      "step": 36481
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.431935180739169e-06,
+      "loss": 0.6487,
+      "step": 36482
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4315676392309819e-06,
+      "loss": 0.5666,
+      "step": 36483
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4312001412615061e-06,
+      "loss": 0.6116,
+      "step": 36484
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.430832686832615e-06,
+      "loss": 0.63,
+      "step": 36485
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.430465275946169e-06,
+      "loss": 0.5678,
+      "step": 36486
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4300979086040445e-06,
+      "loss": 0.6225,
+      "step": 36487
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4297305848080978e-06,
+      "loss": 0.6255,
+      "step": 36488
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4293633045602006e-06,
+      "loss": 0.5759,
+      "step": 36489
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4289960678622173e-06,
+      "loss": 0.6311,
+      "step": 36490
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4286288747160182e-06,
+      "loss": 0.5923,
+      "step": 36491
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4282617251234643e-06,
+      "loss": 0.6186,
+      "step": 36492
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.427894619086424e-06,
+      "loss": 0.586,
+      "step": 36493
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4275275566067592e-06,
+      "loss": 0.5517,
+      "step": 36494
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4271605376863395e-06,
+      "loss": 0.6076,
+      "step": 36495
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.426793562327028e-06,
+      "loss": 0.6064,
+      "step": 36496
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4264266305306874e-06,
+      "loss": 0.6159,
+      "step": 36497
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.426059742299185e-06,
+      "loss": 0.6346,
+      "step": 36498
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4256928976343853e-06,
+      "loss": 0.6565,
+      "step": 36499
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4253260965381498e-06,
+      "loss": 0.6449,
+      "step": 36500
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4249593390123429e-06,
+      "loss": 0.6229,
+      "step": 36501
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4245926250588304e-06,
+      "loss": 0.5438,
+      "step": 36502
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4242259546794712e-06,
+      "loss": 0.6166,
+      "step": 36503
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.423859327876137e-06,
+      "loss": 0.6113,
+      "step": 36504
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4234927446506807e-06,
+      "loss": 0.5739,
+      "step": 36505
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.423126205004971e-06,
+      "loss": 0.6306,
+      "step": 36506
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4227597089408674e-06,
+      "loss": 0.6287,
+      "step": 36507
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.422393256460235e-06,
+      "loss": 0.6669,
+      "step": 36508
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4220268475649356e-06,
+      "loss": 0.6506,
+      "step": 36509
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4216604822568292e-06,
+      "loss": 0.6207,
+      "step": 36510
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4212941605377762e-06,
+      "loss": 0.5821,
+      "step": 36511
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4209278824096418e-06,
+      "loss": 0.603,
+      "step": 36512
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4205616478742857e-06,
+      "loss": 0.5373,
+      "step": 36513
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4201954569335686e-06,
+      "loss": 0.6186,
+      "step": 36514
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.4198293095893489e-06,
+      "loss": 0.598,
+      "step": 36515
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.419463205843492e-06,
+      "loss": 0.5778,
+      "step": 36516
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4190971456978542e-06,
+      "loss": 0.6247,
+      "step": 36517
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.418731129154296e-06,
+      "loss": 0.6276,
+      "step": 36518
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4183651562146793e-06,
+      "loss": 0.6184,
+      "step": 36519
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.417999226880863e-06,
+      "loss": 0.5693,
+      "step": 36520
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.417633341154705e-06,
+      "loss": 0.5737,
+      "step": 36521
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.417267499038064e-06,
+      "loss": 0.5996,
+      "step": 36522
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4169017005328034e-06,
+      "loss": 0.6045,
+      "step": 36523
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4165359456407757e-06,
+      "loss": 0.6174,
+      "step": 36524
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4161702343638474e-06,
+      "loss": 0.633,
+      "step": 36525
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4158045667038678e-06,
+      "loss": 0.6175,
+      "step": 36526
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4154389426627002e-06,
+      "loss": 0.556,
+      "step": 36527
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4150733622422008e-06,
+      "loss": 0.6212,
+      "step": 36528
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4147078254442293e-06,
+      "loss": 0.5626,
+      "step": 36529
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4143423322706395e-06,
+      "loss": 0.567,
+      "step": 36530
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4139768827232947e-06,
+      "loss": 0.6356,
+      "step": 36531
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4136114768040444e-06,
+      "loss": 0.6771,
+      "step": 36532
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4132461145147503e-06,
+      "loss": 0.625,
+      "step": 36533
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4128807958572676e-06,
+      "loss": 0.5832,
+      "step": 36534
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4125155208334508e-06,
+      "loss": 0.559,
+      "step": 36535
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.41215028944516e-06,
+      "loss": 0.6235,
+      "step": 36536
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4117851016942475e-06,
+      "loss": 0.6158,
+      "step": 36537
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4114199575825704e-06,
+      "loss": 0.5284,
+      "step": 36538
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.411054857111982e-06,
+      "loss": 0.5537,
+      "step": 36539
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4106898002843417e-06,
+      "loss": 0.6302,
+      "step": 36540
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4103247871015013e-06,
+      "loss": 0.602,
+      "step": 36541
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4099598175653162e-06,
+      "loss": 0.5232,
+      "step": 36542
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4095948916776393e-06,
+      "loss": 0.5519,
+      "step": 36543
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.409230009440329e-06,
+      "loss": 0.586,
+      "step": 36544
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4088651708552338e-06,
+      "loss": 0.5919,
+      "step": 36545
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4085003759242132e-06,
+      "loss": 0.657,
+      "step": 36546
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4081356246491196e-06,
+      "loss": 0.5921,
+      "step": 36547
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4077709170318033e-06,
+      "loss": 0.5276,
+      "step": 36548
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4074062530741183e-06,
+      "loss": 0.5944,
+      "step": 36549
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4070416327779202e-06,
+      "loss": 0.5279,
+      "step": 36550
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4066770561450605e-06,
+      "loss": 0.5974,
+      "step": 36551
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4063125231773911e-06,
+      "loss": 0.5782,
+      "step": 36552
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4059480338767628e-06,
+      "loss": 0.5643,
+      "step": 36553
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4055835882450309e-06,
+      "loss": 0.5894,
+      "step": 36554
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4052191862840458e-06,
+      "loss": 0.6015,
+      "step": 36555
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4048548279956576e-06,
+      "loss": 0.6054,
+      "step": 36556
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.404490513381721e-06,
+      "loss": 0.6129,
+      "step": 36557
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.404126242444086e-06,
+      "loss": 0.6538,
+      "step": 36558
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4037620151846021e-06,
+      "loss": 0.6623,
+      "step": 36559
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.403397831605119e-06,
+      "loss": 0.5683,
+      "step": 36560
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4030336917074916e-06,
+      "loss": 0.5616,
+      "step": 36561
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4026695954935653e-06,
+      "loss": 0.6336,
+      "step": 36562
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4023055429651977e-06,
+      "loss": 0.6381,
+      "step": 36563
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4019415341242282e-06,
+      "loss": 0.6032,
+      "step": 36564
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.401577568972514e-06,
+      "loss": 0.5966,
+      "step": 36565
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4012136475119008e-06,
+      "loss": 0.5992,
+      "step": 36566
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4008497697442413e-06,
+      "loss": 0.665,
+      "step": 36567
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.4004859356713806e-06,
+      "loss": 0.5504,
+      "step": 36568
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.400122145295173e-06,
+      "loss": 0.6015,
+      "step": 36569
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3997583986174601e-06,
+      "loss": 0.5621,
+      "step": 36570
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.399394695640095e-06,
+      "loss": 0.6213,
+      "step": 36571
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.399031036364924e-06,
+      "loss": 0.5655,
+      "step": 36572
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3986674207937934e-06,
+      "loss": 0.5926,
+      "step": 36573
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.398303848928554e-06,
+      "loss": 0.6077,
+      "step": 36574
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.397940320771053e-06,
+      "loss": 0.5686,
+      "step": 36575
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3975768363231357e-06,
+      "loss": 0.5429,
+      "step": 36576
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3972133955866484e-06,
+      "loss": 0.5803,
+      "step": 36577
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.396849998563441e-06,
+      "loss": 0.5898,
+      "step": 36578
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3964866452553582e-06,
+      "loss": 0.6008,
+      "step": 36579
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3961233356642467e-06,
+      "loss": 0.65,
+      "step": 36580
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3957600697919493e-06,
+      "loss": 0.6156,
+      "step": 36581
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3953968476403179e-06,
+      "loss": 0.6567,
+      "step": 36582
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3950336692111933e-06,
+      "loss": 0.6606,
+      "step": 36583
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3946705345064239e-06,
+      "loss": 0.5597,
+      "step": 36584
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3943074435278548e-06,
+      "loss": 0.6042,
+      "step": 36585
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3939443962773292e-06,
+      "loss": 0.5109,
+      "step": 36586
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3935813927566898e-06,
+      "loss": 0.6353,
+      "step": 36587
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3932184329677866e-06,
+      "loss": 0.5951,
+      "step": 36588
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.392855516912459e-06,
+      "loss": 0.5947,
+      "step": 36589
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.392492644592558e-06,
+      "loss": 0.5948,
+      "step": 36590
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3921298160099172e-06,
+      "loss": 0.5723,
+      "step": 36591
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.391767031166389e-06,
+      "loss": 0.5855,
+      "step": 36592
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3914042900638125e-06,
+      "loss": 0.6381,
+      "step": 36593
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3910415927040311e-06,
+      "loss": 0.5956,
+      "step": 36594
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3906789390888898e-06,
+      "loss": 0.5817,
+      "step": 36595
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3903163292202304e-06,
+      "loss": 0.6436,
+      "step": 36596
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.389953763099895e-06,
+      "loss": 0.6066,
+      "step": 36597
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3895912407297242e-06,
+      "loss": 0.6437,
+      "step": 36598
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.389228762111564e-06,
+      "loss": 0.6181,
+      "step": 36599
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3888663272472524e-06,
+      "loss": 0.5248,
+      "step": 36600
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3885039361386377e-06,
+      "loss": 0.5686,
+      "step": 36601
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3881415887875515e-06,
+      "loss": 0.5374,
+      "step": 36602
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3877792851958416e-06,
+      "loss": 0.5726,
+      "step": 36603
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3874170253653464e-06,
+      "loss": 0.5586,
+      "step": 36604
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3870548092979086e-06,
+      "loss": 0.6743,
+      "step": 36605
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.386692636995368e-06,
+      "loss": 0.5966,
+      "step": 36606
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3863305084595646e-06,
+      "loss": 0.6993,
+      "step": 36607
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3859684236923366e-06,
+      "loss": 0.5962,
+      "step": 36608
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.385606382695528e-06,
+      "loss": 0.5771,
+      "step": 36609
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.385244385470974e-06,
+      "loss": 0.5874,
+      "step": 36610
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.384882432020519e-06,
+      "loss": 0.6321,
+      "step": 36611
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.384520522345999e-06,
+      "loss": 0.6319,
+      "step": 36612
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3841586564492527e-06,
+      "loss": 0.5807,
+      "step": 36613
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3837968343321207e-06,
+      "loss": 0.5453,
+      "step": 36614
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.383435055996437e-06,
+      "loss": 0.5717,
+      "step": 36615
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3830733214440463e-06,
+      "loss": 0.5723,
+      "step": 36616
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3827116306767829e-06,
+      "loss": 0.6023,
+      "step": 36617
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.382349983696486e-06,
+      "loss": 0.5274,
+      "step": 36618
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3819883805049905e-06,
+      "loss": 0.5855,
+      "step": 36619
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3816268211041372e-06,
+      "loss": 0.5698,
+      "step": 36620
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3812653054957604e-06,
+      "loss": 0.5773,
+      "step": 36621
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3809038336817015e-06,
+      "loss": 0.6535,
+      "step": 36622
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.380542405663794e-06,
+      "loss": 0.5907,
+      "step": 36623
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.380181021443875e-06,
+      "loss": 0.5531,
+      "step": 36624
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3798196810237773e-06,
+      "loss": 0.5745,
+      "step": 36625
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3794583844053443e-06,
+      "loss": 0.6396,
+      "step": 36626
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3790971315904045e-06,
+      "loss": 0.6209,
+      "step": 36627
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3787359225808017e-06,
+      "loss": 0.5437,
+      "step": 36628
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3783747573783624e-06,
+      "loss": 0.6445,
+      "step": 36629
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3780136359849272e-06,
+      "loss": 0.6218,
+      "step": 36630
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3776525584023316e-06,
+      "loss": 0.5906,
+      "step": 36631
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3772915246324048e-06,
+      "loss": 0.587,
+      "step": 36632
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3769305346769868e-06,
+      "loss": 0.6106,
+      "step": 36633
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3765695885379115e-06,
+      "loss": 0.612,
+      "step": 36634
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.376208686217011e-06,
+      "loss": 0.5763,
+      "step": 36635
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.375847827716118e-06,
+      "loss": 0.587,
+      "step": 36636
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.375487013037069e-06,
+      "loss": 0.593,
+      "step": 36637
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3751262421816946e-06,
+      "loss": 0.5298,
+      "step": 36638
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3747655151518346e-06,
+      "loss": 0.5812,
+      "step": 36639
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.374404831949312e-06,
+      "loss": 0.5803,
+      "step": 36640
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3740441925759673e-06,
+      "loss": 0.5879,
+      "step": 36641
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3736835970336282e-06,
+      "loss": 0.5804,
+      "step": 36642
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3733230453241308e-06,
+      "loss": 0.5793,
+      "step": 36643
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.372962537449306e-06,
+      "loss": 0.6313,
+      "step": 36644
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3726020734109846e-06,
+      "loss": 0.6494,
+      "step": 36645
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.372241653210996e-06,
+      "loss": 0.6232,
+      "step": 36646
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3718812768511768e-06,
+      "loss": 0.5797,
+      "step": 36647
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3715209443333543e-06,
+      "loss": 0.5779,
+      "step": 36648
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3711606556593626e-06,
+      "loss": 0.6147,
+      "step": 36649
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3708004108310302e-06,
+      "loss": 0.6255,
+      "step": 36650
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.370440209850189e-06,
+      "loss": 0.5998,
+      "step": 36651
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3700800527186675e-06,
+      "loss": 0.5528,
+      "step": 36652
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3697199394382953e-06,
+      "loss": 0.5831,
+      "step": 36653
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3693598700109056e-06,
+      "loss": 0.6554,
+      "step": 36654
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3689998444383256e-06,
+      "loss": 0.588,
+      "step": 36655
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3686398627223851e-06,
+      "loss": 0.5206,
+      "step": 36656
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3682799248649114e-06,
+      "loss": 0.5511,
+      "step": 36657
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3679200308677364e-06,
+      "loss": 0.6029,
+      "step": 36658
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3675601807326867e-06,
+      "loss": 0.5235,
+      "step": 36659
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3672003744615937e-06,
+      "loss": 0.5421,
+      "step": 36660
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3668406120562838e-06,
+      "loss": 0.5532,
+      "step": 36661
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.3664808935185858e-06,
+      "loss": 0.5617,
+      "step": 36662
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3661212188503237e-06,
+      "loss": 0.6544,
+      "step": 36663
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3657615880533314e-06,
+      "loss": 0.5609,
+      "step": 36664
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.36540200112943e-06,
+      "loss": 0.5746,
+      "step": 36665
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3650424580804545e-06,
+      "loss": 0.6104,
+      "step": 36666
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3646829589082222e-06,
+      "loss": 0.6283,
+      "step": 36667
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3643235036145675e-06,
+      "loss": 0.5893,
+      "step": 36668
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.363964092201312e-06,
+      "loss": 0.5641,
+      "step": 36669
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3636047246702865e-06,
+      "loss": 0.6333,
+      "step": 36670
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3632454010233143e-06,
+      "loss": 0.5863,
+      "step": 36671
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3628861212622224e-06,
+      "loss": 0.6303,
+      "step": 36672
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.362526885388834e-06,
+      "loss": 0.5657,
+      "step": 36673
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3621676934049754e-06,
+      "loss": 0.5952,
+      "step": 36674
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.361808545312474e-06,
+      "loss": 0.6122,
+      "step": 36675
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3614494411131508e-06,
+      "loss": 0.5701,
+      "step": 36676
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3610903808088383e-06,
+      "loss": 0.6379,
+      "step": 36677
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3607313644013498e-06,
+      "loss": 0.5808,
+      "step": 36678
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3603723918925182e-06,
+      "loss": 0.5898,
+      "step": 36679
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3600134632841634e-06,
+      "loss": 0.5849,
+      "step": 36680
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3596545785781113e-06,
+      "loss": 0.6374,
+      "step": 36681
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3592957377761862e-06,
+      "loss": 0.6426,
+      "step": 36682
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3589369408802089e-06,
+      "loss": 0.6548,
+      "step": 36683
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3585781878920013e-06,
+      "loss": 0.602,
+      "step": 36684
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3582194788133919e-06,
+      "loss": 0.5743,
+      "step": 36685
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3578608136461968e-06,
+      "loss": 0.6107,
+      "step": 36686
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.357502192392245e-06,
+      "loss": 0.5927,
+      "step": 36687
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3571436150533546e-06,
+      "loss": 0.5722,
+      "step": 36688
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3567850816313498e-06,
+      "loss": 0.5968,
+      "step": 36689
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3564265921280485e-06,
+      "loss": 0.5797,
+      "step": 36690
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3560681465452773e-06,
+      "loss": 0.5994,
+      "step": 36691
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3557097448848555e-06,
+      "loss": 0.5328,
+      "step": 36692
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3553513871486047e-06,
+      "loss": 0.5903,
+      "step": 36693
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3549930733383443e-06,
+      "loss": 0.6083,
+      "step": 36694
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3546348034558944e-06,
+      "loss": 0.5599,
+      "step": 36695
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3542765775030786e-06,
+      "loss": 0.6522,
+      "step": 36696
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3539183954817125e-06,
+      "loss": 0.6855,
+      "step": 36697
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3535602573936247e-06,
+      "loss": 0.6004,
+      "step": 36698
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3532021632406257e-06,
+      "loss": 0.663,
+      "step": 36699
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3528441130245396e-06,
+      "loss": 0.5933,
+      "step": 36700
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.352486106747184e-06,
+      "loss": 0.587,
+      "step": 36701
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3521281444103796e-06,
+      "loss": 0.5792,
+      "step": 36702
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3517702260159438e-06,
+      "loss": 0.6319,
+      "step": 36703
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3514123515657008e-06,
+      "loss": 0.5853,
+      "step": 36704
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.35105452106146e-06,
+      "loss": 0.5976,
+      "step": 36705
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.350696734505046e-06,
+      "loss": 0.6018,
+      "step": 36706
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3503389918982724e-06,
+      "loss": 0.6098,
+      "step": 36707
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3499812932429623e-06,
+      "loss": 0.6105,
+      "step": 36708
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.34962363854093e-06,
+      "loss": 0.6265,
+      "step": 36709
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.349266027793994e-06,
+      "loss": 0.6089,
+      "step": 36710
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3489084610039681e-06,
+      "loss": 0.5964,
+      "step": 36711
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3485509381726747e-06,
+      "loss": 0.5972,
+      "step": 36712
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3481934593019274e-06,
+      "loss": 0.6111,
+      "step": 36713
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3478360243935407e-06,
+      "loss": 0.6243,
+      "step": 36714
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3474786334493373e-06,
+      "loss": 0.6341,
+      "step": 36715
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3471212864711247e-06,
+      "loss": 0.606,
+      "step": 36716
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3467639834607248e-06,
+      "loss": 0.6397,
+      "step": 36717
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3464067244199496e-06,
+      "loss": 0.5923,
+      "step": 36718
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3460495093506187e-06,
+      "loss": 0.6014,
+      "step": 36719
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.345692338254544e-06,
+      "loss": 0.6257,
+      "step": 36720
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3453352111335415e-06,
+      "loss": 0.6381,
+      "step": 36721
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3449781279894226e-06,
+      "loss": 0.643,
+      "step": 36722
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3446210888240075e-06,
+      "loss": 0.5962,
+      "step": 36723
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3442640936391049e-06,
+      "loss": 0.5943,
+      "step": 36724
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3439071424365358e-06,
+      "loss": 0.6617,
+      "step": 36725
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3435502352181052e-06,
+      "loss": 0.5859,
+      "step": 36726
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3431933719856326e-06,
+      "loss": 0.5886,
+      "step": 36727
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3428365527409293e-06,
+      "loss": 0.5317,
+      "step": 36728
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3424797774858088e-06,
+      "loss": 0.5787,
+      "step": 36729
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3421230462220857e-06,
+      "loss": 0.6193,
+      "step": 36730
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3417663589515705e-06,
+      "loss": 0.6437,
+      "step": 36731
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3414097156760765e-06,
+      "loss": 0.6714,
+      "step": 36732
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3410531163974128e-06,
+      "loss": 0.5899,
+      "step": 36733
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.340696561117396e-06,
+      "loss": 0.6385,
+      "step": 36734
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3403400498378338e-06,
+      "loss": 0.6184,
+      "step": 36735
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3399835825605444e-06,
+      "loss": 0.6017,
+      "step": 36736
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3396271592873312e-06,
+      "loss": 0.5994,
+      "step": 36737
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.33927078002001e-06,
+      "loss": 0.6503,
+      "step": 36738
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3389144447603886e-06,
+      "loss": 0.6527,
+      "step": 36739
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.338558153510281e-06,
+      "loss": 0.6125,
+      "step": 36740
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3382019062714947e-06,
+      "loss": 0.6218,
+      "step": 36741
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3378457030458446e-06,
+      "loss": 0.6411,
+      "step": 36742
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3374895438351332e-06,
+      "loss": 0.5823,
+      "step": 36743
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3371334286411775e-06,
+      "loss": 0.6053,
+      "step": 36744
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3367773574657804e-06,
+      "loss": 0.6364,
+      "step": 36745
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3364213303107577e-06,
+      "loss": 0.657,
+      "step": 36746
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.336065347177915e-06,
+      "loss": 0.6405,
+      "step": 36747
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3357094080690624e-06,
+      "loss": 0.617,
+      "step": 36748
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.335353512986005e-06,
+      "loss": 0.5569,
+      "step": 36749
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.334997661930556e-06,
+      "loss": 0.5902,
+      "step": 36750
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3346418549045214e-06,
+      "loss": 0.603,
+      "step": 36751
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3342860919097101e-06,
+      "loss": 0.5936,
+      "step": 36752
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.333930372947928e-06,
+      "loss": 0.5712,
+      "step": 36753
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3335746980209829e-06,
+      "loss": 0.5988,
+      "step": 36754
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3332190671306832e-06,
+      "loss": 0.6246,
+      "step": 36755
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3328634802788355e-06,
+      "loss": 0.6173,
+      "step": 36756
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3325079374672467e-06,
+      "loss": 0.5521,
+      "step": 36757
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3321524386977247e-06,
+      "loss": 0.5469,
+      "step": 36758
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3317969839720746e-06,
+      "loss": 0.5875,
+      "step": 36759
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3314415732921004e-06,
+      "loss": 0.6042,
+      "step": 36760
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3310862066596131e-06,
+      "loss": 0.5471,
+      "step": 36761
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3307308840764121e-06,
+      "loss": 0.5939,
+      "step": 36762
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3303756055443118e-06,
+      "loss": 0.5992,
+      "step": 36763
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3300203710651073e-06,
+      "loss": 0.5772,
+      "step": 36764
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3296651806406102e-06,
+      "loss": 0.6022,
+      "step": 36765
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3293100342726217e-06,
+      "loss": 0.6337,
+      "step": 36766
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3289549319629502e-06,
+      "loss": 0.6391,
+      "step": 36767
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3285998737133986e-06,
+      "loss": 0.6241,
+      "step": 36768
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3282448595257714e-06,
+      "loss": 0.6354,
+      "step": 36769
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3278898894018677e-06,
+      "loss": 0.6583,
+      "step": 36770
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3275349633434986e-06,
+      "loss": 0.6244,
+      "step": 36771
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3271800813524638e-06,
+      "loss": 0.6274,
+      "step": 36772
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.326825243430565e-06,
+      "loss": 0.5793,
+      "step": 36773
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.326470449579611e-06,
+      "loss": 0.5611,
+      "step": 36774
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.326115699801397e-06,
+      "loss": 0.5817,
+      "step": 36775
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3257609940977311e-06,
+      "loss": 0.5994,
+      "step": 36776
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3254063324704126e-06,
+      "loss": 0.5896,
+      "step": 36777
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3250517149212462e-06,
+      "loss": 0.6318,
+      "step": 36778
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3246971414520305e-06,
+      "loss": 0.6138,
+      "step": 36779
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3243426120645742e-06,
+      "loss": 0.5772,
+      "step": 36780
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3239881267606691e-06,
+      "loss": 0.6297,
+      "step": 36781
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.323633685542124e-06,
+      "loss": 0.573,
+      "step": 36782
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3232792884107349e-06,
+      "loss": 0.6939,
+      "step": 36783
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.322924935368306e-06,
+      "loss": 0.5901,
+      "step": 36784
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3225706264166382e-06,
+      "loss": 0.5904,
+      "step": 36785
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.322216361557529e-06,
+      "loss": 0.6448,
+      "step": 36786
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3218621407927778e-06,
+      "loss": 0.5333,
+      "step": 36787
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3215079641241901e-06,
+      "loss": 0.6116,
+      "step": 36788
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.321153831553561e-06,
+      "loss": 0.5682,
+      "step": 36789
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3207997430826913e-06,
+      "loss": 0.6288,
+      "step": 36790
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3204456987133773e-06,
+      "loss": 0.5713,
+      "step": 36791
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3200916984474232e-06,
+      "loss": 0.6383,
+      "step": 36792
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3197377422866253e-06,
+      "loss": 0.6365,
+      "step": 36793
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3193838302327789e-06,
+      "loss": 0.6007,
+      "step": 36794
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.319029962287689e-06,
+      "loss": 0.5984,
+      "step": 36795
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.31867613845315e-06,
+      "loss": 0.5556,
+      "step": 36796
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.318322358730959e-06,
+      "loss": 0.6449,
+      "step": 36797
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.317968623122914e-06,
+      "loss": 0.5596,
+      "step": 36798
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.317614931630814e-06,
+      "loss": 0.566,
+      "step": 36799
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3172612842564537e-06,
+      "loss": 0.5799,
+      "step": 36800
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3169076810016357e-06,
+      "loss": 0.5871,
+      "step": 36801
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3165541218681477e-06,
+      "loss": 0.6253,
+      "step": 36802
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3162006068577948e-06,
+      "loss": 0.6341,
+      "step": 36803
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3158471359723669e-06,
+      "loss": 0.5999,
+      "step": 36804
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3154937092136654e-06,
+      "loss": 0.6284,
+      "step": 36805
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3151403265834838e-06,
+      "loss": 0.563,
+      "step": 36806
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3147869880836173e-06,
+      "loss": 0.6323,
+      "step": 36807
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3144336937158597e-06,
+      "loss": 0.6258,
+      "step": 36808
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.3140804434820108e-06,
+      "loss": 0.639,
+      "step": 36809
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3137272373838616e-06,
+      "loss": 0.5993,
+      "step": 36810
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3133740754232072e-06,
+      "loss": 0.6141,
+      "step": 36811
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.313020957601846e-06,
+      "loss": 0.6385,
+      "step": 36812
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3126678839215678e-06,
+      "loss": 0.5705,
+      "step": 36813
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.312314854384169e-06,
+      "loss": 0.6454,
+      "step": 36814
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3119618689914405e-06,
+      "loss": 0.5816,
+      "step": 36815
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3116089277451804e-06,
+      "loss": 0.5953,
+      "step": 36816
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.31125603064718e-06,
+      "loss": 0.5877,
+      "step": 36817
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.310903177699232e-06,
+      "loss": 0.5938,
+      "step": 36818
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3105503689031274e-06,
+      "loss": 0.575,
+      "step": 36819
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3101976042606635e-06,
+      "loss": 0.5525,
+      "step": 36820
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3098448837736289e-06,
+      "loss": 0.5882,
+      "step": 36821
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3094922074438187e-06,
+      "loss": 0.6152,
+      "step": 36822
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3091395752730252e-06,
+      "loss": 0.5861,
+      "step": 36823
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3087869872630376e-06,
+      "loss": 0.6426,
+      "step": 36824
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.308434443415646e-06,
+      "loss": 0.54,
+      "step": 36825
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3080819437326476e-06,
+      "loss": 0.6617,
+      "step": 36826
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3077294882158297e-06,
+      "loss": 0.6473,
+      "step": 36827
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3073770768669846e-06,
+      "loss": 0.5985,
+      "step": 36828
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3070247096878997e-06,
+      "loss": 0.5769,
+      "step": 36829
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.306672386680371e-06,
+      "loss": 0.6195,
+      "step": 36830
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3063201078461852e-06,
+      "loss": 0.5868,
+      "step": 36831
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3059678731871306e-06,
+      "loss": 0.6775,
+      "step": 36832
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3056156827050016e-06,
+      "loss": 0.5502,
+      "step": 36833
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3052635364015865e-06,
+      "loss": 0.5927,
+      "step": 36834
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3049114342786727e-06,
+      "loss": 0.6698,
+      "step": 36835
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.304559376338048e-06,
+      "loss": 0.5979,
+      "step": 36836
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3042073625815065e-06,
+      "loss": 0.5908,
+      "step": 36837
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3038553930108323e-06,
+      "loss": 0.5849,
+      "step": 36838
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3035034676278192e-06,
+      "loss": 0.5926,
+      "step": 36839
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3031515864342482e-06,
+      "loss": 0.6268,
+      "step": 36840
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3027997494319134e-06,
+      "loss": 0.6139,
+      "step": 36841
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3024479566225967e-06,
+      "loss": 0.6471,
+      "step": 36842
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3020962080080923e-06,
+      "loss": 0.6265,
+      "step": 36843
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3017445035901843e-06,
+      "loss": 0.6596,
+      "step": 36844
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.301392843370659e-06,
+      "loss": 0.5764,
+      "step": 36845
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.301041227351303e-06,
+      "loss": 0.6165,
+      "step": 36846
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.3006896555339056e-06,
+      "loss": 0.5375,
+      "step": 36847
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.300338127920252e-06,
+      "loss": 0.5917,
+      "step": 36848
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2999866445121268e-06,
+      "loss": 0.6483,
+      "step": 36849
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2996352053113182e-06,
+      "loss": 0.6522,
+      "step": 36850
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2992838103196126e-06,
+      "loss": 0.6166,
+      "step": 36851
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.298932459538793e-06,
+      "loss": 0.6105,
+      "step": 36852
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2985811529706438e-06,
+      "loss": 0.5755,
+      "step": 36853
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2982298906169544e-06,
+      "loss": 0.57,
+      "step": 36854
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.297878672479508e-06,
+      "loss": 0.5079,
+      "step": 36855
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2975274985600873e-06,
+      "loss": 0.5822,
+      "step": 36856
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2971763688604765e-06,
+      "loss": 0.6134,
+      "step": 36857
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2968252833824635e-06,
+      "loss": 0.6052,
+      "step": 36858
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2964742421278276e-06,
+      "loss": 0.5047,
+      "step": 36859
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2961232450983574e-06,
+      "loss": 0.6076,
+      "step": 36860
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.295772292295835e-06,
+      "loss": 0.6659,
+      "step": 36861
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2954213837220408e-06,
+      "loss": 0.6129,
+      "step": 36862
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2950705193787595e-06,
+      "loss": 0.5979,
+      "step": 36863
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2947196992677758e-06,
+      "loss": 0.5869,
+      "step": 36864
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2943689233908696e-06,
+      "loss": 0.6418,
+      "step": 36865
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2940181917498262e-06,
+      "loss": 0.6216,
+      "step": 36866
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.293667504346423e-06,
+      "loss": 0.5815,
+      "step": 36867
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2933168611824475e-06,
+      "loss": 0.5798,
+      "step": 36868
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2929662622596784e-06,
+      "loss": 0.5935,
+      "step": 36869
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.292615707579895e-06,
+      "loss": 0.6378,
+      "step": 36870
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2922651971448852e-06,
+      "loss": 0.6423,
+      "step": 36871
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.291914730956425e-06,
+      "loss": 0.6022,
+      "step": 36872
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2915643090162966e-06,
+      "loss": 0.5675,
+      "step": 36873
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2912139313262774e-06,
+      "loss": 0.6424,
+      "step": 36874
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2908635978881534e-06,
+      "loss": 0.6287,
+      "step": 36875
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2905133087037003e-06,
+      "loss": 0.4983,
+      "step": 36876
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2901630637747043e-06,
+      "loss": 0.5032,
+      "step": 36877
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.289812863102935e-06,
+      "loss": 0.6095,
+      "step": 36878
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.28946270669018e-06,
+      "loss": 0.6683,
+      "step": 36879
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2891125945382143e-06,
+      "loss": 0.5975,
+      "step": 36880
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2887625266488214e-06,
+      "loss": 0.6409,
+      "step": 36881
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.288412503023776e-06,
+      "loss": 0.5926,
+      "step": 36882
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.288062523664858e-06,
+      "loss": 0.5877,
+      "step": 36883
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.287712588573844e-06,
+      "loss": 0.576,
+      "step": 36884
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2873626977525166e-06,
+      "loss": 0.6025,
+      "step": 36885
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2870128512026504e-06,
+      "loss": 0.596,
+      "step": 36886
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2866630489260212e-06,
+      "loss": 0.5576,
+      "step": 36887
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2863132909244113e-06,
+      "loss": 0.6401,
+      "step": 36888
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2859635771995959e-06,
+      "loss": 0.6109,
+      "step": 36889
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2856139077533513e-06,
+      "loss": 0.5952,
+      "step": 36890
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2852642825874517e-06,
+      "loss": 0.5633,
+      "step": 36891
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.284914701703679e-06,
+      "loss": 0.59,
+      "step": 36892
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2845651651038082e-06,
+      "loss": 0.62,
+      "step": 36893
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2842156727896126e-06,
+      "loss": 0.5916,
+      "step": 36894
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2838662247628686e-06,
+      "loss": 0.6172,
+      "step": 36895
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2835168210253547e-06,
+      "loss": 0.6248,
+      "step": 36896
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2831674615788425e-06,
+      "loss": 0.6148,
+      "step": 36897
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.282818146425111e-06,
+      "loss": 0.648,
+      "step": 36898
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2824688755659343e-06,
+      "loss": 0.6444,
+      "step": 36899
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.282119649003085e-06,
+      "loss": 0.5901,
+      "step": 36900
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2817704667383378e-06,
+      "loss": 0.5983,
+      "step": 36901
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.28142132877347e-06,
+      "loss": 0.6076,
+      "step": 36902
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2810722351102522e-06,
+      "loss": 0.6092,
+      "step": 36903
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2807231857504632e-06,
+      "loss": 0.6052,
+      "step": 36904
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2803741806958702e-06,
+      "loss": 0.5531,
+      "step": 36905
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.280025219948251e-06,
+      "loss": 0.6069,
+      "step": 36906
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2796763035093774e-06,
+      "loss": 0.5576,
+      "step": 36907
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.27932743138102e-06,
+      "loss": 0.5703,
+      "step": 36908
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2789786035649565e-06,
+      "loss": 0.6065,
+      "step": 36909
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2786298200629565e-06,
+      "loss": 0.57,
+      "step": 36910
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.278281080876792e-06,
+      "loss": 0.6012,
+      "step": 36911
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2779323860082338e-06,
+      "loss": 0.6681,
+      "step": 36912
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2775837354590582e-06,
+      "loss": 0.6484,
+      "step": 36913
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2772351292310315e-06,
+      "loss": 0.5581,
+      "step": 36914
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2768865673259322e-06,
+      "loss": 0.5763,
+      "step": 36915
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2765380497455227e-06,
+      "loss": 0.6179,
+      "step": 36916
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2761895764915799e-06,
+      "loss": 0.5871,
+      "step": 36917
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2758411475658716e-06,
+      "loss": 0.5769,
+      "step": 36918
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2754927629701707e-06,
+      "loss": 0.5371,
+      "step": 36919
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.275144422706246e-06,
+      "loss": 0.5819,
+      "step": 36920
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.274796126775868e-06,
+      "loss": 0.6189,
+      "step": 36921
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2744478751808042e-06,
+      "loss": 0.582,
+      "step": 36922
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.274099667922828e-06,
+      "loss": 0.7061,
+      "step": 36923
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2737515050037074e-06,
+      "loss": 0.6093,
+      "step": 36924
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2734033864252082e-06,
+      "loss": 0.6142,
+      "step": 36925
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2730553121891044e-06,
+      "loss": 0.6108,
+      "step": 36926
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2727072822971631e-06,
+      "loss": 0.6129,
+      "step": 36927
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2723592967511512e-06,
+      "loss": 0.6324,
+      "step": 36928
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2720113555528358e-06,
+      "loss": 0.6086,
+      "step": 36929
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2716634587039888e-06,
+      "loss": 0.6095,
+      "step": 36930
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2713156062063758e-06,
+      "loss": 0.5852,
+      "step": 36931
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2709677980617653e-06,
+      "loss": 0.5778,
+      "step": 36932
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.270620034271921e-06,
+      "loss": 0.539,
+      "step": 36933
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2702723148386154e-06,
+      "loss": 0.5509,
+      "step": 36934
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.26992463976361e-06,
+      "loss": 0.584,
+      "step": 36935
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2695770090486792e-06,
+      "loss": 0.5688,
+      "step": 36936
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2692294226955793e-06,
+      "loss": 0.615,
+      "step": 36937
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2688818807060843e-06,
+      "loss": 0.6098,
+      "step": 36938
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2685343830819552e-06,
+      "loss": 0.6632,
+      "step": 36939
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2681869298249628e-06,
+      "loss": 0.642,
+      "step": 36940
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2678395209368677e-06,
+      "loss": 0.54,
+      "step": 36941
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2674921564194408e-06,
+      "loss": 0.571,
+      "step": 36942
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2671448362744399e-06,
+      "loss": 0.6003,
+      "step": 36943
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2667975605036353e-06,
+      "loss": 0.6133,
+      "step": 36944
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2664503291087904e-06,
+      "loss": 0.6026,
+      "step": 36945
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.266103142091667e-06,
+      "loss": 0.6106,
+      "step": 36946
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2657559994540336e-06,
+      "loss": 0.5666,
+      "step": 36947
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.265408901197651e-06,
+      "loss": 0.6388,
+      "step": 36948
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2650618473242848e-06,
+      "loss": 0.6085,
+      "step": 36949
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2647148378356955e-06,
+      "loss": 0.5932,
+      "step": 36950
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2643678727336494e-06,
+      "loss": 0.6197,
+      "step": 36951
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2640209520199077e-06,
+      "loss": 0.5501,
+      "step": 36952
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.2636740756962373e-06,
+      "loss": 0.6575,
+      "step": 36953
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.263327243764393e-06,
+      "loss": 0.6603,
+      "step": 36954
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.262980456226145e-06,
+      "loss": 0.5987,
+      "step": 36955
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.262633713083249e-06,
+      "loss": 0.5591,
+      "step": 36956
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2622870143374722e-06,
+      "loss": 0.6626,
+      "step": 36957
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2619403599905734e-06,
+      "loss": 0.5941,
+      "step": 36958
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2615937500443154e-06,
+      "loss": 0.6084,
+      "step": 36959
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.261247184500457e-06,
+      "loss": 0.5887,
+      "step": 36960
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2609006633607624e-06,
+      "loss": 0.5401,
+      "step": 36961
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2605541866269887e-06,
+      "loss": 0.6241,
+      "step": 36962
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2602077543009017e-06,
+      "loss": 0.6197,
+      "step": 36963
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2598613663842574e-06,
+      "loss": 0.5773,
+      "step": 36964
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2595150228788177e-06,
+      "loss": 0.6061,
+      "step": 36965
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2591687237863425e-06,
+      "loss": 0.6242,
+      "step": 36966
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2588224691085881e-06,
+      "loss": 0.5805,
+      "step": 36967
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2584762588473188e-06,
+      "loss": 0.5606,
+      "step": 36968
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2581300930042928e-06,
+      "loss": 0.5532,
+      "step": 36969
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2577839715812667e-06,
+      "loss": 0.6551,
+      "step": 36970
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.257437894579998e-06,
+      "loss": 0.581,
+      "step": 36971
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.257091862002251e-06,
+      "loss": 0.599,
+      "step": 36972
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2567458738497784e-06,
+      "loss": 0.6265,
+      "step": 36973
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2563999301243446e-06,
+      "loss": 0.5117,
+      "step": 36974
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2560540308276992e-06,
+      "loss": 0.5937,
+      "step": 36975
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2557081759616062e-06,
+      "loss": 0.5493,
+      "step": 36976
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.255362365527819e-06,
+      "loss": 0.6433,
+      "step": 36977
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2550165995280984e-06,
+      "loss": 0.6099,
+      "step": 36978
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.254670877964198e-06,
+      "loss": 0.6344,
+      "step": 36979
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2543252008378792e-06,
+      "loss": 0.673,
+      "step": 36980
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2539795681508915e-06,
+      "loss": 0.5749,
+      "step": 36981
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2536339799049978e-06,
+      "loss": 0.5842,
+      "step": 36982
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2532884361019492e-06,
+      "loss": 0.6421,
+      "step": 36983
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2529429367435054e-06,
+      "loss": 0.6061,
+      "step": 36984
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2525974818314202e-06,
+      "loss": 0.6125,
+      "step": 36985
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.252252071367448e-06,
+      "loss": 0.6494,
+      "step": 36986
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2519067053533463e-06,
+      "loss": 0.6153,
+      "step": 36987
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2515613837908658e-06,
+      "loss": 0.6195,
+      "step": 36988
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2512161066817662e-06,
+      "loss": 0.5484,
+      "step": 36989
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2508708740277985e-06,
+      "loss": 0.551,
+      "step": 36990
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2505256858307212e-06,
+      "loss": 0.606,
+      "step": 36991
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.250180542092282e-06,
+      "loss": 0.656,
+      "step": 36992
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.24983544281424e-06,
+      "loss": 0.6044,
+      "step": 36993
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2494903879983444e-06,
+      "loss": 0.5391,
+      "step": 36994
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2491453776463524e-06,
+      "loss": 0.6257,
+      "step": 36995
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2488004117600161e-06,
+      "loss": 0.595,
+      "step": 36996
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2484554903410884e-06,
+      "loss": 0.6065,
+      "step": 36997
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2481106133913178e-06,
+      "loss": 0.5877,
+      "step": 36998
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2477657809124632e-06,
+      "loss": 0.6296,
+      "step": 36999
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.247420992906272e-06,
+      "loss": 0.6147,
+      "step": 37000
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2470762493745025e-06,
+      "loss": 0.6262,
+      "step": 37001
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.246731550318897e-06,
+      "loss": 0.5585,
+      "step": 37002
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.246386895741214e-06,
+      "loss": 0.5948,
+      "step": 37003
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2460422856432019e-06,
+      "loss": 0.6087,
+      "step": 37004
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.245697720026614e-06,
+      "loss": 0.53,
+      "step": 37005
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2453531988931999e-06,
+      "loss": 0.6282,
+      "step": 37006
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2450087222447093e-06,
+      "loss": 0.605,
+      "step": 37007
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2446642900828942e-06,
+      "loss": 0.6196,
+      "step": 37008
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.244319902409502e-06,
+      "loss": 0.5905,
+      "step": 37009
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.243975559226286e-06,
+      "loss": 0.5776,
+      "step": 37010
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2436312605349931e-06,
+      "loss": 0.6165,
+      "step": 37011
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.243287006337378e-06,
+      "loss": 0.5347,
+      "step": 37012
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2429427966351826e-06,
+      "loss": 0.5801,
+      "step": 37013
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2425986314301608e-06,
+      "loss": 0.5292,
+      "step": 37014
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.242254510724058e-06,
+      "loss": 0.6099,
+      "step": 37015
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2419104345186273e-06,
+      "loss": 0.6334,
+      "step": 37016
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2415664028156127e-06,
+      "loss": 0.6719,
+      "step": 37017
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2412224156167674e-06,
+      "loss": 0.5984,
+      "step": 37018
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2408784729238333e-06,
+      "loss": 0.5948,
+      "step": 37019
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2405345747385622e-06,
+      "loss": 0.5705,
+      "step": 37020
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2401907210626983e-06,
+      "loss": 0.6336,
+      "step": 37021
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2398469118979928e-06,
+      "loss": 0.6355,
+      "step": 37022
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2395031472461904e-06,
+      "loss": 0.6262,
+      "step": 37023
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.239159427109039e-06,
+      "loss": 0.4731,
+      "step": 37024
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2388157514882838e-06,
+      "loss": 0.6002,
+      "step": 37025
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2384721203856697e-06,
+      "loss": 0.6027,
+      "step": 37026
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2381285338029458e-06,
+      "loss": 0.6266,
+      "step": 37027
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2377849917418583e-06,
+      "loss": 0.616,
+      "step": 37028
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2374414942041501e-06,
+      "loss": 0.6321,
+      "step": 37029
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2370980411915666e-06,
+      "loss": 0.5818,
+      "step": 37030
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2367546327058566e-06,
+      "loss": 0.6126,
+      "step": 37031
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2364112687487605e-06,
+      "loss": 0.589,
+      "step": 37032
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2360679493220273e-06,
+      "loss": 0.5547,
+      "step": 37033
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2357246744273987e-06,
+      "loss": 0.599,
+      "step": 37034
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.23538144406662e-06,
+      "loss": 0.6238,
+      "step": 37035
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2350382582414333e-06,
+      "loss": 0.6409,
+      "step": 37036
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.234695116953587e-06,
+      "loss": 0.6151,
+      "step": 37037
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2343520202048187e-06,
+      "loss": 0.5928,
+      "step": 37038
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2340089679968792e-06,
+      "loss": 0.5952,
+      "step": 37039
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2336659603315038e-06,
+      "loss": 0.6684,
+      "step": 37040
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.233322997210441e-06,
+      "loss": 0.5918,
+      "step": 37041
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2329800786354285e-06,
+      "loss": 0.5382,
+      "step": 37042
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2326372046082147e-06,
+      "loss": 0.5826,
+      "step": 37043
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2322943751305382e-06,
+      "loss": 0.6169,
+      "step": 37044
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2319515902041423e-06,
+      "loss": 0.5911,
+      "step": 37045
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2316088498307689e-06,
+      "loss": 0.5981,
+      "step": 37046
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2312661540121551e-06,
+      "loss": 0.6283,
+      "step": 37047
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2309235027500499e-06,
+      "loss": 0.6007,
+      "step": 37048
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2305808960461873e-06,
+      "loss": 0.5837,
+      "step": 37049
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2302383339023162e-06,
+      "loss": 0.6197,
+      "step": 37050
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.229895816320167e-06,
+      "loss": 0.5898,
+      "step": 37051
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2295533433014884e-06,
+      "loss": 0.5714,
+      "step": 37052
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2292109148480157e-06,
+      "loss": 0.546,
+      "step": 37053
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2288685309614933e-06,
+      "loss": 0.6343,
+      "step": 37054
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2285261916436576e-06,
+      "loss": 0.582,
+      "step": 37055
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2281838968962523e-06,
+      "loss": 0.609,
+      "step": 37056
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2278416467210086e-06,
+      "loss": 0.5522,
+      "step": 37057
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.227499441119674e-06,
+      "loss": 0.605,
+      "step": 37058
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2271572800939812e-06,
+      "loss": 0.6673,
+      "step": 37059
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2268151636456726e-06,
+      "loss": 0.5746,
+      "step": 37060
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2264730917764867e-06,
+      "loss": 0.5915,
+      "step": 37061
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2261310644881608e-06,
+      "loss": 0.6484,
+      "step": 37062
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2257890817824292e-06,
+      "loss": 0.6751,
+      "step": 37063
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.225447143661036e-06,
+      "loss": 0.564,
+      "step": 37064
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2251052501257155e-06,
+      "loss": 0.6134,
+      "step": 37065
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.224763401178205e-06,
+      "loss": 0.5981,
+      "step": 37066
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2244215968202412e-06,
+      "loss": 0.5765,
+      "step": 37067
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.22407983705356e-06,
+      "loss": 0.5885,
+      "step": 37068
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2237381218799005e-06,
+      "loss": 0.5215,
+      "step": 37069
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2233964513009967e-06,
+      "loss": 0.593,
+      "step": 37070
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.223054825318587e-06,
+      "loss": 0.6449,
+      "step": 37071
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.222713243934407e-06,
+      "loss": 0.6178,
+      "step": 37072
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2223717071501907e-06,
+      "loss": 0.5191,
+      "step": 37073
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.222030214967672e-06,
+      "loss": 0.603,
+      "step": 37074
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2216887673885924e-06,
+      "loss": 0.6297,
+      "step": 37075
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2213473644146789e-06,
+      "loss": 0.5906,
+      "step": 37076
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.221006006047676e-06,
+      "loss": 0.6661,
+      "step": 37077
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2206646922893084e-06,
+      "loss": 0.6164,
+      "step": 37078
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2203234231413164e-06,
+      "loss": 0.61,
+      "step": 37079
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2199821986054295e-06,
+      "loss": 0.5959,
+      "step": 37080
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2196410186833873e-06,
+      "loss": 0.6026,
+      "step": 37081
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.219299883376921e-06,
+      "loss": 0.6138,
+      "step": 37082
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2189587926877623e-06,
+      "loss": 0.5807,
+      "step": 37083
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.218617746617644e-06,
+      "loss": 0.5968,
+      "step": 37084
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.218276745168303e-06,
+      "loss": 0.5815,
+      "step": 37085
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.21793578834147e-06,
+      "loss": 0.6247,
+      "step": 37086
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2175948761388745e-06,
+      "loss": 0.5281,
+      "step": 37087
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2172540085622565e-06,
+      "loss": 0.5649,
+      "step": 37088
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2169131856133387e-06,
+      "loss": 0.5821,
+      "step": 37089
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.216572407293859e-06,
+      "loss": 0.5975,
+      "step": 37090
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2162316736055446e-06,
+      "loss": 0.6399,
+      "step": 37091
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.215890984550132e-06,
+      "loss": 0.6283,
+      "step": 37092
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.21555034012935e-06,
+      "loss": 0.597,
+      "step": 37093
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2152097403449292e-06,
+      "loss": 0.5968,
+      "step": 37094
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2148691851985982e-06,
+      "loss": 0.6316,
+      "step": 37095
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2145286746920926e-06,
+      "loss": 0.5875,
+      "step": 37096
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2141882088271374e-06,
+      "loss": 0.5762,
+      "step": 37097
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2138477876054667e-06,
+      "loss": 0.5918,
+      "step": 37098
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.213507411028808e-06,
+      "loss": 0.6086,
+      "step": 37099
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2131670790988926e-06,
+      "loss": 0.5554,
+      "step": 37100
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2128267918174452e-06,
+      "loss": 0.582,
+      "step": 37101
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.2124865491862014e-06,
+      "loss": 0.713,
+      "step": 37102
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2121463512068864e-06,
+      "loss": 0.5501,
+      "step": 37103
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2118061978812301e-06,
+      "loss": 0.5624,
+      "step": 37104
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2114660892109598e-06,
+      "loss": 0.5808,
+      "step": 37105
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2111260251978019e-06,
+      "loss": 0.6505,
+      "step": 37106
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2107860058434895e-06,
+      "loss": 0.5786,
+      "step": 37107
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2104460311497445e-06,
+      "loss": 0.6412,
+      "step": 37108
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2101061011183e-06,
+      "loss": 0.5874,
+      "step": 37109
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2097662157508805e-06,
+      "loss": 0.6156,
+      "step": 37110
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.209426375049213e-06,
+      "loss": 0.6469,
+      "step": 37111
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2090865790150218e-06,
+      "loss": 0.6108,
+      "step": 37112
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2087468276500392e-06,
+      "loss": 0.5766,
+      "step": 37113
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2084071209559867e-06,
+      "loss": 0.5963,
+      "step": 37114
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2080674589345965e-06,
+      "loss": 0.611,
+      "step": 37115
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2077278415875859e-06,
+      "loss": 0.569,
+      "step": 37116
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.207388268916687e-06,
+      "loss": 0.6162,
+      "step": 37117
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2070487409236209e-06,
+      "loss": 0.6023,
+      "step": 37118
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.206709257610118e-06,
+      "loss": 0.5801,
+      "step": 37119
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2063698189779006e-06,
+      "loss": 0.591,
+      "step": 37120
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2060304250286936e-06,
+      "loss": 0.5886,
+      "step": 37121
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2056910757642193e-06,
+      "loss": 0.5811,
+      "step": 37122
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2053517711862062e-06,
+      "loss": 0.6067,
+      "step": 37123
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2050125112963763e-06,
+      "loss": 0.6219,
+      "step": 37124
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2046732960964514e-06,
+      "loss": 0.5646,
+      "step": 37125
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2043341255881613e-06,
+      "loss": 0.5877,
+      "step": 37126
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2039949997732214e-06,
+      "loss": 0.6477,
+      "step": 37127
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2036559186533614e-06,
+      "loss": 0.6061,
+      "step": 37128
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2033168822302988e-06,
+      "loss": 0.602,
+      "step": 37129
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2029778905057621e-06,
+      "loss": 0.5411,
+      "step": 37130
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2026389434814712e-06,
+      "loss": 0.5935,
+      "step": 37131
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.202300041159148e-06,
+      "loss": 0.5706,
+      "step": 37132
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2019611835405131e-06,
+      "loss": 0.5988,
+      "step": 37133
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2016223706272922e-06,
+      "loss": 0.5905,
+      "step": 37134
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2012836024212016e-06,
+      "loss": 0.4816,
+      "step": 37135
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2009448789239687e-06,
+      "loss": 0.5802,
+      "step": 37136
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.200606200137312e-06,
+      "loss": 0.6212,
+      "step": 37137
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.2002675660629526e-06,
+      "loss": 0.5631,
+      "step": 37138
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.199928976702608e-06,
+      "loss": 0.6281,
+      "step": 37139
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1995904320580033e-06,
+      "loss": 0.572,
+      "step": 37140
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.199251932130857e-06,
+      "loss": 0.5964,
+      "step": 37141
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1989134769228894e-06,
+      "loss": 0.5933,
+      "step": 37142
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1985750664358175e-06,
+      "loss": 0.5391,
+      "step": 37143
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1982367006713657e-06,
+      "loss": 0.5444,
+      "step": 37144
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1978983796312515e-06,
+      "loss": 0.6259,
+      "step": 37145
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1975601033171914e-06,
+      "loss": 0.5773,
+      "step": 37146
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.197221871730907e-06,
+      "loss": 0.6091,
+      "step": 37147
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1968836848741173e-06,
+      "loss": 0.6,
+      "step": 37148
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1965455427485395e-06,
+      "loss": 0.6033,
+      "step": 37149
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1962074453558903e-06,
+      "loss": 0.5763,
+      "step": 37150
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1958693926978903e-06,
+      "loss": 0.5982,
+      "step": 37151
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.195531384776255e-06,
+      "loss": 0.6824,
+      "step": 37152
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1951934215927085e-06,
+      "loss": 0.6216,
+      "step": 37153
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1948555031489573e-06,
+      "loss": 0.6606,
+      "step": 37154
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1945176294467265e-06,
+      "loss": 0.5802,
+      "step": 37155
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.194179800487728e-06,
+      "loss": 0.6279,
+      "step": 37156
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.193842016273683e-06,
+      "loss": 0.5177,
+      "step": 37157
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1935042768063065e-06,
+      "loss": 0.5523,
+      "step": 37158
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1931665820873128e-06,
+      "loss": 0.5275,
+      "step": 37159
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1928289321184173e-06,
+      "loss": 0.5999,
+      "step": 37160
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1924913269013393e-06,
+      "loss": 0.5649,
+      "step": 37161
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1921537664377924e-06,
+      "loss": 0.6098,
+      "step": 37162
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1918162507294894e-06,
+      "loss": 0.6005,
+      "step": 37163
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1914787797781503e-06,
+      "loss": 0.6054,
+      "step": 37164
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1911413535854866e-06,
+      "loss": 0.6009,
+      "step": 37165
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1908039721532138e-06,
+      "loss": 0.59,
+      "step": 37166
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.190466635483043e-06,
+      "loss": 0.5948,
+      "step": 37167
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1901293435766948e-06,
+      "loss": 0.5956,
+      "step": 37168
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1897920964358778e-06,
+      "loss": 0.5804,
+      "step": 37169
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1894548940623085e-06,
+      "loss": 0.6257,
+      "step": 37170
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1891177364576967e-06,
+      "loss": 0.5945,
+      "step": 37171
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1887806236237609e-06,
+      "loss": 0.6438,
+      "step": 37172
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1884435555622076e-06,
+      "loss": 0.6361,
+      "step": 37173
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1881065322747564e-06,
+      "loss": 0.5822,
+      "step": 37174
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1877695537631162e-06,
+      "loss": 0.5765,
+      "step": 37175
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1874326200289999e-06,
+      "loss": 0.5781,
+      "step": 37176
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1870957310741172e-06,
+      "loss": 0.52,
+      "step": 37177
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1867588869001844e-06,
+      "loss": 0.5083,
+      "step": 37178
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1864220875089105e-06,
+      "loss": 0.5984,
+      "step": 37179
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1860853329020072e-06,
+      "loss": 0.5113,
+      "step": 37180
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1857486230811833e-06,
+      "loss": 0.6002,
+      "step": 37181
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.185411958048155e-06,
+      "loss": 0.6421,
+      "step": 37182
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1850753378046287e-06,
+      "loss": 0.5931,
+      "step": 37183
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1847387623523143e-06,
+      "loss": 0.6621,
+      "step": 37184
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1844022316929271e-06,
+      "loss": 0.6093,
+      "step": 37185
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1840657458281735e-06,
+      "loss": 0.654,
+      "step": 37186
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1837293047597631e-06,
+      "loss": 0.6103,
+      "step": 37187
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1833929084894047e-06,
+      "loss": 0.5874,
+      "step": 37188
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.18305655701881e-06,
+      "loss": 0.5951,
+      "step": 37189
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.182720250349686e-06,
+      "loss": 0.544,
+      "step": 37190
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1823839884837463e-06,
+      "loss": 0.5888,
+      "step": 37191
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.182047771422692e-06,
+      "loss": 0.5904,
+      "step": 37192
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1817115991682382e-06,
+      "loss": 0.6429,
+      "step": 37193
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1813754717220872e-06,
+      "loss": 0.6382,
+      "step": 37194
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.181039389085953e-06,
+      "loss": 0.6021,
+      "step": 37195
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1807033512615396e-06,
+      "loss": 0.6294,
+      "step": 37196
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1803673582505548e-06,
+      "loss": 0.6024,
+      "step": 37197
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1800314100547062e-06,
+      "loss": 0.5493,
+      "step": 37198
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.179695506675702e-06,
+      "loss": 0.5709,
+      "step": 37199
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.179359648115248e-06,
+      "loss": 0.6239,
+      "step": 37200
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1790238343750494e-06,
+      "loss": 0.58,
+      "step": 37201
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1786880654568155e-06,
+      "loss": 0.5685,
+      "step": 37202
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.178352341362251e-06,
+      "loss": 0.5767,
+      "step": 37203
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.178016662093062e-06,
+      "loss": 0.5725,
+      "step": 37204
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.177681027650952e-06,
+      "loss": 0.6057,
+      "step": 37205
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1773454380376303e-06,
+      "loss": 0.5958,
+      "step": 37206
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1770098932548001e-06,
+      "loss": 0.5873,
+      "step": 37207
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1766743933041668e-06,
+      "loss": 0.5451,
+      "step": 37208
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1763389381874324e-06,
+      "loss": 0.6119,
+      "step": 37209
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1760035279063064e-06,
+      "loss": 0.6147,
+      "step": 37210
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1756681624624888e-06,
+      "loss": 0.6174,
+      "step": 37211
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1753328418576903e-06,
+      "loss": 0.5985,
+      "step": 37212
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1749975660936052e-06,
+      "loss": 0.6477,
+      "step": 37213
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1746623351719443e-06,
+      "loss": 0.6463,
+      "step": 37214
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1743271490944063e-06,
+      "loss": 0.626,
+      "step": 37215
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1739920078626986e-06,
+      "loss": 0.5788,
+      "step": 37216
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1736569114785234e-06,
+      "loss": 0.6602,
+      "step": 37217
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1733218599435814e-06,
+      "loss": 0.6437,
+      "step": 37218
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1729868532595746e-06,
+      "loss": 0.6161,
+      "step": 37219
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.172651891428208e-06,
+      "loss": 0.659,
+      "step": 37220
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.172316974451183e-06,
+      "loss": 0.5917,
+      "step": 37221
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.171982102330198e-06,
+      "loss": 0.6728,
+      "step": 37222
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1716472750669605e-06,
+      "loss": 0.597,
+      "step": 37223
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.171312492663168e-06,
+      "loss": 0.5806,
+      "step": 37224
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1709777551205226e-06,
+      "loss": 0.6025,
+      "step": 37225
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1706430624407229e-06,
+      "loss": 0.6536,
+      "step": 37226
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.170308414625473e-06,
+      "loss": 0.6122,
+      "step": 37227
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1699738116764704e-06,
+      "loss": 0.596,
+      "step": 37228
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1696392535954205e-06,
+      "loss": 0.5473,
+      "step": 37229
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1693047403840164e-06,
+      "loss": 0.5608,
+      "step": 37230
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1689702720439622e-06,
+      "loss": 0.5463,
+      "step": 37231
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1686358485769556e-06,
+      "loss": 0.6325,
+      "step": 37232
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1683014699846972e-06,
+      "loss": 0.6194,
+      "step": 37233
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.167967136268886e-06,
+      "loss": 0.6063,
+      "step": 37234
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1676328474312216e-06,
+      "loss": 0.6484,
+      "step": 37235
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.167298603473398e-06,
+      "loss": 0.5926,
+      "step": 37236
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1669644043971185e-06,
+      "loss": 0.6351,
+      "step": 37237
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1666302502040805e-06,
+      "loss": 0.5885,
+      "step": 37238
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1662961408959795e-06,
+      "loss": 0.6052,
+      "step": 37239
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1659620764745172e-06,
+      "loss": 0.5953,
+      "step": 37240
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.165628056941388e-06,
+      "loss": 0.5986,
+      "step": 37241
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1652940822982894e-06,
+      "loss": 0.6817,
+      "step": 37242
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1649601525469179e-06,
+      "loss": 0.6174,
+      "step": 37243
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.164626267688973e-06,
+      "loss": 0.6321,
+      "step": 37244
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1642924277261492e-06,
+      "loss": 0.6102,
+      "step": 37245
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1639586326601437e-06,
+      "loss": 0.5436,
+      "step": 37246
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1636248824926498e-06,
+      "loss": 0.6219,
+      "step": 37247
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1632911772253686e-06,
+      "loss": 0.6129,
+      "step": 37248
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.1629575168599904e-06,
+      "loss": 0.559,
+      "step": 37249
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1626239013982165e-06,
+      "loss": 0.5739,
+      "step": 37250
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1622903308417355e-06,
+      "loss": 0.5887,
+      "step": 37251
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.161956805192247e-06,
+      "loss": 0.5629,
+      "step": 37252
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.161623324451443e-06,
+      "loss": 0.5593,
+      "step": 37253
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.16128988862102e-06,
+      "loss": 0.6145,
+      "step": 37254
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.160956497702671e-06,
+      "loss": 0.5847,
+      "step": 37255
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1606231516980936e-06,
+      "loss": 0.6174,
+      "step": 37256
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1602898506089756e-06,
+      "loss": 0.6492,
+      "step": 37257
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.159956594437015e-06,
+      "loss": 0.5809,
+      "step": 37258
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1596233831839043e-06,
+      "loss": 0.5588,
+      "step": 37259
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1592902168513343e-06,
+      "loss": 0.545,
+      "step": 37260
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1589570954410024e-06,
+      "loss": 0.6414,
+      "step": 37261
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1586240189545971e-06,
+      "loss": 0.6011,
+      "step": 37262
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1582909873938142e-06,
+      "loss": 0.6364,
+      "step": 37263
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.157958000760341e-06,
+      "loss": 0.5883,
+      "step": 37264
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1576250590558746e-06,
+      "loss": 0.6385,
+      "step": 37265
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1572921622821042e-06,
+      "loss": 0.5629,
+      "step": 37266
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.156959310440725e-06,
+      "loss": 0.5581,
+      "step": 37267
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1566265035334213e-06,
+      "loss": 0.6084,
+      "step": 37268
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1562937415618902e-06,
+      "loss": 0.5855,
+      "step": 37269
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1559610245278196e-06,
+      "loss": 0.5403,
+      "step": 37270
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1556283524329026e-06,
+      "loss": 0.6144,
+      "step": 37271
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1552957252788278e-06,
+      "loss": 0.5879,
+      "step": 37272
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.154963143067286e-06,
+      "loss": 0.6009,
+      "step": 37273
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1546306057999645e-06,
+      "loss": 0.6184,
+      "step": 37274
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1542981134785568e-06,
+      "loss": 0.588,
+      "step": 37275
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1539656661047505e-06,
+      "loss": 0.6204,
+      "step": 37276
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1536332636802383e-06,
+      "loss": 0.6394,
+      "step": 37277
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1533009062067024e-06,
+      "loss": 0.6256,
+      "step": 37278
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.152968593685838e-06,
+      "loss": 0.5741,
+      "step": 37279
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1526363261193307e-06,
+      "loss": 0.6084,
+      "step": 37280
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1523041035088679e-06,
+      "loss": 0.616,
+      "step": 37281
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.15197192585614e-06,
+      "loss": 0.5873,
+      "step": 37282
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1516397931628342e-06,
+      "loss": 0.5612,
+      "step": 37283
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1513077054306387e-06,
+      "loss": 0.6071,
+      "step": 37284
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1509756626612378e-06,
+      "loss": 0.5796,
+      "step": 37285
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1506436648563235e-06,
+      "loss": 0.5792,
+      "step": 37286
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1503117120175778e-06,
+      "loss": 0.6601,
+      "step": 37287
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1499798041466947e-06,
+      "loss": 0.6077,
+      "step": 37288
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1496479412453521e-06,
+      "loss": 0.5502,
+      "step": 37289
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1493161233152416e-06,
+      "loss": 0.6356,
+      "step": 37290
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1489843503580456e-06,
+      "loss": 0.6102,
+      "step": 37291
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1486526223754558e-06,
+      "loss": 0.6619,
+      "step": 37292
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1483209393691508e-06,
+      "loss": 0.6807,
+      "step": 37293
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.147989301340824e-06,
+      "loss": 0.5645,
+      "step": 37294
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1476577082921513e-06,
+      "loss": 0.6013,
+      "step": 37295
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1473261602248255e-06,
+      "loss": 0.6195,
+      "step": 37296
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1469946571405276e-06,
+      "loss": 0.5985,
+      "step": 37297
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1466631990409405e-06,
+      "loss": 0.5856,
+      "step": 37298
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1463317859277524e-06,
+      "loss": 0.568,
+      "step": 37299
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.146000417802645e-06,
+      "loss": 0.6336,
+      "step": 37300
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1456690946673032e-06,
+      "loss": 0.5704,
+      "step": 37301
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1453378165234086e-06,
+      "loss": 0.6169,
+      "step": 37302
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1450065833726465e-06,
+      "loss": 0.6748,
+      "step": 37303
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1446753952167e-06,
+      "loss": 0.6189,
+      "step": 37304
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1443442520572501e-06,
+      "loss": 0.5965,
+      "step": 37305
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1440131538959797e-06,
+      "loss": 0.6036,
+      "step": 37306
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1436821007345733e-06,
+      "loss": 0.614,
+      "step": 37307
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1433510925747103e-06,
+      "loss": 0.5478,
+      "step": 37308
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1430201294180766e-06,
+      "loss": 0.6159,
+      "step": 37309
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1426892112663502e-06,
+      "loss": 0.5653,
+      "step": 37310
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1423583381212144e-06,
+      "loss": 0.6355,
+      "step": 37311
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.142027509984348e-06,
+      "loss": 0.557,
+      "step": 37312
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1416967268574364e-06,
+      "loss": 0.5978,
+      "step": 37313
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1413659887421557e-06,
+      "loss": 0.5686,
+      "step": 37314
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1410352956401938e-06,
+      "loss": 0.5846,
+      "step": 37315
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1407046475532213e-06,
+      "loss": 0.6143,
+      "step": 37316
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1403740444829248e-06,
+      "loss": 0.528,
+      "step": 37317
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1400434864309817e-06,
+      "loss": 0.5205,
+      "step": 37318
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.139712973399072e-06,
+      "loss": 0.6324,
+      "step": 37319
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1393825053888762e-06,
+      "loss": 0.5686,
+      "step": 37320
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1390520824020735e-06,
+      "loss": 0.6077,
+      "step": 37321
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.138721704440342e-06,
+      "loss": 0.5531,
+      "step": 37322
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1383913715053597e-06,
+      "loss": 0.6584,
+      "step": 37323
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1380610835988071e-06,
+      "loss": 0.6029,
+      "step": 37324
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1377308407223586e-06,
+      "loss": 0.6031,
+      "step": 37325
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1374006428777007e-06,
+      "loss": 0.5797,
+      "step": 37326
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1370704900665007e-06,
+      "loss": 0.6093,
+      "step": 37327
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1367403822904444e-06,
+      "loss": 0.5709,
+      "step": 37328
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1364103195512021e-06,
+      "loss": 0.6247,
+      "step": 37329
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1360803018504584e-06,
+      "loss": 0.5343,
+      "step": 37330
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1357503291898842e-06,
+      "loss": 0.5845,
+      "step": 37331
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1354204015711623e-06,
+      "loss": 0.6264,
+      "step": 37332
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1350905189959615e-06,
+      "loss": 0.6547,
+      "step": 37333
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.134760681465964e-06,
+      "loss": 0.6173,
+      "step": 37334
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1344308889828415e-06,
+      "loss": 0.6266,
+      "step": 37335
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.134101141548274e-06,
+      "loss": 0.5749,
+      "step": 37336
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1337714391639355e-06,
+      "loss": 0.613,
+      "step": 37337
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1334417818315014e-06,
+      "loss": 0.5857,
+      "step": 37338
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.133112169552646e-06,
+      "loss": 0.5796,
+      "step": 37339
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1327826023290422e-06,
+      "loss": 0.6137,
+      "step": 37340
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.132453080162368e-06,
+      "loss": 0.5965,
+      "step": 37341
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.132123603054298e-06,
+      "loss": 0.5841,
+      "step": 37342
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.131794171006505e-06,
+      "loss": 0.608,
+      "step": 37343
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1314647840206604e-06,
+      "loss": 0.5699,
+      "step": 37344
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1311354420984423e-06,
+      "loss": 0.6328,
+      "step": 37345
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1308061452415208e-06,
+      "loss": 0.6008,
+      "step": 37346
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1304768934515731e-06,
+      "loss": 0.5709,
+      "step": 37347
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1301476867302696e-06,
+      "loss": 0.661,
+      "step": 37348
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1298185250792826e-06,
+      "loss": 0.5543,
+      "step": 37349
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.129489408500285e-06,
+      "loss": 0.6263,
+      "step": 37350
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.12916033699495e-06,
+      "loss": 0.5418,
+      "step": 37351
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1288313105649485e-06,
+      "loss": 0.6291,
+      "step": 37352
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1285023292119568e-06,
+      "loss": 0.5481,
+      "step": 37353
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1281733929376383e-06,
+      "loss": 0.5572,
+      "step": 37354
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1278445017436712e-06,
+      "loss": 0.5789,
+      "step": 37355
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1275156556317224e-06,
+      "loss": 0.6366,
+      "step": 37356
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1271868546034681e-06,
+      "loss": 0.6046,
+      "step": 37357
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1268580986605747e-06,
+      "loss": 0.5786,
+      "step": 37358
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1265293878047145e-06,
+      "loss": 0.6646,
+      "step": 37359
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.126200722037557e-06,
+      "loss": 0.5797,
+      "step": 37360
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.125872101360771e-06,
+      "loss": 0.5937,
+      "step": 37361
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1255435257760284e-06,
+      "loss": 0.6103,
+      "step": 37362
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1252149952849977e-06,
+      "loss": 0.6136,
+      "step": 37363
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1248865098893512e-06,
+      "loss": 0.5452,
+      "step": 37364
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1245580695907532e-06,
+      "loss": 0.5854,
+      "step": 37365
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1242296743908755e-06,
+      "loss": 0.6071,
+      "step": 37366
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1239013242913854e-06,
+      "loss": 0.5832,
+      "step": 37367
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.123573019293953e-06,
+      "loss": 0.5796,
+      "step": 37368
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.123244759400246e-06,
+      "loss": 0.6202,
+      "step": 37369
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1229165446119328e-06,
+      "loss": 0.6308,
+      "step": 37370
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1225883749306777e-06,
+      "loss": 0.5814,
+      "step": 37371
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1222602503581537e-06,
+      "loss": 0.5984,
+      "step": 37372
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1219321708960218e-06,
+      "loss": 0.6083,
+      "step": 37373
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1216041365459563e-06,
+      "loss": 0.5672,
+      "step": 37374
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.121276147309619e-06,
+      "loss": 0.6266,
+      "step": 37375
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1209482031886787e-06,
+      "loss": 0.6415,
+      "step": 37376
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1206203041847985e-06,
+      "loss": 0.5972,
+      "step": 37377
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1202924502996494e-06,
+      "loss": 0.6782,
+      "step": 37378
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1199646415348952e-06,
+      "loss": 0.6155,
+      "step": 37379
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1196368778922006e-06,
+      "loss": 0.6036,
+      "step": 37380
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1193091593732318e-06,
+      "loss": 0.6266,
+      "step": 37381
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1189814859796521e-06,
+      "loss": 0.5719,
+      "step": 37382
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1186538577131302e-06,
+      "loss": 0.5624,
+      "step": 37383
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1183262745753276e-06,
+      "loss": 0.6122,
+      "step": 37384
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1179987365679123e-06,
+      "loss": 0.6437,
+      "step": 37385
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1176712436925463e-06,
+      "loss": 0.6441,
+      "step": 37386
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1173437959508936e-06,
+      "loss": 0.5878,
+      "step": 37387
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1170163933446176e-06,
+      "loss": 0.6099,
+      "step": 37388
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1166890358753835e-06,
+      "loss": 0.6261,
+      "step": 37389
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.116361723544853e-06,
+      "loss": 0.5903,
+      "step": 37390
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1160344563546942e-06,
+      "loss": 0.6404,
+      "step": 37391
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.115707234306561e-06,
+      "loss": 0.6769,
+      "step": 37392
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1153800574021246e-06,
+      "loss": 0.631,
+      "step": 37393
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.115052925643042e-06,
+      "loss": 0.551,
+      "step": 37394
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1147258390309789e-06,
+      "loss": 0.5505,
+      "step": 37395
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.1143987975675962e-06,
+      "loss": 0.566,
+      "step": 37396
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.114071801254556e-06,
+      "loss": 0.5706,
+      "step": 37397
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1137448500935189e-06,
+      "loss": 0.6726,
+      "step": 37398
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1134179440861438e-06,
+      "loss": 0.594,
+      "step": 37399
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.113091083234098e-06,
+      "loss": 0.6033,
+      "step": 37400
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.112764267539036e-06,
+      "loss": 0.559,
+      "step": 37401
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1124374970026263e-06,
+      "loss": 0.5979,
+      "step": 37402
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1121107716265212e-06,
+      "loss": 0.5675,
+      "step": 37403
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1117840914123844e-06,
+      "loss": 0.6319,
+      "step": 37404
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1114574563618752e-06,
+      "loss": 0.599,
+      "step": 37405
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.111130866476655e-06,
+      "loss": 0.6242,
+      "step": 37406
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.110804321758382e-06,
+      "loss": 0.5685,
+      "step": 37407
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1104778222087166e-06,
+      "loss": 0.6831,
+      "step": 37408
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.110151367829313e-06,
+      "loss": 0.5979,
+      "step": 37409
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1098249586218379e-06,
+      "loss": 0.6452,
+      "step": 37410
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1094985945879422e-06,
+      "loss": 0.5707,
+      "step": 37411
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1091722757292899e-06,
+      "loss": 0.5963,
+      "step": 37412
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1088460020475377e-06,
+      "loss": 0.6558,
+      "step": 37413
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.108519773544343e-06,
+      "loss": 0.5749,
+      "step": 37414
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1081935902213613e-06,
+      "loss": 0.6036,
+      "step": 37415
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1078674520802535e-06,
+      "loss": 0.5675,
+      "step": 37416
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1075413591226758e-06,
+      "loss": 0.6119,
+      "step": 37417
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1072153113502849e-06,
+      "loss": 0.6041,
+      "step": 37418
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.106889308764737e-06,
+      "loss": 0.5553,
+      "step": 37419
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1065633513676865e-06,
+      "loss": 0.5957,
+      "step": 37420
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1062374391607955e-06,
+      "loss": 0.6738,
+      "step": 37421
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1059115721457137e-06,
+      "loss": 0.6567,
+      "step": 37422
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.105585750324102e-06,
+      "loss": 0.6121,
+      "step": 37423
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1052599736976144e-06,
+      "loss": 0.548,
+      "step": 37424
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1049342422679054e-06,
+      "loss": 0.6048,
+      "step": 37425
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1046085560366281e-06,
+      "loss": 0.5589,
+      "step": 37426
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1042829150054436e-06,
+      "loss": 0.5902,
+      "step": 37427
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1039573191759988e-06,
+      "loss": 0.6103,
+      "step": 37428
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1036317685499575e-06,
+      "loss": 0.546,
+      "step": 37429
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1033062631289648e-06,
+      "loss": 0.599,
+      "step": 37430
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1029808029146794e-06,
+      "loss": 0.5841,
+      "step": 37431
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1026553879087532e-06,
+      "loss": 0.5571,
+      "step": 37432
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1023300181128416e-06,
+      "loss": 0.6042,
+      "step": 37433
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1020046935285977e-06,
+      "loss": 0.594,
+      "step": 37434
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1016794141576737e-06,
+      "loss": 0.5274,
+      "step": 37435
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1013541800017203e-06,
+      "loss": 0.5661,
+      "step": 37436
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.101028991062395e-06,
+      "loss": 0.6266,
+      "step": 37437
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.100703847341348e-06,
+      "loss": 0.5899,
+      "step": 37438
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1003787488402273e-06,
+      "loss": 0.5968,
+      "step": 37439
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.1000536955606944e-06,
+      "loss": 0.6132,
+      "step": 37440
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.09972868750439e-06,
+      "loss": 0.6208,
+      "step": 37441
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0994037246729727e-06,
+      "loss": 0.5869,
+      "step": 37442
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0990788070680903e-06,
+      "loss": 0.618,
+      "step": 37443
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0987539346913967e-06,
+      "loss": 0.6155,
+      "step": 37444
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.098429107544542e-06,
+      "loss": 0.5899,
+      "step": 37445
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0981043256291746e-06,
+      "loss": 0.6088,
+      "step": 37446
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0977795889469455e-06,
+      "loss": 0.574,
+      "step": 37447
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.097454897499507e-06,
+      "loss": 0.596,
+      "step": 37448
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0971302512885051e-06,
+      "loss": 0.5632,
+      "step": 37449
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0968056503155943e-06,
+      "loss": 0.5965,
+      "step": 37450
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0964810945824212e-06,
+      "loss": 0.575,
+      "step": 37451
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.096156584090635e-06,
+      "loss": 0.6237,
+      "step": 37452
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.095832118841883e-06,
+      "loss": 0.5413,
+      "step": 37453
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0955076988378178e-06,
+      "loss": 0.578,
+      "step": 37454
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0951833240800846e-06,
+      "loss": 0.5922,
+      "step": 37455
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0948589945703335e-06,
+      "loss": 0.6082,
+      "step": 37456
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.094534710310211e-06,
+      "loss": 0.5697,
+      "step": 37457
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0942104713013667e-06,
+      "loss": 0.5943,
+      "step": 37458
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0938862775454474e-06,
+      "loss": 0.5761,
+      "step": 37459
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.093562129044098e-06,
+      "loss": 0.6165,
+      "step": 37460
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0932380257989716e-06,
+      "loss": 0.6069,
+      "step": 37461
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0929139678117073e-06,
+      "loss": 0.5992,
+      "step": 37462
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0925899550839581e-06,
+      "loss": 0.5672,
+      "step": 37463
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0922659876173658e-06,
+      "loss": 0.5701,
+      "step": 37464
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0919420654135815e-06,
+      "loss": 0.6176,
+      "step": 37465
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.091618188474246e-06,
+      "loss": 0.6128,
+      "step": 37466
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0912943568010104e-06,
+      "loss": 0.5755,
+      "step": 37467
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0909705703955142e-06,
+      "loss": 0.6502,
+      "step": 37468
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0906468292594074e-06,
+      "loss": 0.6142,
+      "step": 37469
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.090323133394331e-06,
+      "loss": 0.5319,
+      "step": 37470
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0899994828019346e-06,
+      "loss": 0.5925,
+      "step": 37471
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0896758774838601e-06,
+      "loss": 0.6069,
+      "step": 37472
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0893523174417519e-06,
+      "loss": 0.5661,
+      "step": 37473
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0890288026772521e-06,
+      "loss": 0.6286,
+      "step": 37474
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0887053331920095e-06,
+      "loss": 0.5943,
+      "step": 37475
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0883819089876635e-06,
+      "loss": 0.6305,
+      "step": 37476
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0880585300658576e-06,
+      "loss": 0.6255,
+      "step": 37477
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0877351964282378e-06,
+      "loss": 0.6074,
+      "step": 37478
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0874119080764468e-06,
+      "loss": 0.6011,
+      "step": 37479
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0870886650121249e-06,
+      "loss": 0.5734,
+      "step": 37480
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0867654672369144e-06,
+      "loss": 0.5906,
+      "step": 37481
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0864423147524605e-06,
+      "loss": 0.5968,
+      "step": 37482
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0861192075604044e-06,
+      "loss": 0.6214,
+      "step": 37483
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0857961456623856e-06,
+      "loss": 0.5522,
+      "step": 37484
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0854731290600462e-06,
+      "loss": 0.5148,
+      "step": 37485
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0851501577550305e-06,
+      "loss": 0.5676,
+      "step": 37486
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0848272317489761e-06,
+      "loss": 0.5745,
+      "step": 37487
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0845043510435294e-06,
+      "loss": 0.5812,
+      "step": 37488
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0841815156403223e-06,
+      "loss": 0.6188,
+      "step": 37489
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0838587255410027e-06,
+      "loss": 0.6061,
+      "step": 37490
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0835359807472056e-06,
+      "loss": 0.6764,
+      "step": 37491
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0832132812605766e-06,
+      "loss": 0.643,
+      "step": 37492
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0828906270827511e-06,
+      "loss": 0.5736,
+      "step": 37493
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0825680182153709e-06,
+      "loss": 0.5966,
+      "step": 37494
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0822454546600713e-06,
+      "loss": 0.6174,
+      "step": 37495
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.081922936418497e-06,
+      "loss": 0.5937,
+      "step": 37496
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0816004634922827e-06,
+      "loss": 0.5635,
+      "step": 37497
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0812780358830676e-06,
+      "loss": 0.628,
+      "step": 37498
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0809556535924947e-06,
+      "loss": 0.5743,
+      "step": 37499
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0806333166221938e-06,
+      "loss": 0.5975,
+      "step": 37500
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.080311024973809e-06,
+      "loss": 0.5988,
+      "step": 37501
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0799887786489738e-06,
+      "loss": 0.6465,
+      "step": 37502
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.079666577649331e-06,
+      "loss": 0.5439,
+      "step": 37503
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0793444219765126e-06,
+      "loss": 0.6453,
+      "step": 37504
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0790223116321607e-06,
+      "loss": 0.6255,
+      "step": 37505
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0787002466179065e-06,
+      "loss": 0.597,
+      "step": 37506
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0783782269353905e-06,
+      "loss": 0.5512,
+      "step": 37507
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0780562525862449e-06,
+      "loss": 0.5771,
+      "step": 37508
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0777343235721106e-06,
+      "loss": 0.581,
+      "step": 37509
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0774124398946218e-06,
+      "loss": 0.5535,
+      "step": 37510
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0770906015554128e-06,
+      "loss": 0.6037,
+      "step": 37511
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.076768808556119e-06,
+      "loss": 0.58,
+      "step": 37512
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0764470608983767e-06,
+      "loss": 0.5825,
+      "step": 37513
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0761253585838205e-06,
+      "loss": 0.6248,
+      "step": 37514
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0758037016140831e-06,
+      "loss": 0.5506,
+      "step": 37515
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0754820899908024e-06,
+      "loss": 0.5973,
+      "step": 37516
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0751605237156115e-06,
+      "loss": 0.586,
+      "step": 37517
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0748390027901433e-06,
+      "loss": 0.5766,
+      "step": 37518
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0745175272160302e-06,
+      "loss": 0.5876,
+      "step": 37519
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0741960969949095e-06,
+      "loss": 0.5448,
+      "step": 37520
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.073874712128412e-06,
+      "loss": 0.6353,
+      "step": 37521
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0735533726181723e-06,
+      "loss": 0.5766,
+      "step": 37522
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.073232078465819e-06,
+      "loss": 0.6064,
+      "step": 37523
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0729108296729907e-06,
+      "loss": 0.6072,
+      "step": 37524
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0725896262413137e-06,
+      "loss": 0.5783,
+      "step": 37525
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0722684681724282e-06,
+      "loss": 0.5673,
+      "step": 37526
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0719473554679571e-06,
+      "loss": 0.5873,
+      "step": 37527
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0716262881295381e-06,
+      "loss": 0.6333,
+      "step": 37528
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0713052661587986e-06,
+      "loss": 0.5776,
+      "step": 37529
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.070984289557374e-06,
+      "loss": 0.6235,
+      "step": 37530
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0706633583268933e-06,
+      "loss": 0.6806,
+      "step": 37531
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.070342472468987e-06,
+      "loss": 0.6213,
+      "step": 37532
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0700216319852841e-06,
+      "loss": 0.6088,
+      "step": 37533
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0697008368774187e-06,
+      "loss": 0.5778,
+      "step": 37534
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0693800871470184e-06,
+      "loss": 0.5844,
+      "step": 37535
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.069059382795712e-06,
+      "loss": 0.5534,
+      "step": 37536
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0687387238251312e-06,
+      "loss": 0.5835,
+      "step": 37537
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.068418110236905e-06,
+      "loss": 0.6089,
+      "step": 37538
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0680975420326622e-06,
+      "loss": 0.6196,
+      "step": 37539
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0677770192140301e-06,
+      "loss": 0.6128,
+      "step": 37540
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.06745654178264e-06,
+      "loss": 0.6254,
+      "step": 37541
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0671361097401178e-06,
+      "loss": 0.5339,
+      "step": 37542
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.0668157230880972e-06,
+      "loss": 0.5891,
+      "step": 37543
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0664953818281976e-06,
+      "loss": 0.5866,
+      "step": 37544
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0661750859620545e-06,
+      "loss": 0.6293,
+      "step": 37545
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.065854835491289e-06,
+      "loss": 0.6233,
+      "step": 37546
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.065534630417534e-06,
+      "loss": 0.5799,
+      "step": 37547
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.065214470742414e-06,
+      "loss": 0.6458,
+      "step": 37548
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0648943564675552e-06,
+      "loss": 0.5973,
+      "step": 37549
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.064574287594584e-06,
+      "loss": 0.5548,
+      "step": 37550
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0642542641251297e-06,
+      "loss": 0.5996,
+      "step": 37551
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0639342860608159e-06,
+      "loss": 0.5925,
+      "step": 37552
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0636143534032694e-06,
+      "loss": 0.5977,
+      "step": 37553
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0632944661541133e-06,
+      "loss": 0.6042,
+      "step": 37554
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0629746243149773e-06,
+      "loss": 0.5816,
+      "step": 37555
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0626548278874848e-06,
+      "loss": 0.6144,
+      "step": 37556
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0623350768732577e-06,
+      "loss": 0.5571,
+      "step": 37557
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0620153712739268e-06,
+      "loss": 0.5479,
+      "step": 37558
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.061695711091113e-06,
+      "loss": 0.5934,
+      "step": 37559
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0613760963264408e-06,
+      "loss": 0.5942,
+      "step": 37560
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.061056526981532e-06,
+      "loss": 0.6247,
+      "step": 37561
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0607370030580156e-06,
+      "loss": 0.5929,
+      "step": 37562
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.06041752455751e-06,
+      "loss": 0.5902,
+      "step": 37563
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.060098091481645e-06,
+      "loss": 0.6046,
+      "step": 37564
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0597787038320362e-06,
+      "loss": 0.6184,
+      "step": 37565
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.059459361610312e-06,
+      "loss": 0.594,
+      "step": 37566
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0591400648180915e-06,
+      "loss": 0.5919,
+      "step": 37567
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.058820813457e-06,
+      "loss": 0.5417,
+      "step": 37568
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0585016075286569e-06,
+      "loss": 0.5783,
+      "step": 37569
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.058182447034689e-06,
+      "loss": 0.5917,
+      "step": 37570
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0578633319767117e-06,
+      "loss": 0.6375,
+      "step": 37571
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0575442623563515e-06,
+      "loss": 0.5359,
+      "step": 37572
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.057225238175228e-06,
+      "loss": 0.5792,
+      "step": 37573
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.05690625943496e-06,
+      "loss": 0.6029,
+      "step": 37574
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0565873261371716e-06,
+      "loss": 0.624,
+      "step": 37575
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.056268438283483e-06,
+      "loss": 0.6132,
+      "step": 37576
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.055949595875514e-06,
+      "loss": 0.588,
+      "step": 37577
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0556307989148818e-06,
+      "loss": 0.6148,
+      "step": 37578
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0553120474032119e-06,
+      "loss": 0.5728,
+      "step": 37579
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.05499334134212e-06,
+      "loss": 0.5973,
+      "step": 37580
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0546746807332276e-06,
+      "loss": 0.5428,
+      "step": 37581
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0543560655781504e-06,
+      "loss": 0.5773,
+      "step": 37582
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0540374958785116e-06,
+      "loss": 0.6071,
+      "step": 37583
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0537189716359264e-06,
+      "loss": 0.5675,
+      "step": 37584
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.053400492852017e-06,
+      "loss": 0.5184,
+      "step": 37585
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0530820595283996e-06,
+      "loss": 0.5764,
+      "step": 37586
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0527636716666934e-06,
+      "loss": 0.6161,
+      "step": 37587
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0524453292685133e-06,
+      "loss": 0.6118,
+      "step": 37588
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0521270323354804e-06,
+      "loss": 0.6226,
+      "step": 37589
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.05180878086921e-06,
+      "loss": 0.5814,
+      "step": 37590
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.05149057487132e-06,
+      "loss": 0.5145,
+      "step": 37591
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0511724143434254e-06,
+      "loss": 0.6212,
+      "step": 37592
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0508542992871472e-06,
+      "loss": 0.5738,
+      "step": 37593
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0505362297040989e-06,
+      "loss": 0.6175,
+      "step": 37594
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0502182055958942e-06,
+      "loss": 0.6195,
+      "step": 37595
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0499002269641546e-06,
+      "loss": 0.6196,
+      "step": 37596
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0495822938104927e-06,
+      "loss": 0.6079,
+      "step": 37597
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0492644061365242e-06,
+      "loss": 0.5604,
+      "step": 37598
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0489465639438635e-06,
+      "loss": 0.6112,
+      "step": 37599
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0486287672341278e-06,
+      "loss": 0.5522,
+      "step": 37600
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0483110160089293e-06,
+      "loss": 0.5604,
+      "step": 37601
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.047993310269888e-06,
+      "loss": 0.6064,
+      "step": 37602
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0476756500186103e-06,
+      "loss": 0.5568,
+      "step": 37603
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0473580352567169e-06,
+      "loss": 0.5898,
+      "step": 37604
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0470404659858168e-06,
+      "loss": 0.567,
+      "step": 37605
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0467229422075287e-06,
+      "loss": 0.643,
+      "step": 37606
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.046405463923461e-06,
+      "loss": 0.5565,
+      "step": 37607
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0460880311352328e-06,
+      "loss": 0.602,
+      "step": 37608
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0457706438444503e-06,
+      "loss": 0.6574,
+      "step": 37609
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0454533020527314e-06,
+      "loss": 0.5274,
+      "step": 37610
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0451360057616866e-06,
+      "loss": 0.6535,
+      "step": 37611
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.044818754972926e-06,
+      "loss": 0.5287,
+      "step": 37612
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0445015496880662e-06,
+      "loss": 0.6257,
+      "step": 37613
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0441843899087178e-06,
+      "loss": 0.6136,
+      "step": 37614
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0438672756364898e-06,
+      "loss": 0.6114,
+      "step": 37615
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0435502068729941e-06,
+      "loss": 0.5894,
+      "step": 37616
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.043233183619844e-06,
+      "loss": 0.589,
+      "step": 37617
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0429162058786502e-06,
+      "loss": 0.5935,
+      "step": 37618
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0425992736510215e-06,
+      "loss": 0.5599,
+      "step": 37619
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0422823869385678e-06,
+      "loss": 0.5919,
+      "step": 37620
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0419655457429023e-06,
+      "loss": 0.5736,
+      "step": 37621
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0416487500656314e-06,
+      "loss": 0.6962,
+      "step": 37622
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.041331999908368e-06,
+      "loss": 0.5874,
+      "step": 37623
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0410152952727215e-06,
+      "loss": 0.5842,
+      "step": 37624
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0406986361603e-06,
+      "loss": 0.6361,
+      "step": 37625
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0403820225727101e-06,
+      "loss": 0.6469,
+      "step": 37626
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0400654545115652e-06,
+      "loss": 0.6089,
+      "step": 37627
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0397489319784703e-06,
+      "loss": 0.6576,
+      "step": 37628
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.039432454975039e-06,
+      "loss": 0.6038,
+      "step": 37629
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0391160235028707e-06,
+      "loss": 0.5799,
+      "step": 37630
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0387996375635812e-06,
+      "loss": 0.6304,
+      "step": 37631
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0384832971587744e-06,
+      "loss": 0.6161,
+      "step": 37632
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.038167002290057e-06,
+      "loss": 0.6179,
+      "step": 37633
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0378507529590398e-06,
+      "loss": 0.6439,
+      "step": 37634
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0375345491673272e-06,
+      "loss": 0.5489,
+      "step": 37635
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0372183909165267e-06,
+      "loss": 0.652,
+      "step": 37636
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0369022782082415e-06,
+      "loss": 0.583,
+      "step": 37637
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0365862110440838e-06,
+      "loss": 0.6208,
+      "step": 37638
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0362701894256543e-06,
+      "loss": 0.5839,
+      "step": 37639
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0359542133545652e-06,
+      "loss": 0.5768,
+      "step": 37640
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.035638282832414e-06,
+      "loss": 0.6213,
+      "step": 37641
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0353223978608117e-06,
+      "loss": 0.5243,
+      "step": 37642
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0350065584413604e-06,
+      "loss": 0.5906,
+      "step": 37643
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0346907645756677e-06,
+      "loss": 0.6474,
+      "step": 37644
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0343750162653377e-06,
+      "loss": 0.6323,
+      "step": 37645
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0340593135119726e-06,
+      "loss": 0.6082,
+      "step": 37646
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0337436563171765e-06,
+      "loss": 0.5901,
+      "step": 37647
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0334280446825574e-06,
+      "loss": 0.6251,
+      "step": 37648
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0331124786097134e-06,
+      "loss": 0.5744,
+      "step": 37649
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0327969581002538e-06,
+      "loss": 0.589,
+      "step": 37650
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0324814831557784e-06,
+      "loss": 0.5731,
+      "step": 37651
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.032166053777891e-06,
+      "loss": 0.5433,
+      "step": 37652
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0318506699681941e-06,
+      "loss": 0.6251,
+      "step": 37653
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0315353317282895e-06,
+      "loss": 0.5688,
+      "step": 37654
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0312200390597816e-06,
+      "loss": 0.5729,
+      "step": 37655
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0309047919642713e-06,
+      "loss": 0.5939,
+      "step": 37656
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0305895904433593e-06,
+      "loss": 0.616,
+      "step": 37657
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0302744344986482e-06,
+      "loss": 0.5694,
+      "step": 37658
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0299593241317407e-06,
+      "loss": 0.5523,
+      "step": 37659
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0296442593442346e-06,
+      "loss": 0.5444,
+      "step": 37660
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0293292401377365e-06,
+      "loss": 0.6586,
+      "step": 37661
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0290142665138425e-06,
+      "loss": 0.5972,
+      "step": 37662
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.028699338474155e-06,
+      "loss": 0.6333,
+      "step": 37663
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0283844560202705e-06,
+      "loss": 0.6539,
+      "step": 37664
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0280696191537953e-06,
+      "loss": 0.6226,
+      "step": 37665
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0277548278763228e-06,
+      "loss": 0.5731,
+      "step": 37666
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0274400821894603e-06,
+      "loss": 0.6365,
+      "step": 37667
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0271253820947979e-06,
+      "loss": 0.6013,
+      "step": 37668
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0268107275939422e-06,
+      "loss": 0.6026,
+      "step": 37669
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.026496118688487e-06,
+      "loss": 0.5942,
+      "step": 37670
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0261815553800335e-06,
+      "loss": 0.57,
+      "step": 37671
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0258670376701817e-06,
+      "loss": 0.6223,
+      "step": 37672
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.025552565560526e-06,
+      "loss": 0.664,
+      "step": 37673
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0252381390526667e-06,
+      "loss": 0.6329,
+      "step": 37674
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0249237581481998e-06,
+      "loss": 0.5963,
+      "step": 37675
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0246094228487247e-06,
+      "loss": 0.594,
+      "step": 37676
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.024295133155835e-06,
+      "loss": 0.5799,
+      "step": 37677
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0239808890711356e-06,
+      "loss": 0.5941,
+      "step": 37678
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0236666905962133e-06,
+      "loss": 0.5577,
+      "step": 37679
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0233525377326713e-06,
+      "loss": 0.6179,
+      "step": 37680
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0230384304821028e-06,
+      "loss": 0.601,
+      "step": 37681
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0227243688461064e-06,
+      "loss": 0.5977,
+      "step": 37682
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0224103528262775e-06,
+      "loss": 0.6072,
+      "step": 37683
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0220963824242103e-06,
+      "loss": 0.5875,
+      "step": 37684
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0217824576414982e-06,
+      "loss": 0.6053,
+      "step": 37685
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0214685784797406e-06,
+      "loss": 0.6749,
+      "step": 37686
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0211547449405301e-06,
+      "loss": 0.6112,
+      "step": 37687
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.020840957025463e-06,
+      "loss": 0.6131,
+      "step": 37688
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0205272147361322e-06,
+      "loss": 0.6229,
+      "step": 37689
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.0202135180741335e-06,
+      "loss": 0.5942,
+      "step": 37690
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0198998670410587e-06,
+      "loss": 0.6136,
+      "step": 37691
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.019586261638501e-06,
+      "loss": 0.6478,
+      "step": 37692
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.019272701868058e-06,
+      "loss": 0.6232,
+      "step": 37693
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0189591877313198e-06,
+      "loss": 0.6361,
+      "step": 37694
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0186457192298805e-06,
+      "loss": 0.6351,
+      "step": 37695
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0183322963653307e-06,
+      "loss": 0.5917,
+      "step": 37696
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0180189191392664e-06,
+      "loss": 0.6691,
+      "step": 37697
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0177055875532771e-06,
+      "loss": 0.6464,
+      "step": 37698
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0173923016089582e-06,
+      "loss": 0.6139,
+      "step": 37699
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0170790613078996e-06,
+      "loss": 0.5934,
+      "step": 37700
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0167658666516934e-06,
+      "loss": 0.6046,
+      "step": 37701
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0164527176419282e-06,
+      "loss": 0.6639,
+      "step": 37702
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0161396142801994e-06,
+      "loss": 0.5961,
+      "step": 37703
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0158265565680947e-06,
+      "loss": 0.6039,
+      "step": 37704
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0155135445072106e-06,
+      "loss": 0.6005,
+      "step": 37705
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0152005780991291e-06,
+      "loss": 0.5697,
+      "step": 37706
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.014887657345447e-06,
+      "loss": 0.5799,
+      "step": 37707
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0145747822477504e-06,
+      "loss": 0.584,
+      "step": 37708
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0142619528076326e-06,
+      "loss": 0.6232,
+      "step": 37709
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0139491690266812e-06,
+      "loss": 0.5559,
+      "step": 37710
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.013636430906485e-06,
+      "loss": 0.6133,
+      "step": 37711
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.013323738448635e-06,
+      "loss": 0.5005,
+      "step": 37712
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0130110916547164e-06,
+      "loss": 0.5488,
+      "step": 37713
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0126984905263226e-06,
+      "loss": 0.584,
+      "step": 37714
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.012385935065039e-06,
+      "loss": 0.5929,
+      "step": 37715
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0120734252724573e-06,
+      "loss": 0.6019,
+      "step": 37716
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0117609611501588e-06,
+      "loss": 0.5994,
+      "step": 37717
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0114485426997367e-06,
+      "loss": 0.5812,
+      "step": 37718
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0111361699227763e-06,
+      "loss": 0.6305,
+      "step": 37719
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0108238428208662e-06,
+      "loss": 0.6087,
+      "step": 37720
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.010511561395594e-06,
+      "loss": 0.5416,
+      "step": 37721
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0101993256485442e-06,
+      "loss": 0.6311,
+      "step": 37722
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0098871355813033e-06,
+      "loss": 0.5898,
+      "step": 37723
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.00957499119546e-06,
+      "loss": 0.5848,
+      "step": 37724
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0092628924925974e-06,
+      "loss": 0.5168,
+      "step": 37725
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0089508394743053e-06,
+      "loss": 0.5876,
+      "step": 37726
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0086388321421669e-06,
+      "loss": 0.6191,
+      "step": 37727
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0083268704977678e-06,
+      "loss": 0.6011,
+      "step": 37728
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.008014954542692e-06,
+      "loss": 0.5255,
+      "step": 37729
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0077030842785263e-06,
+      "loss": 0.6042,
+      "step": 37730
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0073912597068558e-06,
+      "loss": 0.5824,
+      "step": 37731
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0070794808292638e-06,
+      "loss": 0.5901,
+      "step": 37732
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0067677476473337e-06,
+      "loss": 0.5803,
+      "step": 37733
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0064560601626495e-06,
+      "loss": 0.6,
+      "step": 37734
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0061444183767977e-06,
+      "loss": 0.5528,
+      "step": 37735
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0058328222913582e-06,
+      "loss": 0.6427,
+      "step": 37736
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.00552127190792e-06,
+      "loss": 0.5835,
+      "step": 37737
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.005209767228058e-06,
+      "loss": 0.5527,
+      "step": 37738
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0048983082533614e-06,
+      "loss": 0.6295,
+      "step": 37739
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.00458689498541e-06,
+      "loss": 0.5978,
+      "step": 37740
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0042755274257877e-06,
+      "loss": 0.6318,
+      "step": 37741
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0039642055760746e-06,
+      "loss": 0.5632,
+      "step": 37742
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0036529294378572e-06,
+      "loss": 0.5477,
+      "step": 37743
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.003341699012711e-06,
+      "loss": 0.5917,
+      "step": 37744
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0030305143022223e-06,
+      "loss": 0.5643,
+      "step": 37745
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0027193753079679e-06,
+      "loss": 0.5901,
+      "step": 37746
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0024082820315339e-06,
+      "loss": 0.6119,
+      "step": 37747
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0020972344744984e-06,
+      "loss": 0.6249,
+      "step": 37748
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0017862326384419e-06,
+      "loss": 0.6854,
+      "step": 37749
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0014752765249425e-06,
+      "loss": 0.6093,
+      "step": 37750
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.001164366135584e-06,
+      "loss": 0.5768,
+      "step": 37751
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0008535014719467e-06,
+      "loss": 0.5588,
+      "step": 37752
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0005426825356046e-06,
+      "loss": 0.5802,
+      "step": 37753
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.0002319093281454e-06,
+      "loss": 0.5884,
+      "step": 37754
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.99921181851139e-07,
+      "loss": 0.548,
+      "step": 37755
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.996105001061719e-07,
+      "loss": 0.5759,
+      "step": 37756
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.992998640948171e-07,
+      "loss": 0.6197,
+      "step": 37757
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.98989273818658e-07,
+      "loss": 0.5838,
+      "step": 37758
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.986787292792698e-07,
+      "loss": 0.6011,
+      "step": 37759
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.983682304782316e-07,
+      "loss": 0.5869,
+      "step": 37760
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.980577774171185e-07,
+      "loss": 0.5732,
+      "step": 37761
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.977473700975126e-07,
+      "loss": 0.5832,
+      "step": 37762
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.974370085209862e-07,
+      "loss": 0.5936,
+      "step": 37763
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.971266926891231e-07,
+      "loss": 0.6134,
+      "step": 37764
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.968164226034915e-07,
+      "loss": 0.6376,
+      "step": 37765
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.965061982656754e-07,
+      "loss": 0.6157,
+      "step": 37766
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.961960196772458e-07,
+      "loss": 0.5386,
+      "step": 37767
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.958858868397825e-07,
+      "loss": 0.5578,
+      "step": 37768
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.9557579975486e-07,
+      "loss": 0.5929,
+      "step": 37769
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.952657584240545e-07,
+      "loss": 0.6298,
+      "step": 37770
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.949557628489403e-07,
+      "loss": 0.5536,
+      "step": 37771
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.94645813031091e-07,
+      "loss": 0.5845,
+      "step": 37772
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.943359089720861e-07,
+      "loss": 0.551,
+      "step": 37773
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.940260506734956e-07,
+      "loss": 0.6198,
+      "step": 37774
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.937162381369004e-07,
+      "loss": 0.5838,
+      "step": 37775
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.934064713638669e-07,
+      "loss": 0.6152,
+      "step": 37776
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.930967503559753e-07,
+      "loss": 0.6408,
+      "step": 37777
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.927870751147938e-07,
+      "loss": 0.5961,
+      "step": 37778
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.924774456419017e-07,
+      "loss": 0.5484,
+      "step": 37779
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.921678619388674e-07,
+      "loss": 0.6088,
+      "step": 37780
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.918583240072699e-07,
+      "loss": 0.5869,
+      "step": 37781
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.915488318486744e-07,
+      "loss": 0.6091,
+      "step": 37782
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.912393854646606e-07,
+      "loss": 0.5542,
+      "step": 37783
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.909299848567944e-07,
+      "loss": 0.6559,
+      "step": 37784
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.906206300266542e-07,
+      "loss": 0.7399,
+      "step": 37785
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.903113209758098e-07,
+      "loss": 0.6451,
+      "step": 37786
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.90002057705831e-07,
+      "loss": 0.5972,
+      "step": 37787
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.896928402182892e-07,
+      "loss": 0.5663,
+      "step": 37788
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.893836685147585e-07,
+      "loss": 0.5941,
+      "step": 37789
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.890745425968073e-07,
+      "loss": 0.6583,
+      "step": 37790
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.88765462466007e-07,
+      "loss": 0.6505,
+      "step": 37791
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.884564281239318e-07,
+      "loss": 0.6179,
+      "step": 37792
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.881474395721446e-07,
+      "loss": 0.568,
+      "step": 37793
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.878384968122211e-07,
+      "loss": 0.5579,
+      "step": 37794
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.875295998457279e-07,
+      "loss": 0.6189,
+      "step": 37795
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.87220748674238e-07,
+      "loss": 0.6293,
+      "step": 37796
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.8691194329932e-07,
+      "loss": 0.6701,
+      "step": 37797
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.86603183722541e-07,
+      "loss": 0.5498,
+      "step": 37798
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.862944699454691e-07,
+      "loss": 0.6296,
+      "step": 37799
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.859858019696766e-07,
+      "loss": 0.6251,
+      "step": 37800
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.85677179796729e-07,
+      "loss": 0.5971,
+      "step": 37801
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.853686034281985e-07,
+      "loss": 0.573,
+      "step": 37802
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.850600728656468e-07,
+      "loss": 0.5874,
+      "step": 37803
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.847515881106473e-07,
+      "loss": 0.5964,
+      "step": 37804
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.84443149164762e-07,
+      "loss": 0.6248,
+      "step": 37805
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.841347560295645e-07,
+      "loss": 0.5817,
+      "step": 37806
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.838264087066185e-07,
+      "loss": 0.5724,
+      "step": 37807
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.83518107197491e-07,
+      "loss": 0.6023,
+      "step": 37808
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.83209851503747e-07,
+      "loss": 0.538,
+      "step": 37809
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.829016416269554e-07,
+      "loss": 0.5792,
+      "step": 37810
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.82593477568682e-07,
+      "loss": 0.6267,
+      "step": 37811
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.822853593304905e-07,
+      "loss": 0.5732,
+      "step": 37812
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.81977286913951e-07,
+      "loss": 0.5959,
+      "step": 37813
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.816692603206224e-07,
+      "loss": 0.5404,
+      "step": 37814
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.813612795520754e-07,
+      "loss": 0.6449,
+      "step": 37815
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.810533446098713e-07,
+      "loss": 0.5783,
+      "step": 37816
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.807454554955776e-07,
+      "loss": 0.6638,
+      "step": 37817
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.804376122107573e-07,
+      "loss": 0.5636,
+      "step": 37818
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.801298147569771e-07,
+      "loss": 0.6089,
+      "step": 37819
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.798220631357957e-07,
+      "loss": 0.6323,
+      "step": 37820
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.79514357348782e-07,
+      "loss": 0.5927,
+      "step": 37821
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.792066973974968e-07,
+      "loss": 0.62,
+      "step": 37822
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.788990832835043e-07,
+      "loss": 0.6313,
+      "step": 37823
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.785915150083692e-07,
+      "loss": 0.577,
+      "step": 37824
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.78283992573651e-07,
+      "loss": 0.5832,
+      "step": 37825
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.77976515980913e-07,
+      "loss": 0.5603,
+      "step": 37826
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.776690852317206e-07,
+      "loss": 0.6525,
+      "step": 37827
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.773617003276338e-07,
+      "loss": 0.5912,
+      "step": 37828
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.770543612702144e-07,
+      "loss": 0.6068,
+      "step": 37829
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.767470680610225e-07,
+      "loss": 0.5515,
+      "step": 37830
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.764398207016223e-07,
+      "loss": 0.5765,
+      "step": 37831
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.761326191935749e-07,
+      "loss": 0.616,
+      "step": 37832
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.75825463538439e-07,
+      "loss": 0.6267,
+      "step": 37833
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.755183537377788e-07,
+      "loss": 0.5386,
+      "step": 37834
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.75211289793152e-07,
+      "loss": 0.5587,
+      "step": 37835
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.749042717061196e-07,
+      "loss": 0.6088,
+      "step": 37836
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.745972994782405e-07,
+      "loss": 0.5877,
+      "step": 37837
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.742903731110764e-07,
+      "loss": 0.6135,
+      "step": 37838
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.739834926061852e-07,
+      "loss": 0.613,
+      "step": 37839
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.736766579651314e-07,
+      "loss": 0.5612,
+      "step": 37840
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.733698691894655e-07,
+      "loss": 0.5904,
+      "step": 37841
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.730631262807532e-07,
+      "loss": 0.5781,
+      "step": 37842
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.72756429240549e-07,
+      "loss": 0.6057,
+      "step": 37843
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.724497780704144e-07,
+      "loss": 0.6565,
+      "step": 37844
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.721431727719067e-07,
+      "loss": 0.5929,
+      "step": 37845
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.71836613346583e-07,
+      "loss": 0.6542,
+      "step": 37846
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.715300997960009e-07,
+      "loss": 0.5516,
+      "step": 37847
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.712236321217194e-07,
+      "loss": 0.6652,
+      "step": 37848
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.70917210325295e-07,
+      "loss": 0.616,
+      "step": 37849
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.706108344082822e-07,
+      "loss": 0.6886,
+      "step": 37850
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.703045043722425e-07,
+      "loss": 0.6385,
+      "step": 37851
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.699982202187297e-07,
+      "loss": 0.588,
+      "step": 37852
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.696919819493012e-07,
+      "loss": 0.5866,
+      "step": 37853
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.693857895655101e-07,
+      "loss": 0.576,
+      "step": 37854
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.690796430689164e-07,
+      "loss": 0.5828,
+      "step": 37855
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.687735424610734e-07,
+      "loss": 0.6074,
+      "step": 37856
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.684674877435374e-07,
+      "loss": 0.6058,
+      "step": 37857
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.681614789178618e-07,
+      "loss": 0.5986,
+      "step": 37858
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.67855515985604e-07,
+      "loss": 0.563,
+      "step": 37859
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.675495989483152e-07,
+      "loss": 0.6412,
+      "step": 37860
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.672437278075552e-07,
+      "loss": 0.5747,
+      "step": 37861
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.669379025648751e-07,
+      "loss": 0.5587,
+      "step": 37862
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.66632123221828e-07,
+      "loss": 0.5752,
+      "step": 37863
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.663263897799669e-07,
+      "loss": 0.5992,
+      "step": 37864
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.660207022408497e-07,
+      "loss": 0.6555,
+      "step": 37865
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.657150606060262e-07,
+      "loss": 0.6452,
+      "step": 37866
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.654094648770507e-07,
+      "loss": 0.5395,
+      "step": 37867
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.65103915055474e-07,
+      "loss": 0.5482,
+      "step": 37868
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.64798411142852e-07,
+      "loss": 0.6135,
+      "step": 37869
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.644929531407354e-07,
+      "loss": 0.5691,
+      "step": 37870
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.64187541050674e-07,
+      "loss": 0.65,
+      "step": 37871
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.638821748742243e-07,
+      "loss": 0.6631,
+      "step": 37872
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.635768546129354e-07,
+      "loss": 0.6063,
+      "step": 37873
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.63271580268359e-07,
+      "loss": 0.6012,
+      "step": 37874
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.62966351842044e-07,
+      "loss": 0.6059,
+      "step": 37875
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.62661169335546e-07,
+      "loss": 0.5504,
+      "step": 37876
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.623560327504111e-07,
+      "loss": 0.6172,
+      "step": 37877
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.620509420881962e-07,
+      "loss": 0.5285,
+      "step": 37878
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.617458973504423e-07,
+      "loss": 0.6115,
+      "step": 37879
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.61440898538708e-07,
+      "loss": 0.6306,
+      "step": 37880
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.611359456545365e-07,
+      "loss": 0.5805,
+      "step": 37881
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.608310386994835e-07,
+      "loss": 0.5999,
+      "step": 37882
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.60526177675094e-07,
+      "loss": 0.566,
+      "step": 37883
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.602213625829193e-07,
+      "loss": 0.5871,
+      "step": 37884
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.599165934245048e-07,
+      "loss": 0.6082,
+      "step": 37885
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.596118702014034e-07,
+      "loss": 0.6065,
+      "step": 37886
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.59307192915162e-07,
+      "loss": 0.6374,
+      "step": 37887
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.590025615673259e-07,
+      "loss": 0.5708,
+      "step": 37888
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.586979761594483e-07,
+      "loss": 0.6433,
+      "step": 37889
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.583934366930736e-07,
+      "loss": 0.5483,
+      "step": 37890
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.580889431697505e-07,
+      "loss": 0.6307,
+      "step": 37891
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.577844955910231e-07,
+      "loss": 0.6209,
+      "step": 37892
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.574800939584428e-07,
+      "loss": 0.5794,
+      "step": 37893
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.571757382735548e-07,
+      "loss": 0.6286,
+      "step": 37894
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.568714285379045e-07,
+      "loss": 0.5412,
+      "step": 37895
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.565671647530373e-07,
+      "loss": 0.6271,
+      "step": 37896
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.56262946920502e-07,
+      "loss": 0.6032,
+      "step": 37897
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.55958775041842e-07,
+      "loss": 0.576,
+      "step": 37898
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.55654649118606e-07,
+      "loss": 0.6444,
+      "step": 37899
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.55350569152338e-07,
+      "loss": 0.5694,
+      "step": 37900
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.550465351445814e-07,
+      "loss": 0.6085,
+      "step": 37901
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.54742547096882e-07,
+      "loss": 0.5745,
+      "step": 37902
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.544386050107856e-07,
+      "loss": 0.6345,
+      "step": 37903
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.54134708887836e-07,
+      "loss": 0.5768,
+      "step": 37904
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.538308587295764e-07,
+      "loss": 0.6076,
+      "step": 37905
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.535270545375508e-07,
+      "loss": 0.5572,
+      "step": 37906
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.532232963133048e-07,
+      "loss": 0.605,
+      "step": 37907
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.529195840583815e-07,
+      "loss": 0.5787,
+      "step": 37908
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.5261591777432e-07,
+      "loss": 0.5752,
+      "step": 37909
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.523122974626697e-07,
+      "loss": 0.5971,
+      "step": 37910
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.520087231249697e-07,
+      "loss": 0.613,
+      "step": 37911
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.517051947627642e-07,
+      "loss": 0.6533,
+      "step": 37912
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.51401712377592e-07,
+      "loss": 0.5584,
+      "step": 37913
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.510982759709997e-07,
+      "loss": 0.588,
+      "step": 37914
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.507948855445248e-07,
+      "loss": 0.6023,
+      "step": 37915
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.504915410997162e-07,
+      "loss": 0.6055,
+      "step": 37916
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.501882426381048e-07,
+      "loss": 0.5939,
+      "step": 37917
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.498849901612405e-07,
+      "loss": 0.5954,
+      "step": 37918
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.49581783670659e-07,
+      "loss": 0.6652,
+      "step": 37919
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.492786231679052e-07,
+      "loss": 0.6715,
+      "step": 37920
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.48975508654516e-07,
+      "loss": 0.6159,
+      "step": 37921
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.486724401320346e-07,
+      "loss": 0.579,
+      "step": 37922
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.483694176019964e-07,
+      "loss": 0.6325,
+      "step": 37923
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.480664410659457e-07,
+      "loss": 0.5555,
+      "step": 37924
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.477635105254212e-07,
+      "loss": 0.5748,
+      "step": 37925
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.474606259819596e-07,
+      "loss": 0.59,
+      "step": 37926
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.47157787437103e-07,
+      "loss": 0.6133,
+      "step": 37927
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.468549948923888e-07,
+      "loss": 0.579,
+      "step": 37928
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.465522483493561e-07,
+      "loss": 0.5563,
+      "step": 37929
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.462495478095412e-07,
+      "loss": 0.6182,
+      "step": 37930
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.459468932744853e-07,
+      "loss": 0.6665,
+      "step": 37931
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.456442847457248e-07,
+      "loss": 0.5783,
+      "step": 37932
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.453417222247974e-07,
+      "loss": 0.5906,
+      "step": 37933
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.450392057132385e-07,
+      "loss": 0.6045,
+      "step": 37934
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.447367352125902e-07,
+      "loss": 0.6292,
+      "step": 37935
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.444343107243836e-07,
+      "loss": 0.6007,
+      "step": 37936
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.441319322501608e-07,
+      "loss": 0.5491,
+      "step": 37937
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.438295997914549e-07,
+      "loss": 0.6194,
+      "step": 37938
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.435273133498036e-07,
+      "loss": 0.5695,
+      "step": 37939
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.4322507292674e-07,
+      "loss": 0.5689,
+      "step": 37940
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.429228785238054e-07,
+      "loss": 0.6031,
+      "step": 37941
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.426207301425283e-07,
+      "loss": 0.5802,
+      "step": 37942
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.423186277844532e-07,
+      "loss": 0.5648,
+      "step": 37943
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.420165714511054e-07,
+      "loss": 0.6114,
+      "step": 37944
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.417145611440259e-07,
+      "loss": 0.5438,
+      "step": 37945
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.414125968647481e-07,
+      "loss": 0.5875,
+      "step": 37946
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.411106786148039e-07,
+      "loss": 0.6947,
+      "step": 37947
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.408088063957311e-07,
+      "loss": 0.5987,
+      "step": 37948
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.405069802090617e-07,
+      "loss": 0.531,
+      "step": 37949
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.402052000563289e-07,
+      "loss": 0.5541,
+      "step": 37950
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.39903465939066e-07,
+      "loss": 0.6594,
+      "step": 37951
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.396017778588085e-07,
+      "loss": 0.5997,
+      "step": 37952
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.393001358170861e-07,
+      "loss": 0.5645,
+      "step": 37953
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.389985398154366e-07,
+      "loss": 0.5348,
+      "step": 37954
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.386969898553855e-07,
+      "loss": 0.5931,
+      "step": 37955
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.383954859384703e-07,
+      "loss": 0.517,
+      "step": 37956
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.380940280662199e-07,
+      "loss": 0.5994,
+      "step": 37957
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.377926162401696e-07,
+      "loss": 0.5411,
+      "step": 37958
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.374912504618494e-07,
+      "loss": 0.5536,
+      "step": 37959
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.371899307327892e-07,
+      "loss": 0.5937,
+      "step": 37960
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.368886570545199e-07,
+      "loss": 0.5929,
+      "step": 37961
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.365874294285759e-07,
+      "loss": 0.594,
+      "step": 37962
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.362862478564849e-07,
+      "loss": 0.6566,
+      "step": 37963
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.359851123397756e-07,
+      "loss": 0.5268,
+      "step": 37964
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.356840228799835e-07,
+      "loss": 0.6141,
+      "step": 37965
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.353829794786351e-07,
+      "loss": 0.5899,
+      "step": 37966
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.350819821372603e-07,
+      "loss": 0.6137,
+      "step": 37967
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.347810308573868e-07,
+      "loss": 0.5988,
+      "step": 37968
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.344801256405478e-07,
+      "loss": 0.6458,
+      "step": 37969
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.341792664882699e-07,
+      "loss": 0.6279,
+      "step": 37970
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.338784534020817e-07,
+      "loss": 0.558,
+      "step": 37971
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.335776863835111e-07,
+      "loss": 0.6334,
+      "step": 37972
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.332769654340901e-07,
+      "loss": 0.676,
+      "step": 37973
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.329762905553408e-07,
+      "loss": 0.6364,
+      "step": 37974
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.326756617487976e-07,
+      "loss": 0.5989,
+      "step": 37975
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.323750790159836e-07,
+      "loss": 0.6081,
+      "step": 37976
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.320745423584288e-07,
+      "loss": 0.6213,
+      "step": 37977
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.317740517776552e-07,
+      "loss": 0.5916,
+      "step": 37978
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.314736072751962e-07,
+      "loss": 0.6428,
+      "step": 37979
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.311732088525749e-07,
+      "loss": 0.5911,
+      "step": 37980
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.308728565113212e-07,
+      "loss": 0.546,
+      "step": 37981
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.305725502529539e-07,
+      "loss": 0.6168,
+      "step": 37982
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.302722900790062e-07,
+      "loss": 0.5515,
+      "step": 37983
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.299720759910014e-07,
+      "loss": 0.6257,
+      "step": 37984
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.296719079904625e-07,
+      "loss": 0.5901,
+      "step": 37985
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.293717860789197e-07,
+      "loss": 0.6024,
+      "step": 37986
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.29071710257895e-07,
+      "loss": 0.5919,
+      "step": 37987
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.287716805289127e-07,
+      "loss": 0.6611,
+      "step": 37988
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.284716968934971e-07,
+      "loss": 0.5589,
+      "step": 37989
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.281717593531747e-07,
+      "loss": 0.5803,
+      "step": 37990
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.278718679094678e-07,
+      "loss": 0.6304,
+      "step": 37991
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.275720225639029e-07,
+      "loss": 0.5822,
+      "step": 37992
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.272722233179988e-07,
+      "loss": 0.6278,
+      "step": 37993
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.26972470173284e-07,
+      "loss": 0.639,
+      "step": 37994
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.266727631312766e-07,
+      "loss": 0.6002,
+      "step": 37995
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.26373102193504e-07,
+      "loss": 0.6077,
+      "step": 37996
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.260734873614863e-07,
+      "loss": 0.5391,
+      "step": 37997
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.257739186367476e-07,
+      "loss": 0.6327,
+      "step": 37998
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.254743960208068e-07,
+      "loss": 0.5426,
+      "step": 37999
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.251749195151904e-07,
+      "loss": 0.6343,
+      "step": 38000
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.248754891214162e-07,
+      "loss": 0.6177,
+      "step": 38001
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.245761048410095e-07,
+      "loss": 0.6249,
+      "step": 38002
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.242767666754881e-07,
+      "loss": 0.5504,
+      "step": 38003
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.239774746263763e-07,
+      "loss": 0.6066,
+      "step": 38004
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.236782286951918e-07,
+      "loss": 0.6384,
+      "step": 38005
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.233790288834543e-07,
+      "loss": 0.5946,
+      "step": 38006
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.230798751926884e-07,
+      "loss": 0.5361,
+      "step": 38007
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.227807676244127e-07,
+      "loss": 0.59,
+      "step": 38008
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.224817061801449e-07,
+      "loss": 0.5859,
+      "step": 38009
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.221826908614051e-07,
+      "loss": 0.6675,
+      "step": 38010
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.218837216697152e-07,
+      "loss": 0.5696,
+      "step": 38011
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.215847986065907e-07,
+      "loss": 0.6081,
+      "step": 38012
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.21285921673557e-07,
+      "loss": 0.6617,
+      "step": 38013
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.209870908721241e-07,
+      "loss": 0.5517,
+      "step": 38014
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.206883062038163e-07,
+      "loss": 0.6037,
+      "step": 38015
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.203895676701491e-07,
+      "loss": 0.5783,
+      "step": 38016
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.200908752726423e-07,
+      "loss": 0.6227,
+      "step": 38017
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.197922290128125e-07,
+      "loss": 0.6282,
+      "step": 38018
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.194936288921796e-07,
+      "loss": 0.6232,
+      "step": 38019
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.191950749122569e-07,
+      "loss": 0.6096,
+      "step": 38020
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.188965670745642e-07,
+      "loss": 0.574,
+      "step": 38021
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.18598105380617e-07,
+      "loss": 0.586,
+      "step": 38022
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.18299689831933e-07,
+      "loss": 0.5883,
+      "step": 38023
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.180013204300286e-07,
+      "loss": 0.5368,
+      "step": 38024
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.177029971764195e-07,
+      "loss": 0.6099,
+      "step": 38025
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.174047200726211e-07,
+      "loss": 0.5742,
+      "step": 38026
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.171064891201464e-07,
+      "loss": 0.6696,
+      "step": 38027
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.168083043205167e-07,
+      "loss": 0.6456,
+      "step": 38028
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.165101656752418e-07,
+      "loss": 0.5657,
+      "step": 38029
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.162120731858437e-07,
+      "loss": 0.618,
+      "step": 38030
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.159140268538269e-07,
+      "loss": 0.6152,
+      "step": 38031
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.156160266807146e-07,
+      "loss": 0.6272,
+      "step": 38032
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.153180726680156e-07,
+      "loss": 0.5609,
+      "step": 38033
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.150201648172475e-07,
+      "loss": 0.6251,
+      "step": 38034
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.147223031299235e-07,
+      "loss": 0.6351,
+      "step": 38035
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.144244876075548e-07,
+      "loss": 0.6175,
+      "step": 38036
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.141267182516556e-07,
+      "loss": 0.5853,
+      "step": 38037
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.138289950637403e-07,
+      "loss": 0.6688,
+      "step": 38038
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.135313180453198e-07,
+      "loss": 0.627,
+      "step": 38039
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.132336871979108e-07,
+      "loss": 0.6226,
+      "step": 38040
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.129361025230199e-07,
+      "loss": 0.5934,
+      "step": 38041
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.126385640221625e-07,
+      "loss": 0.5737,
+      "step": 38042
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.123410716968495e-07,
+      "loss": 0.604,
+      "step": 38043
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.120436255485931e-07,
+      "loss": 0.6024,
+      "step": 38044
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.117462255789056e-07,
+      "loss": 0.6543,
+      "step": 38045
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.114488717892977e-07,
+      "loss": 0.6731,
+      "step": 38046
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.111515641812785e-07,
+      "loss": 0.6053,
+      "step": 38047
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.108543027563588e-07,
+      "loss": 0.625,
+      "step": 38048
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.105570875160518e-07,
+      "loss": 0.5275,
+      "step": 38049
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.102599184618654e-07,
+      "loss": 0.6373,
+      "step": 38050
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.099627955953128e-07,
+      "loss": 0.6603,
+      "step": 38051
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.096657189178982e-07,
+      "loss": 0.5907,
+      "step": 38052
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.09368688431137e-07,
+      "loss": 0.5684,
+      "step": 38053
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.090717041365327e-07,
+      "loss": 0.6042,
+      "step": 38054
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.087747660355995e-07,
+      "loss": 0.6601,
+      "step": 38055
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.084778741298428e-07,
+      "loss": 0.6256,
+      "step": 38056
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.081810284207759e-07,
+      "loss": 0.5524,
+      "step": 38057
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.07884228909901e-07,
+      "loss": 0.6343,
+      "step": 38058
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.075874755987302e-07,
+      "loss": 0.5997,
+      "step": 38059
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.072907684887689e-07,
+      "loss": 0.6462,
+      "step": 38060
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.069941075815269e-07,
+      "loss": 0.5517,
+      "step": 38061
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.066974928785121e-07,
+      "loss": 0.6135,
+      "step": 38062
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.064009243812299e-07,
+      "loss": 0.5909,
+      "step": 38063
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.06104402091188e-07,
+      "loss": 0.607,
+      "step": 38064
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.058079260098906e-07,
+      "loss": 0.5837,
+      "step": 38065
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.055114961388478e-07,
+      "loss": 0.6687,
+      "step": 38066
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.052151124795638e-07,
+      "loss": 0.5909,
+      "step": 38067
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.049187750335487e-07,
+      "loss": 0.5798,
+      "step": 38068
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.04622483802301e-07,
+      "loss": 0.6186,
+      "step": 38069
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.043262387873331e-07,
+      "loss": 0.5782,
+      "step": 38070
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.040300399901447e-07,
+      "loss": 0.5947,
+      "step": 38071
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.037338874122448e-07,
+      "loss": 0.6461,
+      "step": 38072
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.034377810551375e-07,
+      "loss": 0.635,
+      "step": 38073
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.031417209203275e-07,
+      "loss": 0.5441,
+      "step": 38074
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.028457070093155e-07,
+      "loss": 0.604,
+      "step": 38075
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.025497393236115e-07,
+      "loss": 0.5613,
+      "step": 38076
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.022538178647144e-07,
+      "loss": 0.5767,
+      "step": 38077
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.019579426341341e-07,
+      "loss": 0.6769,
+      "step": 38078
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.016621136333659e-07,
+      "loss": 0.5902,
+      "step": 38079
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.013663308639187e-07,
+      "loss": 0.6183,
+      "step": 38080
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.010705943272924e-07,
+      "loss": 0.6001,
+      "step": 38081
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.007749040249925e-07,
+      "loss": 0.6369,
+      "step": 38082
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.00479259958521e-07,
+      "loss": 0.6297,
+      "step": 38083
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.00183662129378e-07,
+      "loss": 0.6414,
+      "step": 38084
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.998881105390678e-07,
+      "loss": 0.5716,
+      "step": 38085
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.995926051890891e-07,
+      "loss": 0.5793,
+      "step": 38086
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.992971460809474e-07,
+      "loss": 0.597,
+      "step": 38087
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.990017332161405e-07,
+      "loss": 0.609,
+      "step": 38088
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.987063665961748e-07,
+      "loss": 0.5882,
+      "step": 38089
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.984110462225437e-07,
+      "loss": 0.5768,
+      "step": 38090
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.981157720967537e-07,
+      "loss": 0.6598,
+      "step": 38091
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.978205442203014e-07,
+      "loss": 0.6414,
+      "step": 38092
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.975253625946901e-07,
+      "loss": 0.5982,
+      "step": 38093
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.972302272214161e-07,
+      "loss": 0.5472,
+      "step": 38094
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.969351381019842e-07,
+      "loss": 0.5938,
+      "step": 38095
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.966400952378885e-07,
+      "loss": 0.5791,
+      "step": 38096
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.963450986306321e-07,
+      "loss": 0.5499,
+      "step": 38097
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.960501482817097e-07,
+      "loss": 0.6002,
+      "step": 38098
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.957552441926254e-07,
+      "loss": 0.6146,
+      "step": 38099
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.954603863648747e-07,
+      "loss": 0.5865,
+      "step": 38100
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.951655747999566e-07,
+      "loss": 0.6623,
+      "step": 38101
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.948708094993664e-07,
+      "loss": 0.6036,
+      "step": 38102
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.945760904646061e-07,
+      "loss": 0.6023,
+      "step": 38103
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.942814176971714e-07,
+      "loss": 0.617,
+      "step": 38104
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.939867911985601e-07,
+      "loss": 0.5566,
+      "step": 38105
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.936922109702673e-07,
+      "loss": 0.6192,
+      "step": 38106
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.933976770137909e-07,
+      "loss": 0.624,
+      "step": 38107
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.931031893306285e-07,
+      "loss": 0.6154,
+      "step": 38108
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.928087479222747e-07,
+      "loss": 0.5885,
+      "step": 38109
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.925143527902291e-07,
+      "loss": 0.5943,
+      "step": 38110
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.92220003935984e-07,
+      "loss": 0.6234,
+      "step": 38111
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.919257013610371e-07,
+      "loss": 0.6268,
+      "step": 38112
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.916314450668817e-07,
+      "loss": 0.5842,
+      "step": 38113
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.913372350550165e-07,
+      "loss": 0.5796,
+      "step": 38114
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.910430713269313e-07,
+      "loss": 0.6826,
+      "step": 38115
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.907489538841286e-07,
+      "loss": 0.5798,
+      "step": 38116
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.904548827280945e-07,
+      "loss": 0.5914,
+      "step": 38117
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.901608578603283e-07,
+      "loss": 0.5734,
+      "step": 38118
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.898668792823218e-07,
+      "loss": 0.6177,
+      "step": 38119
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.895729469955705e-07,
+      "loss": 0.5616,
+      "step": 38120
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.892790610015678e-07,
+      "loss": 0.6016,
+      "step": 38121
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.889852213018069e-07,
+      "loss": 0.5909,
+      "step": 38122
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.886914278977776e-07,
+      "loss": 0.5725,
+      "step": 38123
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.883976807909778e-07,
+      "loss": 0.6101,
+      "step": 38124
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.881039799828972e-07,
+      "loss": 0.6157,
+      "step": 38125
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.878103254750281e-07,
+      "loss": 0.5718,
+      "step": 38126
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.875167172688659e-07,
+      "loss": 0.5755,
+      "step": 38127
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.872231553658973e-07,
+      "loss": 0.6202,
+      "step": 38128
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.869296397676174e-07,
+      "loss": 0.6164,
+      "step": 38129
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.866361704755155e-07,
+      "loss": 0.6014,
+      "step": 38130
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 8.863427474910869e-07,
+      "loss": 0.561,
+      "step": 38131
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.860493708158191e-07,
+      "loss": 0.566,
+      "step": 38132
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.857560404512034e-07,
+      "loss": 0.5891,
+      "step": 38133
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.854627563987284e-07,
+      "loss": 0.6515,
+      "step": 38134
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.851695186598897e-07,
+      "loss": 0.6312,
+      "step": 38135
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.848763272361705e-07,
+      "loss": 0.6025,
+      "step": 38136
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.845831821290673e-07,
+      "loss": 0.5619,
+      "step": 38137
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.842900833400658e-07,
+      "loss": 0.5739,
+      "step": 38138
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.839970308706558e-07,
+      "loss": 0.5898,
+      "step": 38139
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.837040247223261e-07,
+      "loss": 0.5495,
+      "step": 38140
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.834110648965666e-07,
+      "loss": 0.61,
+      "step": 38141
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.831181513948662e-07,
+      "loss": 0.5841,
+      "step": 38142
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.828252842187123e-07,
+      "loss": 0.59,
+      "step": 38143
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.825324633695931e-07,
+      "loss": 0.5471,
+      "step": 38144
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.822396888489948e-07,
+      "loss": 0.6173,
+      "step": 38145
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.819469606584085e-07,
+      "loss": 0.6281,
+      "step": 38146
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.816542787993188e-07,
+      "loss": 0.5844,
+      "step": 38147
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.813616432732153e-07,
+      "loss": 0.5772,
+      "step": 38148
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.810690540815836e-07,
+      "loss": 0.5961,
+      "step": 38149
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.807765112259114e-07,
+      "loss": 0.6121,
+      "step": 38150
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.804840147076821e-07,
+      "loss": 0.597,
+      "step": 38151
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.801915645283865e-07,
+      "loss": 0.6228,
+      "step": 38152
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.798991606895069e-07,
+      "loss": 0.6188,
+      "step": 38153
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.796068031925332e-07,
+      "loss": 0.616,
+      "step": 38154
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.793144920389451e-07,
+      "loss": 0.543,
+      "step": 38155
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.79022227230234e-07,
+      "loss": 0.586,
+      "step": 38156
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.787300087678796e-07,
+      "loss": 0.6059,
+      "step": 38157
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.784378366533708e-07,
+      "loss": 0.5352,
+      "step": 38158
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.781457108881908e-07,
+      "loss": 0.6412,
+      "step": 38159
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.778536314738251e-07,
+      "loss": 0.5553,
+      "step": 38160
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.775615984117536e-07,
+      "loss": 0.6099,
+      "step": 38161
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.772696117034651e-07,
+      "loss": 0.5964,
+      "step": 38162
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.769776713504419e-07,
+      "loss": 0.576,
+      "step": 38163
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.766857773541648e-07,
+      "loss": 0.6271,
+      "step": 38164
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.763939297161228e-07,
+      "loss": 0.5519,
+      "step": 38165
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.761021284377913e-07,
+      "loss": 0.6299,
+      "step": 38166
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.758103735206592e-07,
+      "loss": 0.6121,
+      "step": 38167
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.75518664966205e-07,
+      "loss": 0.542,
+      "step": 38168
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.752270027759147e-07,
+      "loss": 0.5507,
+      "step": 38169
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.749353869512678e-07,
+      "loss": 0.5971,
+      "step": 38170
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.746438174937466e-07,
+      "loss": 0.5894,
+      "step": 38171
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.74352294404831e-07,
+      "loss": 0.6043,
+      "step": 38172
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.740608176860055e-07,
+      "loss": 0.6628,
+      "step": 38173
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.737693873387487e-07,
+      "loss": 0.5959,
+      "step": 38174
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.734780033645429e-07,
+      "loss": 0.6535,
+      "step": 38175
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.731866657648691e-07,
+      "loss": 0.5941,
+      "step": 38176
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.728953745412072e-07,
+      "loss": 0.6038,
+      "step": 38177
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.726041296950338e-07,
+      "loss": 0.652,
+      "step": 38178
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.723129312278345e-07,
+      "loss": 0.5844,
+      "step": 38179
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.720217791410856e-07,
+      "loss": 0.5847,
+      "step": 38180
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.717306734362685e-07,
+      "loss": 0.614,
+      "step": 38181
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.714396141148585e-07,
+      "loss": 0.6489,
+      "step": 38182
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.711486011783398e-07,
+      "loss": 0.5247,
+      "step": 38183
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.708576346281883e-07,
+      "loss": 0.6069,
+      "step": 38184
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.705667144658802e-07,
+      "loss": 0.586,
+      "step": 38185
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.702758406928991e-07,
+      "loss": 0.6408,
+      "step": 38186
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.699850133107191e-07,
+      "loss": 0.5631,
+      "step": 38187
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.696942323208201e-07,
+      "loss": 0.5571,
+      "step": 38188
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.694034977246768e-07,
+      "loss": 0.5375,
+      "step": 38189
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.6911280952377e-07,
+      "loss": 0.676,
+      "step": 38190
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.688221677195729e-07,
+      "loss": 0.5443,
+      "step": 38191
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.685315723135679e-07,
+      "loss": 0.5925,
+      "step": 38192
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.682410233072258e-07,
+      "loss": 0.667,
+      "step": 38193
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.679505207020267e-07,
+      "loss": 0.5982,
+      "step": 38194
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.676600644994426e-07,
+      "loss": 0.5825,
+      "step": 38195
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.673696547009558e-07,
+      "loss": 0.6682,
+      "step": 38196
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.670792913080373e-07,
+      "loss": 0.6197,
+      "step": 38197
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.667889743221636e-07,
+      "loss": 0.5912,
+      "step": 38198
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.664987037448091e-07,
+      "loss": 0.6314,
+      "step": 38199
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.662084795774505e-07,
+      "loss": 0.5852,
+      "step": 38200
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.659183018215611e-07,
+      "loss": 0.57,
+      "step": 38201
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.656281704786152e-07,
+      "loss": 0.5773,
+      "step": 38202
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.653380855500892e-07,
+      "loss": 0.5668,
+      "step": 38203
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.650480470374545e-07,
+      "loss": 0.6504,
+      "step": 38204
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.647580549421874e-07,
+      "loss": 0.6583,
+      "step": 38205
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.64468109265758e-07,
+      "loss": 0.5997,
+      "step": 38206
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.64178210009643e-07,
+      "loss": 0.585,
+      "step": 38207
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.638883571753142e-07,
+      "loss": 0.5483,
+      "step": 38208
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.635985507642442e-07,
+      "loss": 0.6179,
+      "step": 38209
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.633087907779036e-07,
+      "loss": 0.59,
+      "step": 38210
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.630190772177682e-07,
+      "loss": 0.5644,
+      "step": 38211
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.627294100853078e-07,
+      "loss": 0.6254,
+      "step": 38212
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.624397893819969e-07,
+      "loss": 0.6598,
+      "step": 38213
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.621502151093064e-07,
+      "loss": 0.5584,
+      "step": 38214
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.618606872687052e-07,
+      "loss": 0.6202,
+      "step": 38215
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.615712058616654e-07,
+      "loss": 0.6087,
+      "step": 38216
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.612817708896604e-07,
+      "loss": 0.5855,
+      "step": 38217
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.609923823541577e-07,
+      "loss": 0.622,
+      "step": 38218
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.607030402566307e-07,
+      "loss": 0.5799,
+      "step": 38219
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.604137445985461e-07,
+      "loss": 0.6096,
+      "step": 38220
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.60124495381377e-07,
+      "loss": 0.5971,
+      "step": 38221
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.598352926065923e-07,
+      "loss": 0.5561,
+      "step": 38222
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.595461362756596e-07,
+      "loss": 0.5813,
+      "step": 38223
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.592570263900524e-07,
+      "loss": 0.5701,
+      "step": 38224
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.58967962951237e-07,
+      "loss": 0.5543,
+      "step": 38225
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.586789459606815e-07,
+      "loss": 0.6014,
+      "step": 38226
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.583899754198544e-07,
+      "loss": 0.5371,
+      "step": 38227
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.581010513302268e-07,
+      "loss": 0.5774,
+      "step": 38228
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.578121736932632e-07,
+      "loss": 0.5685,
+      "step": 38229
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.575233425104368e-07,
+      "loss": 0.6475,
+      "step": 38230
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.572345577832086e-07,
+      "loss": 0.588,
+      "step": 38231
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.569458195130509e-07,
+      "loss": 0.6636,
+      "step": 38232
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.566571277014268e-07,
+      "loss": 0.6064,
+      "step": 38233
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.563684823498075e-07,
+      "loss": 0.6157,
+      "step": 38234
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.560798834596584e-07,
+      "loss": 0.6444,
+      "step": 38235
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.557913310324451e-07,
+      "loss": 0.6471,
+      "step": 38236
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.555028250696318e-07,
+      "loss": 0.6001,
+      "step": 38237
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.552143655726897e-07,
+      "loss": 0.6243,
+      "step": 38238
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.549259525430809e-07,
+      "loss": 0.6208,
+      "step": 38239
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.546375859822697e-07,
+      "loss": 0.5739,
+      "step": 38240
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.543492658917241e-07,
+      "loss": 0.6238,
+      "step": 38241
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.540609922729093e-07,
+      "loss": 0.5877,
+      "step": 38242
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.537727651272898e-07,
+      "loss": 0.5758,
+      "step": 38243
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.534845844563267e-07,
+      "loss": 0.5603,
+      "step": 38244
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.531964502614887e-07,
+      "loss": 0.6063,
+      "step": 38245
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.529083625442392e-07,
+      "loss": 0.678,
+      "step": 38246
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.526203213060402e-07,
+      "loss": 0.6597,
+      "step": 38247
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.523323265483552e-07,
+      "loss": 0.6247,
+      "step": 38248
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.520443782726506e-07,
+      "loss": 0.6295,
+      "step": 38249
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.517564764803854e-07,
+      "loss": 0.5974,
+      "step": 38250
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.514686211730272e-07,
+      "loss": 0.5805,
+      "step": 38251
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.511808123520349e-07,
+      "loss": 0.5197,
+      "step": 38252
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.508930500188728e-07,
+      "loss": 0.5991,
+      "step": 38253
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.50605334175001e-07,
+      "loss": 0.559,
+      "step": 38254
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.503176648218847e-07,
+      "loss": 0.6,
+      "step": 38255
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.500300419609841e-07,
+      "loss": 0.4953,
+      "step": 38256
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.497424655937602e-07,
+      "loss": 0.6,
+      "step": 38257
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.49454935721673e-07,
+      "loss": 0.5418,
+      "step": 38258
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.491674523461879e-07,
+      "loss": 0.5826,
+      "step": 38259
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.488800154687615e-07,
+      "loss": 0.6351,
+      "step": 38260
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.485926250908549e-07,
+      "loss": 0.6545,
+      "step": 38261
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.483052812139303e-07,
+      "loss": 0.5483,
+      "step": 38262
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.480179838394475e-07,
+      "loss": 0.6013,
+      "step": 38263
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.477307329688644e-07,
+      "loss": 0.5896,
+      "step": 38264
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.474435286036408e-07,
+      "loss": 0.6254,
+      "step": 38265
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.47156370745239e-07,
+      "loss": 0.6482,
+      "step": 38266
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.468692593951145e-07,
+      "loss": 0.6227,
+      "step": 38267
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.465821945547315e-07,
+      "loss": 0.594,
+      "step": 38268
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.462951762255412e-07,
+      "loss": 0.6254,
+      "step": 38269
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.460082044090079e-07,
+      "loss": 0.6582,
+      "step": 38270
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.457212791065861e-07,
+      "loss": 0.6281,
+      "step": 38271
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.454344003197368e-07,
+      "loss": 0.6714,
+      "step": 38272
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.451475680499166e-07,
+      "loss": 0.5716,
+      "step": 38273
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.448607822985832e-07,
+      "loss": 0.6209,
+      "step": 38274
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.445740430671911e-07,
+      "loss": 0.5956,
+      "step": 38275
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.442873503572013e-07,
+      "loss": 0.549,
+      "step": 38276
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.440007041700693e-07,
+      "loss": 0.6099,
+      "step": 38277
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 8.437141045072483e-07,
+      "loss": 0.5634,
+      "step": 38278
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.434275513702006e-07,
+      "loss": 0.5663,
+      "step": 38279
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.431410447603782e-07,
+      "loss": 0.6344,
+      "step": 38280
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.42854584679238e-07,
+      "loss": 0.5902,
+      "step": 38281
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.425681711282341e-07,
+      "loss": 0.6001,
+      "step": 38282
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.422818041088243e-07,
+      "loss": 0.6123,
+      "step": 38283
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.419954836224631e-07,
+      "loss": 0.6127,
+      "step": 38284
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.417092096706047e-07,
+      "loss": 0.6056,
+      "step": 38285
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.414229822547026e-07,
+      "loss": 0.5763,
+      "step": 38286
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.411368013762144e-07,
+      "loss": 0.6077,
+      "step": 38287
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.408506670365902e-07,
+      "loss": 0.5693,
+      "step": 38288
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.405645792372896e-07,
+      "loss": 0.6003,
+      "step": 38289
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.402785379797595e-07,
+      "loss": 0.6068,
+      "step": 38290
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.399925432654588e-07,
+      "loss": 0.604,
+      "step": 38291
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.39706595095836e-07,
+      "loss": 0.6139,
+      "step": 38292
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.394206934723493e-07,
+      "loss": 0.6081,
+      "step": 38293
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.391348383964459e-07,
+      "loss": 0.598,
+      "step": 38294
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.388490298695861e-07,
+      "loss": 0.6268,
+      "step": 38295
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.385632678932131e-07,
+      "loss": 0.6416,
+      "step": 38296
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.382775524687858e-07,
+      "loss": 0.619,
+      "step": 38297
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.379918835977518e-07,
+      "loss": 0.6288,
+      "step": 38298
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.377062612815645e-07,
+      "loss": 0.6087,
+      "step": 38299
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.374206855216749e-07,
+      "loss": 0.5948,
+      "step": 38300
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.371351563195351e-07,
+      "loss": 0.6359,
+      "step": 38301
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.368496736765952e-07,
+      "loss": 0.5848,
+      "step": 38302
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.36564237594304e-07,
+      "loss": 0.658,
+      "step": 38303
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.362788480741146e-07,
+      "loss": 0.5864,
+      "step": 38304
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.35993505117475e-07,
+      "loss": 0.5873,
+      "step": 38305
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.357082087258395e-07,
+      "loss": 0.5774,
+      "step": 38306
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.354229589006502e-07,
+      "loss": 0.6754,
+      "step": 38307
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.351377556433638e-07,
+      "loss": 0.6035,
+      "step": 38308
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.348525989554235e-07,
+      "loss": 0.6321,
+      "step": 38309
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.345674888382838e-07,
+      "loss": 0.615,
+      "step": 38310
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.342824252933912e-07,
+      "loss": 0.622,
+      "step": 38311
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.339974083221935e-07,
+      "loss": 0.6091,
+      "step": 38312
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.337124379261374e-07,
+      "loss": 0.5616,
+      "step": 38313
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.33427514106675e-07,
+      "loss": 0.6195,
+      "step": 38314
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.331426368652506e-07,
+      "loss": 0.5671,
+      "step": 38315
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.328578062033155e-07,
+      "loss": 0.6194,
+      "step": 38316
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.325730221223116e-07,
+      "loss": 0.5351,
+      "step": 38317
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.322882846236912e-07,
+      "loss": 0.6094,
+      "step": 38318
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.320035937088989e-07,
+      "loss": 0.6109,
+      "step": 38319
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.317189493793787e-07,
+      "loss": 0.5957,
+      "step": 38320
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.314343516365819e-07,
+      "loss": 0.5977,
+      "step": 38321
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.311498004819518e-07,
+      "loss": 0.6195,
+      "step": 38322
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.308652959169361e-07,
+      "loss": 0.6492,
+      "step": 38323
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.305808379429758e-07,
+      "loss": 0.66,
+      "step": 38324
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.302964265615221e-07,
+      "loss": 0.6087,
+      "step": 38325
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.30012061774016e-07,
+      "loss": 0.5855,
+      "step": 38326
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.297277435819085e-07,
+      "loss": 0.5853,
+      "step": 38327
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.294434719866351e-07,
+      "loss": 0.619,
+      "step": 38328
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.29159246989647e-07,
+      "loss": 0.577,
+      "step": 38329
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.288750685923852e-07,
+      "loss": 0.6394,
+      "step": 38330
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.285909367962974e-07,
+      "loss": 0.6076,
+      "step": 38331
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.283068516028225e-07,
+      "loss": 0.5854,
+      "step": 38332
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.280228130134104e-07,
+      "loss": 0.557,
+      "step": 38333
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.277388210294968e-07,
+      "loss": 0.5724,
+      "step": 38334
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.274548756525303e-07,
+      "loss": 0.6214,
+      "step": 38335
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.271709768839498e-07,
+      "loss": 0.5585,
+      "step": 38336
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.268871247252031e-07,
+      "loss": 0.6058,
+      "step": 38337
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.266033191777279e-07,
+      "loss": 0.6289,
+      "step": 38338
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.263195602429686e-07,
+      "loss": 0.658,
+      "step": 38339
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.260358479223651e-07,
+      "loss": 0.5768,
+      "step": 38340
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.257521822173597e-07,
+      "loss": 0.6114,
+      "step": 38341
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.254685631293957e-07,
+      "loss": 0.5304,
+      "step": 38342
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.251849906599108e-07,
+      "loss": 0.6264,
+      "step": 38343
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.249014648103515e-07,
+      "loss": 0.6042,
+      "step": 38344
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.246179855821513e-07,
+      "loss": 0.5459,
+      "step": 38345
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.243345529767555e-07,
+      "loss": 0.5758,
+      "step": 38346
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.24051166995602e-07,
+      "loss": 0.6134,
+      "step": 38347
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.23767827640134e-07,
+      "loss": 0.5838,
+      "step": 38348
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.234845349117881e-07,
+      "loss": 0.6136,
+      "step": 38349
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.232012888120055e-07,
+      "loss": 0.5995,
+      "step": 38350
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.229180893422229e-07,
+      "loss": 0.5368,
+      "step": 38351
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.226349365038822e-07,
+      "loss": 0.5736,
+      "step": 38352
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.223518302984201e-07,
+      "loss": 0.6046,
+      "step": 38353
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.220687707272789e-07,
+      "loss": 0.6196,
+      "step": 38354
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.21785757791892e-07,
+      "loss": 0.5827,
+      "step": 38355
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.215027914937002e-07,
+      "loss": 0.5979,
+      "step": 38356
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.212198718341413e-07,
+      "loss": 0.6102,
+      "step": 38357
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.209369988146498e-07,
+      "loss": 0.5499,
+      "step": 38358
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.206541724366679e-07,
+      "loss": 0.5981,
+      "step": 38359
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.203713927016311e-07,
+      "loss": 0.5695,
+      "step": 38360
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.200886596109748e-07,
+      "loss": 0.545,
+      "step": 38361
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.198059731661345e-07,
+      "loss": 0.6851,
+      "step": 38362
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.195233333685504e-07,
+      "loss": 0.5734,
+      "step": 38363
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.192407402196556e-07,
+      "loss": 0.5462,
+      "step": 38364
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.189581937208913e-07,
+      "loss": 0.5813,
+      "step": 38365
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.186756938736851e-07,
+      "loss": 0.5788,
+      "step": 38366
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.183932406794781e-07,
+      "loss": 0.559,
+      "step": 38367
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.181108341397026e-07,
+      "loss": 0.6091,
+      "step": 38368
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.178284742557963e-07,
+      "loss": 0.6727,
+      "step": 38369
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.175461610291913e-07,
+      "loss": 0.6305,
+      "step": 38370
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.172638944613276e-07,
+      "loss": 0.6135,
+      "step": 38371
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.169816745536308e-07,
+      "loss": 0.657,
+      "step": 38372
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.166995013075418e-07,
+      "loss": 0.5469,
+      "step": 38373
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.164173747244897e-07,
+      "loss": 0.5827,
+      "step": 38374
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.161352948059121e-07,
+      "loss": 0.5814,
+      "step": 38375
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.158532615532422e-07,
+      "loss": 0.6115,
+      "step": 38376
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.155712749679101e-07,
+      "loss": 0.5936,
+      "step": 38377
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.152893350513502e-07,
+      "loss": 0.6148,
+      "step": 38378
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.150074418049936e-07,
+      "loss": 0.6378,
+      "step": 38379
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.147255952302768e-07,
+      "loss": 0.6003,
+      "step": 38380
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.144437953286277e-07,
+      "loss": 0.6032,
+      "step": 38381
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.141620421014806e-07,
+      "loss": 0.5943,
+      "step": 38382
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.138803355502656e-07,
+      "loss": 0.6263,
+      "step": 38383
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.135986756764147e-07,
+      "loss": 0.636,
+      "step": 38384
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.133170624813592e-07,
+      "loss": 0.622,
+      "step": 38385
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.13035495966531e-07,
+      "loss": 0.6669,
+      "step": 38386
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.127539761333602e-07,
+      "loss": 0.6025,
+      "step": 38387
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.12472502983277e-07,
+      "loss": 0.6502,
+      "step": 38388
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.121910765177099e-07,
+      "loss": 0.5688,
+      "step": 38389
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.119096967380924e-07,
+      "loss": 0.6392,
+      "step": 38390
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.11628363645851e-07,
+      "loss": 0.6069,
+      "step": 38391
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.113470772424204e-07,
+      "loss": 0.6731,
+      "step": 38392
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.110658375292224e-07,
+      "loss": 0.5787,
+      "step": 38393
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.107846445076928e-07,
+      "loss": 0.5609,
+      "step": 38394
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.105034981792548e-07,
+      "loss": 0.5903,
+      "step": 38395
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.102223985453417e-07,
+      "loss": 0.649,
+      "step": 38396
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.099413456073801e-07,
+      "loss": 0.6275,
+      "step": 38397
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.096603393667978e-07,
+      "loss": 0.6119,
+      "step": 38398
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.093793798250227e-07,
+      "loss": 0.6373,
+      "step": 38399
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.090984669834811e-07,
+      "loss": 0.6304,
+      "step": 38400
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.088176008436044e-07,
+      "loss": 0.5329,
+      "step": 38401
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.085367814068146e-07,
+      "loss": 0.5668,
+      "step": 38402
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.08256008674545e-07,
+      "loss": 0.6222,
+      "step": 38403
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.079752826482146e-07,
+      "loss": 0.5647,
+      "step": 38404
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.076946033292554e-07,
+      "loss": 0.6302,
+      "step": 38405
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.074139707190909e-07,
+      "loss": 0.6387,
+      "step": 38406
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.071333848191498e-07,
+      "loss": 0.6045,
+      "step": 38407
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.068528456308567e-07,
+      "loss": 0.5876,
+      "step": 38408
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.065723531556358e-07,
+      "loss": 0.578,
+      "step": 38409
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.062919073949116e-07,
+      "loss": 0.6167,
+      "step": 38410
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.06011508350113e-07,
+      "loss": 0.5575,
+      "step": 38411
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.057311560226599e-07,
+      "loss": 0.5036,
+      "step": 38412
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.054508504139813e-07,
+      "loss": 0.5766,
+      "step": 38413
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.051705915255004e-07,
+      "loss": 0.5949,
+      "step": 38414
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.048903793586405e-07,
+      "loss": 0.5832,
+      "step": 38415
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.046102139148237e-07,
+      "loss": 0.6262,
+      "step": 38416
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.043300951954769e-07,
+      "loss": 0.6157,
+      "step": 38417
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.04050023202022e-07,
+      "loss": 0.6066,
+      "step": 38418
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.037699979358815e-07,
+      "loss": 0.6552,
+      "step": 38419
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.034900193984796e-07,
+      "loss": 0.5295,
+      "step": 38420
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.032100875912364e-07,
+      "loss": 0.6171,
+      "step": 38421
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.029302025155783e-07,
+      "loss": 0.6451,
+      "step": 38422
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.026503641729233e-07,
+      "loss": 0.5764,
+      "step": 38423
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.023705725646969e-07,
+      "loss": 0.5627,
+      "step": 38424
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 8.02090827692319e-07,
+      "loss": 0.582,
+      "step": 38425
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.018111295572118e-07,
+      "loss": 0.5787,
+      "step": 38426
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.015314781607941e-07,
+      "loss": 0.5494,
+      "step": 38427
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.012518735044906e-07,
+      "loss": 0.663,
+      "step": 38428
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.009723155897186e-07,
+      "loss": 0.577,
+      "step": 38429
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.006928044179041e-07,
+      "loss": 0.5582,
+      "step": 38430
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.004133399904601e-07,
+      "loss": 0.555,
+      "step": 38431
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 8.001339223088112e-07,
+      "loss": 0.5825,
+      "step": 38432
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.99854551374375e-07,
+      "loss": 0.6039,
+      "step": 38433
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.995752271885737e-07,
+      "loss": 0.6054,
+      "step": 38434
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.992959497528252e-07,
+      "loss": 0.6015,
+      "step": 38435
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.990167190685482e-07,
+      "loss": 0.6526,
+      "step": 38436
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.987375351371629e-07,
+      "loss": 0.5863,
+      "step": 38437
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.984583979600847e-07,
+      "loss": 0.5752,
+      "step": 38438
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.981793075387357e-07,
+      "loss": 0.5844,
+      "step": 38439
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.979002638745303e-07,
+      "loss": 0.544,
+      "step": 38440
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.976212669688932e-07,
+      "loss": 0.6427,
+      "step": 38441
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.973423168232331e-07,
+      "loss": 0.5266,
+      "step": 38442
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.970634134389732e-07,
+      "loss": 0.5868,
+      "step": 38443
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.967845568175292e-07,
+      "loss": 0.6702,
+      "step": 38444
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.965057469603188e-07,
+      "loss": 0.6362,
+      "step": 38445
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.962269838687575e-07,
+      "loss": 0.5717,
+      "step": 38446
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.959482675442632e-07,
+      "loss": 0.6279,
+      "step": 38447
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.956695979882501e-07,
+      "loss": 0.6137,
+      "step": 38448
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.953909752021361e-07,
+      "loss": 0.5698,
+      "step": 38449
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.951123991873344e-07,
+      "loss": 0.5634,
+      "step": 38450
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.948338699452651e-07,
+      "loss": 0.6237,
+      "step": 38451
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.945553874773393e-07,
+      "loss": 0.5294,
+      "step": 38452
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.942769517849746e-07,
+      "loss": 0.663,
+      "step": 38453
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.939985628695834e-07,
+      "loss": 0.5644,
+      "step": 38454
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.937202207325823e-07,
+      "loss": 0.6155,
+      "step": 38455
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.934419253753856e-07,
+      "loss": 0.6298,
+      "step": 38456
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.931636767994067e-07,
+      "loss": 0.5781,
+      "step": 38457
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.928854750060588e-07,
+      "loss": 0.585,
+      "step": 38458
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.926073199967555e-07,
+      "loss": 0.6052,
+      "step": 38459
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.92329211772912e-07,
+      "loss": 0.58,
+      "step": 38460
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.920511503359385e-07,
+      "loss": 0.6188,
+      "step": 38461
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.917731356872527e-07,
+      "loss": 0.5938,
+      "step": 38462
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.914951678282634e-07,
+      "loss": 0.5971,
+      "step": 38463
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.91217246760384e-07,
+      "loss": 0.6519,
+      "step": 38464
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.909393724850245e-07,
+      "loss": 0.5715,
+      "step": 38465
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.906615450036014e-07,
+      "loss": 0.551,
+      "step": 38466
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.903837643175228e-07,
+      "loss": 0.6018,
+      "step": 38467
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.901060304282038e-07,
+      "loss": 0.6447,
+      "step": 38468
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.898283433370501e-07,
+      "loss": 0.5855,
+      "step": 38469
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.895507030454785e-07,
+      "loss": 0.5818,
+      "step": 38470
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.892731095548944e-07,
+      "loss": 0.5998,
+      "step": 38471
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.889955628667134e-07,
+      "loss": 0.579,
+      "step": 38472
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.887180629823432e-07,
+      "loss": 0.6483,
+      "step": 38473
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.884406099031949e-07,
+      "loss": 0.6295,
+      "step": 38474
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.881632036306752e-07,
+      "loss": 0.6133,
+      "step": 38475
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.878858441661985e-07,
+      "loss": 0.5984,
+      "step": 38476
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.876085315111715e-07,
+      "loss": 0.5141,
+      "step": 38477
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.873312656670007e-07,
+      "loss": 0.6309,
+      "step": 38478
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.870540466351017e-07,
+      "loss": 0.5917,
+      "step": 38479
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.867768744168769e-07,
+      "loss": 0.6032,
+      "step": 38480
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.864997490137383e-07,
+      "loss": 0.6079,
+      "step": 38481
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.862226704270904e-07,
+      "loss": 0.5661,
+      "step": 38482
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.859456386583453e-07,
+      "loss": 0.5868,
+      "step": 38483
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.856686537089086e-07,
+      "loss": 0.6022,
+      "step": 38484
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.853917155801893e-07,
+      "loss": 0.604,
+      "step": 38485
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.851148242735907e-07,
+      "loss": 0.652,
+      "step": 38486
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.848379797905237e-07,
+      "loss": 0.5953,
+      "step": 38487
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.845611821323928e-07,
+      "loss": 0.6209,
+      "step": 38488
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.84284431300607e-07,
+      "loss": 0.5912,
+      "step": 38489
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.840077272965707e-07,
+      "loss": 0.6153,
+      "step": 38490
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.837310701216904e-07,
+      "loss": 0.6223,
+      "step": 38491
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.834544597773696e-07,
+      "loss": 0.6804,
+      "step": 38492
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.831778962650182e-07,
+      "loss": 0.6239,
+      "step": 38493
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.829013795860385e-07,
+      "loss": 0.6366,
+      "step": 38494
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.826249097418359e-07,
+      "loss": 0.5192,
+      "step": 38495
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.82348486733815e-07,
+      "loss": 0.5742,
+      "step": 38496
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.820721105633822e-07,
+      "loss": 0.5695,
+      "step": 38497
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.817957812319399e-07,
+      "loss": 0.6182,
+      "step": 38498
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.815194987408925e-07,
+      "loss": 0.606,
+      "step": 38499
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.812432630916467e-07,
+      "loss": 0.6433,
+      "step": 38500
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.809670742856001e-07,
+      "loss": 0.5794,
+      "step": 38501
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.806909323241629e-07,
+      "loss": 0.5809,
+      "step": 38502
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.804148372087317e-07,
+      "loss": 0.6382,
+      "step": 38503
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.801387889407153e-07,
+      "loss": 0.5605,
+      "step": 38504
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.798627875215125e-07,
+      "loss": 0.6515,
+      "step": 38505
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.795868329525303e-07,
+      "loss": 0.6334,
+      "step": 38506
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.793109252351638e-07,
+      "loss": 0.6575,
+      "step": 38507
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.790350643708211e-07,
+      "loss": 0.5649,
+      "step": 38508
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.787592503608998e-07,
+      "loss": 0.626,
+      "step": 38509
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.784834832068044e-07,
+      "loss": 0.6572,
+      "step": 38510
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.782077629099361e-07,
+      "loss": 0.5797,
+      "step": 38511
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.779320894716946e-07,
+      "loss": 0.6013,
+      "step": 38512
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.776564628934791e-07,
+      "loss": 0.6132,
+      "step": 38513
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.773808831766938e-07,
+      "loss": 0.6332,
+      "step": 38514
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.771053503227366e-07,
+      "loss": 0.63,
+      "step": 38515
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.768298643330074e-07,
+      "loss": 0.5895,
+      "step": 38516
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.765544252089086e-07,
+      "loss": 0.5783,
+      "step": 38517
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.762790329518378e-07,
+      "loss": 0.5473,
+      "step": 38518
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.760036875631938e-07,
+      "loss": 0.6018,
+      "step": 38519
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.757283890443746e-07,
+      "loss": 0.6312,
+      "step": 38520
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.754531373967833e-07,
+      "loss": 0.6455,
+      "step": 38521
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.751779326218167e-07,
+      "loss": 0.5524,
+      "step": 38522
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.749027747208715e-07,
+      "loss": 0.5419,
+      "step": 38523
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.746276636953454e-07,
+      "loss": 0.5946,
+      "step": 38524
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.743525995466406e-07,
+      "loss": 0.5437,
+      "step": 38525
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.740775822761493e-07,
+      "loss": 0.6203,
+      "step": 38526
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.738026118852737e-07,
+      "loss": 0.5583,
+      "step": 38527
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.735276883754095e-07,
+      "loss": 0.5985,
+      "step": 38528
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.732528117479532e-07,
+      "loss": 0.6422,
+      "step": 38529
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.729779820042993e-07,
+      "loss": 0.6126,
+      "step": 38530
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.727031991458489e-07,
+      "loss": 0.5562,
+      "step": 38531
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.724284631739953e-07,
+      "loss": 0.5725,
+      "step": 38532
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.721537740901352e-07,
+      "loss": 0.5997,
+      "step": 38533
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.718791318956631e-07,
+      "loss": 0.6074,
+      "step": 38534
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.716045365919766e-07,
+      "loss": 0.673,
+      "step": 38535
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.713299881804703e-07,
+      "loss": 0.6425,
+      "step": 38536
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.710554866625375e-07,
+      "loss": 0.6135,
+      "step": 38537
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.707810320395782e-07,
+      "loss": 0.6354,
+      "step": 38538
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.705066243129799e-07,
+      "loss": 0.5767,
+      "step": 38539
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.702322634841419e-07,
+      "loss": 0.6611,
+      "step": 38540
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.699579495544562e-07,
+      "loss": 0.5995,
+      "step": 38541
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.696836825253173e-07,
+      "loss": 0.6266,
+      "step": 38542
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.694094623981185e-07,
+      "loss": 0.6682,
+      "step": 38543
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.691352891742565e-07,
+      "loss": 0.613,
+      "step": 38544
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.688611628551179e-07,
+      "loss": 0.5566,
+      "step": 38545
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.685870834421006e-07,
+      "loss": 0.5801,
+      "step": 38546
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.683130509365943e-07,
+      "loss": 0.6837,
+      "step": 38547
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.680390653399949e-07,
+      "loss": 0.5423,
+      "step": 38548
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.677651266536934e-07,
+      "loss": 0.5664,
+      "step": 38549
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.674912348790797e-07,
+      "loss": 0.5906,
+      "step": 38550
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.672173900175461e-07,
+      "loss": 0.5782,
+      "step": 38551
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.669435920704871e-07,
+      "loss": 0.6244,
+      "step": 38552
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.666698410392903e-07,
+      "loss": 0.664,
+      "step": 38553
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.663961369253481e-07,
+      "loss": 0.6122,
+      "step": 38554
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.661224797300515e-07,
+      "loss": 0.6513,
+      "step": 38555
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.658488694547916e-07,
+      "loss": 0.5672,
+      "step": 38556
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.655753061009575e-07,
+      "loss": 0.5478,
+      "step": 38557
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.653017896699388e-07,
+      "loss": 0.629,
+      "step": 38558
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.65028320163127e-07,
+      "loss": 0.6198,
+      "step": 38559
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.647548975819119e-07,
+      "loss": 0.6342,
+      "step": 38560
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.644815219276813e-07,
+      "loss": 0.6141,
+      "step": 38561
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.642081932018231e-07,
+      "loss": 0.6438,
+      "step": 38562
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.639349114057293e-07,
+      "loss": 0.6177,
+      "step": 38563
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.636616765407856e-07,
+      "loss": 0.6607,
+      "step": 38564
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.633884886083854e-07,
+      "loss": 0.6058,
+      "step": 38565
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.631153476099096e-07,
+      "loss": 0.5964,
+      "step": 38566
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.628422535467517e-07,
+      "loss": 0.5492,
+      "step": 38567
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.625692064202961e-07,
+      "loss": 0.5401,
+      "step": 38568
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.622962062319339e-07,
+      "loss": 0.5904,
+      "step": 38569
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.620232529830495e-07,
+      "loss": 0.6294,
+      "step": 38570
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.617503466750298e-07,
+      "loss": 0.5866,
+      "step": 38571
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.614774873092612e-07,
+      "loss": 0.6481,
+      "step": 38572
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.612046748871327e-07,
+      "loss": 0.6007,
+      "step": 38573
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.609319094100287e-07,
+      "loss": 0.6343,
+      "step": 38574
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.606591908793348e-07,
+      "loss": 0.5973,
+      "step": 38575
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.603865192964377e-07,
+      "loss": 0.5619,
+      "step": 38576
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.601138946627229e-07,
+      "loss": 0.6099,
+      "step": 38577
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.59841316979576e-07,
+      "loss": 0.6261,
+      "step": 38578
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.595687862483803e-07,
+      "loss": 0.633,
+      "step": 38579
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.592963024705225e-07,
+      "loss": 0.5907,
+      "step": 38580
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.590238656473848e-07,
+      "loss": 0.6512,
+      "step": 38581
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.587514757803572e-07,
+      "loss": 0.5839,
+      "step": 38582
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.584791328708152e-07,
+      "loss": 0.6217,
+      "step": 38583
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.5820683692015e-07,
+      "loss": 0.5992,
+      "step": 38584
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.579345879297406e-07,
+      "loss": 0.5923,
+      "step": 38585
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.576623859009735e-07,
+      "loss": 0.554,
+      "step": 38586
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.57390230835231e-07,
+      "loss": 0.6064,
+      "step": 38587
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.571181227338953e-07,
+      "loss": 0.5958,
+      "step": 38588
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.568460615983475e-07,
+      "loss": 0.5889,
+      "step": 38589
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.565740474299732e-07,
+      "loss": 0.6871,
+      "step": 38590
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.563020802301535e-07,
+      "loss": 0.5824,
+      "step": 38591
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.560301600002706e-07,
+      "loss": 0.5833,
+      "step": 38592
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.557582867417035e-07,
+      "loss": 0.6052,
+      "step": 38593
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.554864604558376e-07,
+      "loss": 0.5861,
+      "step": 38594
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.552146811440531e-07,
+      "loss": 0.6427,
+      "step": 38595
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.549429488077276e-07,
+      "loss": 0.5387,
+      "step": 38596
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.546712634482467e-07,
+      "loss": 0.5748,
+      "step": 38597
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.543996250669893e-07,
+      "loss": 0.5104,
+      "step": 38598
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.541280336653345e-07,
+      "loss": 0.6512,
+      "step": 38599
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.538564892446609e-07,
+      "loss": 0.5954,
+      "step": 38600
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.53584991806352e-07,
+      "loss": 0.5707,
+      "step": 38601
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.533135413517845e-07,
+      "loss": 0.6951,
+      "step": 38602
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.530421378823416e-07,
+      "loss": 0.6246,
+      "step": 38603
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.527707813993967e-07,
+      "loss": 0.6103,
+      "step": 38604
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.524994719043344e-07,
+      "loss": 0.5605,
+      "step": 38605
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.522282093985279e-07,
+      "loss": 0.5716,
+      "step": 38606
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.519569938833593e-07,
+      "loss": 0.5785,
+      "step": 38607
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.516858253602044e-07,
+      "loss": 0.6362,
+      "step": 38608
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.514147038304464e-07,
+      "loss": 0.529,
+      "step": 38609
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.511436292954544e-07,
+      "loss": 0.5813,
+      "step": 38610
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.508726017566126e-07,
+      "loss": 0.6211,
+      "step": 38611
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.506016212152956e-07,
+      "loss": 0.6629,
+      "step": 38612
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.503306876728788e-07,
+      "loss": 0.5957,
+      "step": 38613
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.500598011307425e-07,
+      "loss": 0.6113,
+      "step": 38614
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.497889615902621e-07,
+      "loss": 0.598,
+      "step": 38615
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.495181690528119e-07,
+      "loss": 0.6409,
+      "step": 38616
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.492474235197677e-07,
+      "loss": 0.5265,
+      "step": 38617
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.489767249925084e-07,
+      "loss": 0.5292,
+      "step": 38618
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.487060734724072e-07,
+      "loss": 0.6125,
+      "step": 38619
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.484354689608408e-07,
+      "loss": 0.602,
+      "step": 38620
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.481649114591804e-07,
+      "loss": 0.6084,
+      "step": 38621
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.478944009688061e-07,
+      "loss": 0.6074,
+      "step": 38622
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.476239374910876e-07,
+      "loss": 0.6074,
+      "step": 38623
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.473535210274042e-07,
+      "loss": 0.6536,
+      "step": 38624
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.470831515791266e-07,
+      "loss": 0.5499,
+      "step": 38625
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.468128291476295e-07,
+      "loss": 0.5947,
+      "step": 38626
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.46542553734284e-07,
+      "loss": 0.5998,
+      "step": 38627
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.46272325340468e-07,
+      "loss": 0.6497,
+      "step": 38628
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.460021439675525e-07,
+      "loss": 0.6033,
+      "step": 38629
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.457320096169107e-07,
+      "loss": 0.5955,
+      "step": 38630
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.454619222899128e-07,
+      "loss": 0.5519,
+      "step": 38631
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.451918819879345e-07,
+      "loss": 0.6188,
+      "step": 38632
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.449218887123466e-07,
+      "loss": 0.5821,
+      "step": 38633
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.446519424645193e-07,
+      "loss": 0.5633,
+      "step": 38634
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.443820432458282e-07,
+      "loss": 0.61,
+      "step": 38635
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.441121910576432e-07,
+      "loss": 0.6431,
+      "step": 38636
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.438423859013333e-07,
+      "loss": 0.6114,
+      "step": 38637
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.435726277782707e-07,
+      "loss": 0.6108,
+      "step": 38638
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.433029166898276e-07,
+      "loss": 0.5206,
+      "step": 38639
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.430332526373718e-07,
+      "loss": 0.5796,
+      "step": 38640
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.427636356222779e-07,
+      "loss": 0.5229,
+      "step": 38641
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.424940656459101e-07,
+      "loss": 0.6236,
+      "step": 38642
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.422245427096431e-07,
+      "loss": 0.6058,
+      "step": 38643
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.419550668148423e-07,
+      "loss": 0.5783,
+      "step": 38644
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.416856379628811e-07,
+      "loss": 0.5737,
+      "step": 38645
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.414162561551241e-07,
+      "loss": 0.6077,
+      "step": 38646
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.411469213929468e-07,
+      "loss": 0.6092,
+      "step": 38647
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.408776336777102e-07,
+      "loss": 0.6103,
+      "step": 38648
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.406083930107877e-07,
+      "loss": 0.5708,
+      "step": 38649
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.40339199393546e-07,
+      "loss": 0.5964,
+      "step": 38650
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.400700528273508e-07,
+      "loss": 0.5732,
+      "step": 38651
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.39800953313572e-07,
+      "loss": 0.6383,
+      "step": 38652
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.395319008535784e-07,
+      "loss": 0.6128,
+      "step": 38653
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.392628954487335e-07,
+      "loss": 0.5911,
+      "step": 38654
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.389939371004051e-07,
+      "loss": 0.6052,
+      "step": 38655
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.387250258099632e-07,
+      "loss": 0.5708,
+      "step": 38656
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.384561615787711e-07,
+      "loss": 0.5594,
+      "step": 38657
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.381873444081955e-07,
+      "loss": 0.5826,
+      "step": 38658
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.37918574299601e-07,
+      "loss": 0.6275,
+      "step": 38659
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.376498512543562e-07,
+      "loss": 0.5762,
+      "step": 38660
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.373811752738225e-07,
+      "loss": 0.599,
+      "step": 38661
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.371125463593709e-07,
+      "loss": 0.6331,
+      "step": 38662
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.368439645123626e-07,
+      "loss": 0.6121,
+      "step": 38663
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.365754297341631e-07,
+      "loss": 0.5991,
+      "step": 38664
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.363069420261348e-07,
+      "loss": 0.6169,
+      "step": 38665
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.360385013896454e-07,
+      "loss": 0.634,
+      "step": 38666
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.35770107826057e-07,
+      "loss": 0.5874,
+      "step": 38667
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.355017613367354e-07,
+      "loss": 0.668,
+      "step": 38668
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.352334619230406e-07,
+      "loss": 0.6186,
+      "step": 38669
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.349652095863391e-07,
+      "loss": 0.6411,
+      "step": 38670
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.346970043279933e-07,
+      "loss": 0.5939,
+      "step": 38671
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.344288461493632e-07,
+      "loss": 0.6165,
+      "step": 38672
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.341607350518154e-07,
+      "loss": 0.5804,
+      "step": 38673
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.338926710367111e-07,
+      "loss": 0.6129,
+      "step": 38674
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.336246541054115e-07,
+      "loss": 0.6061,
+      "step": 38675
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.333566842592765e-07,
+      "loss": 0.6617,
+      "step": 38676
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.330887614996717e-07,
+      "loss": 0.6231,
+      "step": 38677
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.328208858279562e-07,
+      "loss": 0.6114,
+      "step": 38678
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.325530572454942e-07,
+      "loss": 0.6412,
+      "step": 38679
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.322852757536414e-07,
+      "loss": 0.6388,
+      "step": 38680
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.320175413537634e-07,
+      "loss": 0.6044,
+      "step": 38681
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.317498540472168e-07,
+      "loss": 0.573,
+      "step": 38682
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.314822138353639e-07,
+      "loss": 0.6297,
+      "step": 38683
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.312146207195658e-07,
+      "loss": 0.5963,
+      "step": 38684
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.309470747011793e-07,
+      "loss": 0.6383,
+      "step": 38685
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.306795757815644e-07,
+      "loss": 0.5593,
+      "step": 38686
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.304121239620832e-07,
+      "loss": 0.625,
+      "step": 38687
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.301447192440902e-07,
+      "loss": 0.5917,
+      "step": 38688
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.29877361628949e-07,
+      "loss": 0.5802,
+      "step": 38689
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.296100511180159e-07,
+      "loss": 0.6099,
+      "step": 38690
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.293427877126491e-07,
+      "loss": 0.5672,
+      "step": 38691
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.290755714142061e-07,
+      "loss": 0.5763,
+      "step": 38692
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.288084022240438e-07,
+      "loss": 0.5761,
+      "step": 38693
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.285412801435232e-07,
+      "loss": 0.6271,
+      "step": 38694
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.282742051739988e-07,
+      "loss": 0.5798,
+      "step": 38695
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.280071773168295e-07,
+      "loss": 0.5821,
+      "step": 38696
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.277401965733688e-07,
+      "loss": 0.5969,
+      "step": 38697
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.274732629449777e-07,
+      "loss": 0.5851,
+      "step": 38698
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.272063764330084e-07,
+      "loss": 0.5594,
+      "step": 38699
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.26939537038821e-07,
+      "loss": 0.5923,
+      "step": 38700
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.2667274476377e-07,
+      "loss": 0.6368,
+      "step": 38701
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.26405999609211e-07,
+      "loss": 0.6295,
+      "step": 38702
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.261393015764961e-07,
+      "loss": 0.648,
+      "step": 38703
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.258726506669855e-07,
+      "loss": 0.6084,
+      "step": 38704
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.256060468820303e-07,
+      "loss": 0.555,
+      "step": 38705
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.253394902229905e-07,
+      "loss": 0.6361,
+      "step": 38706
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.250729806912138e-07,
+      "loss": 0.5999,
+      "step": 38707
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.248065182880592e-07,
+      "loss": 0.6453,
+      "step": 38708
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.245401030148769e-07,
+      "loss": 0.5857,
+      "step": 38709
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.242737348730245e-07,
+      "loss": 0.5674,
+      "step": 38710
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.240074138638542e-07,
+      "loss": 0.6149,
+      "step": 38711
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.237411399887185e-07,
+      "loss": 0.604,
+      "step": 38712
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.234749132489716e-07,
+      "loss": 0.6386,
+      "step": 38713
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.232087336459625e-07,
+      "loss": 0.5863,
+      "step": 38714
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.229426011810492e-07,
+      "loss": 0.6509,
+      "step": 38715
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.226765158555793e-07,
+      "loss": 0.5947,
+      "step": 38716
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.224104776709106e-07,
+      "loss": 0.686,
+      "step": 38717
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.221444866283878e-07,
+      "loss": 0.5998,
+      "step": 38718
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.218785427293673e-07,
+      "loss": 0.5534,
+      "step": 38719
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.216126459751982e-07,
+      "loss": 0.5875,
+      "step": 38720
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.213467963672338e-07,
+      "loss": 0.6048,
+      "step": 38721
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.210809939068242e-07,
+      "loss": 0.6532,
+      "step": 38722
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.208152385953194e-07,
+      "loss": 0.6568,
+      "step": 38723
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.205495304340682e-07,
+      "loss": 0.687,
+      "step": 38724
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.202838694244241e-07,
+      "loss": 0.5982,
+      "step": 38725
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.200182555677327e-07,
+      "loss": 0.5663,
+      "step": 38726
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.197526888653495e-07,
+      "loss": 0.5534,
+      "step": 38727
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.1948716931862e-07,
+      "loss": 0.5924,
+      "step": 38728
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.192216969288935e-07,
+      "loss": 0.615,
+      "step": 38729
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.189562716975207e-07,
+      "loss": 0.5578,
+      "step": 38730
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.186908936258474e-07,
+      "loss": 0.6214,
+      "step": 38731
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.184255627152248e-07,
+      "loss": 0.6678,
+      "step": 38732
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.181602789670006e-07,
+      "loss": 0.6023,
+      "step": 38733
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.178950423825215e-07,
+      "loss": 0.5414,
+      "step": 38734
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.176298529631353e-07,
+      "loss": 0.5941,
+      "step": 38735
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.173647107101922e-07,
+      "loss": 0.5697,
+      "step": 38736
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.170996156250354e-07,
+      "loss": 0.6289,
+      "step": 38737
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.16834567709016e-07,
+      "loss": 0.5852,
+      "step": 38738
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.165695669634787e-07,
+      "loss": 0.5955,
+      "step": 38739
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.1630461338977e-07,
+      "loss": 0.6127,
+      "step": 38740
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.160397069892345e-07,
+      "loss": 0.6221,
+      "step": 38741
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.15774847763222e-07,
+      "loss": 0.605,
+      "step": 38742
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.155100357130751e-07,
+      "loss": 0.6214,
+      "step": 38743
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.152452708401447e-07,
+      "loss": 0.624,
+      "step": 38744
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.149805531457688e-07,
+      "loss": 0.6443,
+      "step": 38745
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.147158826312972e-07,
+      "loss": 0.534,
+      "step": 38746
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.144512592980712e-07,
+      "loss": 0.5674,
+      "step": 38747
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.141866831474398e-07,
+      "loss": 0.6756,
+      "step": 38748
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.139221541807462e-07,
+      "loss": 0.6378,
+      "step": 38749
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.136576723993326e-07,
+      "loss": 0.612,
+      "step": 38750
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.13393237804545e-07,
+      "loss": 0.5792,
+      "step": 38751
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.13128850397724e-07,
+      "loss": 0.5947,
+      "step": 38752
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.128645101802167e-07,
+      "loss": 0.7097,
+      "step": 38753
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.12600217153363e-07,
+      "loss": 0.6324,
+      "step": 38754
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.123359713185107e-07,
+      "loss": 0.5907,
+      "step": 38755
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.120717726769954e-07,
+      "loss": 0.6073,
+      "step": 38756
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.118076212301661e-07,
+      "loss": 0.5723,
+      "step": 38757
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.115435169793605e-07,
+      "loss": 0.6286,
+      "step": 38758
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.112794599259243e-07,
+      "loss": 0.6199,
+      "step": 38759
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.110154500711974e-07,
+      "loss": 0.5904,
+      "step": 38760
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.107514874165222e-07,
+      "loss": 0.5206,
+      "step": 38761
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.104875719632365e-07,
+      "loss": 0.6315,
+      "step": 38762
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.102237037126858e-07,
+      "loss": 0.664,
+      "step": 38763
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.09959882666208e-07,
+      "loss": 0.5383,
+      "step": 38764
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.096961088251464e-07,
+      "loss": 0.6308,
+      "step": 38765
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.094323821908388e-07,
+      "loss": 0.578,
+      "step": 38766
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.091687027646266e-07,
+      "loss": 0.6032,
+      "step": 38767
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.089050705478473e-07,
+      "loss": 0.6316,
+      "step": 38768
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.086414855418444e-07,
+      "loss": 0.5273,
+      "step": 38769
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.083779477479557e-07,
+      "loss": 0.6371,
+      "step": 38770
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.081144571675192e-07,
+      "loss": 0.5629,
+      "step": 38771
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.078510138018745e-07,
+      "loss": 0.6301,
+      "step": 38772
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.075876176523578e-07,
+      "loss": 0.6168,
+      "step": 38773
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.073242687203119e-07,
+      "loss": 0.601,
+      "step": 38774
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.070609670070716e-07,
+      "loss": 0.5788,
+      "step": 38775
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.067977125139791e-07,
+      "loss": 0.5432,
+      "step": 38776
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.065345052423655e-07,
+      "loss": 0.5658,
+      "step": 38777
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.062713451935743e-07,
+      "loss": 0.6194,
+      "step": 38778
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.060082323689377e-07,
+      "loss": 0.6148,
+      "step": 38779
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.057451667697979e-07,
+      "loss": 0.5293,
+      "step": 38780
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.054821483974861e-07,
+      "loss": 0.5845,
+      "step": 38781
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.052191772533445e-07,
+      "loss": 0.5846,
+      "step": 38782
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.049562533387044e-07,
+      "loss": 0.592,
+      "step": 38783
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.046933766549047e-07,
+      "loss": 0.5688,
+      "step": 38784
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.044305472032786e-07,
+      "loss": 0.619,
+      "step": 38785
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.041677649851653e-07,
+      "loss": 0.6861,
+      "step": 38786
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.039050300018968e-07,
+      "loss": 0.6587,
+      "step": 38787
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.0364234225481e-07,
+      "loss": 0.5961,
+      "step": 38788
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.033797017452382e-07,
+      "loss": 0.6177,
+      "step": 38789
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.03117108474517e-07,
+      "loss": 0.528,
+      "step": 38790
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.028545624439809e-07,
+      "loss": 0.56,
+      "step": 38791
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.025920636549632e-07,
+      "loss": 0.6112,
+      "step": 38792
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.023296121087997e-07,
+      "loss": 0.5724,
+      "step": 38793
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.020672078068202e-07,
+      "loss": 0.5384,
+      "step": 38794
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.018048507503616e-07,
+      "loss": 0.5612,
+      "step": 38795
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.015425409407539e-07,
+      "loss": 0.6508,
+      "step": 38796
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.012802783793327e-07,
+      "loss": 0.6054,
+      "step": 38797
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.010180630674301e-07,
+      "loss": 0.5426,
+      "step": 38798
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.007558950063786e-07,
+      "loss": 0.6265,
+      "step": 38799
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.004937741975082e-07,
+      "loss": 0.6343,
+      "step": 38800
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.002317006421532e-07,
+      "loss": 0.6396,
+      "step": 38801
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.999696743416439e-07,
+      "loss": 0.6225,
+      "step": 38802
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.997076952973147e-07,
+      "loss": 0.5819,
+      "step": 38803
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.994457635104912e-07,
+      "loss": 0.5532,
+      "step": 38804
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.9918387898251e-07,
+      "loss": 0.5523,
+      "step": 38805
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.989220417146969e-07,
+      "loss": 0.5857,
+      "step": 38806
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.986602517083874e-07,
+      "loss": 0.5903,
+      "step": 38807
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.983985089649081e-07,
+      "loss": 0.5704,
+      "step": 38808
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.981368134855904e-07,
+      "loss": 0.5904,
+      "step": 38809
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.978751652717641e-07,
+      "loss": 0.6297,
+      "step": 38810
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.976135643247573e-07,
+      "loss": 0.6305,
+      "step": 38811
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.973520106459008e-07,
+      "loss": 0.6286,
+      "step": 38812
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.970905042365217e-07,
+      "loss": 0.5804,
+      "step": 38813
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.968290450979532e-07,
+      "loss": 0.6086,
+      "step": 38814
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.965676332315185e-07,
+      "loss": 0.5593,
+      "step": 38815
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.963062686385491e-07,
+      "loss": 0.546,
+      "step": 38816
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.960449513203716e-07,
+      "loss": 0.6159,
+      "step": 38817
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.957836812783158e-07,
+      "loss": 0.5494,
+      "step": 38818
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.955224585137055e-07,
+      "loss": 0.6469,
+      "step": 38819
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.952612830278749e-07,
+      "loss": 0.5698,
+      "step": 38820
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.95000154822143e-07,
+      "loss": 0.6214,
+      "step": 38821
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.947390738978421e-07,
+      "loss": 0.6042,
+      "step": 38822
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.944780402562956e-07,
+      "loss": 0.5728,
+      "step": 38823
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.942170538988324e-07,
+      "loss": 0.6545,
+      "step": 38824
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.939561148267782e-07,
+      "loss": 0.5776,
+      "step": 38825
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.936952230414584e-07,
+      "loss": 0.5133,
+      "step": 38826
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.934343785441966e-07,
+      "loss": 0.5871,
+      "step": 38827
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.931735813363228e-07,
+      "loss": 0.5481,
+      "step": 38828
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.929128314191592e-07,
+      "loss": 0.5891,
+      "step": 38829
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.926521287940291e-07,
+      "loss": 0.6317,
+      "step": 38830
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.923914734622627e-07,
+      "loss": 0.6257,
+      "step": 38831
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.92130865425179e-07,
+      "loss": 0.6168,
+      "step": 38832
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.918703046841046e-07,
+      "loss": 0.611,
+      "step": 38833
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.916097912403619e-07,
+      "loss": 0.6341,
+      "step": 38834
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.913493250952774e-07,
+      "loss": 0.5997,
+      "step": 38835
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.910889062501735e-07,
+      "loss": 0.5672,
+      "step": 38836
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.908285347063737e-07,
+      "loss": 0.5869,
+      "step": 38837
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.905682104651979e-07,
+      "loss": 0.6547,
+      "step": 38838
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.903079335279739e-07,
+      "loss": 0.6425,
+      "step": 38839
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.900477038960184e-07,
+      "loss": 0.6058,
+      "step": 38840
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.897875215706618e-07,
+      "loss": 0.6237,
+      "step": 38841
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.895273865532181e-07,
+      "loss": 0.59,
+      "step": 38842
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.892672988450133e-07,
+      "loss": 0.6017,
+      "step": 38843
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.890072584473673e-07,
+      "loss": 0.5987,
+      "step": 38844
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.887472653616034e-07,
+      "loss": 0.5812,
+      "step": 38845
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.884873195890418e-07,
+      "loss": 0.696,
+      "step": 38846
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.882274211310036e-07,
+      "loss": 0.5623,
+      "step": 38847
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.879675699888067e-07,
+      "loss": 0.6212,
+      "step": 38848
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.877077661637754e-07,
+      "loss": 0.5837,
+      "step": 38849
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.874480096572277e-07,
+      "loss": 0.6026,
+      "step": 38850
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.871883004704838e-07,
+      "loss": 0.6439,
+      "step": 38851
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.869286386048657e-07,
+      "loss": 0.6117,
+      "step": 38852
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.866690240616869e-07,
+      "loss": 0.6087,
+      "step": 38853
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.86409456842273e-07,
+      "loss": 0.5895,
+      "step": 38854
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.861499369479386e-07,
+      "loss": 0.5826,
+      "step": 38855
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.858904643800046e-07,
+      "loss": 0.6059,
+      "step": 38856
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.856310391397881e-07,
+      "loss": 0.5688,
+      "step": 38857
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.853716612286121e-07,
+      "loss": 0.6328,
+      "step": 38858
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.851123306477858e-07,
+      "loss": 0.6365,
+      "step": 38859
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.848530473986348e-07,
+      "loss": 0.5859,
+      "step": 38860
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.845938114824713e-07,
+      "loss": 0.6152,
+      "step": 38861
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.843346229006164e-07,
+      "loss": 0.6406,
+      "step": 38862
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.840754816543849e-07,
+      "loss": 0.5991,
+      "step": 38863
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.838163877450943e-07,
+      "loss": 0.5793,
+      "step": 38864
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.835573411740604e-07,
+      "loss": 0.5995,
+      "step": 38865
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 6.832983419426009e-07,
+      "loss": 0.6334,
+      "step": 38866
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.830393900520316e-07,
+      "loss": 0.5928,
+      "step": 38867
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.827804855036668e-07,
+      "loss": 0.5716,
+      "step": 38868
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.825216282988222e-07,
+      "loss": 0.6058,
+      "step": 38869
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.822628184388158e-07,
+      "loss": 0.635,
+      "step": 38870
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.820040559249608e-07,
+      "loss": 0.5896,
+      "step": 38871
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.817453407585695e-07,
+      "loss": 0.5656,
+      "step": 38872
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.814866729409619e-07,
+      "loss": 0.6275,
+      "step": 38873
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.812280524734483e-07,
+      "loss": 0.6398,
+      "step": 38874
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.809694793573452e-07,
+      "loss": 0.6498,
+      "step": 38875
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.807109535939638e-07,
+      "loss": 0.6006,
+      "step": 38876
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.804524751846198e-07,
+      "loss": 0.5707,
+      "step": 38877
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.801940441306254e-07,
+      "loss": 0.631,
+      "step": 38878
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.799356604332974e-07,
+      "loss": 0.6421,
+      "step": 38879
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.796773240939425e-07,
+      "loss": 0.6502,
+      "step": 38880
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.794190351138785e-07,
+      "loss": 0.598,
+      "step": 38881
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.791607934944145e-07,
+      "loss": 0.6104,
+      "step": 38882
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.789025992368659e-07,
+      "loss": 0.5643,
+      "step": 38883
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.786444523425428e-07,
+      "loss": 0.6439,
+      "step": 38884
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.783863528127576e-07,
+      "loss": 0.5773,
+      "step": 38885
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.781283006488193e-07,
+      "loss": 0.5196,
+      "step": 38886
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.778702958520433e-07,
+      "loss": 0.6034,
+      "step": 38887
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.776123384237376e-07,
+      "loss": 0.5395,
+      "step": 38888
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.773544283652123e-07,
+      "loss": 0.5917,
+      "step": 38889
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.770965656777817e-07,
+      "loss": 0.6353,
+      "step": 38890
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.768387503627538e-07,
+      "loss": 0.5705,
+      "step": 38891
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.765809824214376e-07,
+      "loss": 0.6353,
+      "step": 38892
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.76323261855143e-07,
+      "loss": 0.6312,
+      "step": 38893
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.760655886651824e-07,
+      "loss": 0.568,
+      "step": 38894
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.758079628528625e-07,
+      "loss": 0.6089,
+      "step": 38895
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.755503844194932e-07,
+      "loss": 0.6313,
+      "step": 38896
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.752928533663816e-07,
+      "loss": 0.6566,
+      "step": 38897
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.750353696948397e-07,
+      "loss": 0.6292,
+      "step": 38898
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.747779334061722e-07,
+      "loss": 0.6091,
+      "step": 38899
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.745205445016911e-07,
+      "loss": 0.6242,
+      "step": 38900
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.742632029827012e-07,
+      "loss": 0.5823,
+      "step": 38901
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.740059088505124e-07,
+      "loss": 0.6008,
+      "step": 38902
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.737486621064282e-07,
+      "loss": 0.5918,
+      "step": 38903
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.734914627517608e-07,
+      "loss": 0.5377,
+      "step": 38904
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.732343107878136e-07,
+      "loss": 0.6411,
+      "step": 38905
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.729772062158946e-07,
+      "loss": 0.5883,
+      "step": 38906
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.727201490373092e-07,
+      "loss": 0.618,
+      "step": 38907
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.724631392533654e-07,
+      "loss": 0.5528,
+      "step": 38908
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.722061768653687e-07,
+      "loss": 0.6188,
+      "step": 38909
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.719492618746215e-07,
+      "loss": 0.5818,
+      "step": 38910
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.716923942824338e-07,
+      "loss": 0.587,
+      "step": 38911
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.714355740901101e-07,
+      "loss": 0.5704,
+      "step": 38912
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.711788012989529e-07,
+      "loss": 0.6043,
+      "step": 38913
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.709220759102675e-07,
+      "loss": 0.5591,
+      "step": 38914
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.706653979253597e-07,
+      "loss": 0.5964,
+      "step": 38915
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.704087673455328e-07,
+      "loss": 0.63,
+      "step": 38916
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.701521841720937e-07,
+      "loss": 0.5631,
+      "step": 38917
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.698956484063412e-07,
+      "loss": 0.6156,
+      "step": 38918
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.696391600495833e-07,
+      "loss": 0.5932,
+      "step": 38919
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.693827191031188e-07,
+      "loss": 0.618,
+      "step": 38920
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.691263255682546e-07,
+      "loss": 0.5912,
+      "step": 38921
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.688699794462928e-07,
+      "loss": 0.5469,
+      "step": 38922
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.686136807385358e-07,
+      "loss": 0.5982,
+      "step": 38923
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.683574294462835e-07,
+      "loss": 0.5282,
+      "step": 38924
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.681012255708419e-07,
+      "loss": 0.6533,
+      "step": 38925
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.678450691135108e-07,
+      "loss": 0.5657,
+      "step": 38926
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.675889600755892e-07,
+      "loss": 0.6309,
+      "step": 38927
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.673328984583838e-07,
+      "loss": 0.555,
+      "step": 38928
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.670768842631936e-07,
+      "loss": 0.6377,
+      "step": 38929
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.668209174913187e-07,
+      "loss": 0.6475,
+      "step": 38930
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.66564998144058e-07,
+      "loss": 0.5661,
+      "step": 38931
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.663091262227161e-07,
+      "loss": 0.5944,
+      "step": 38932
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.660533017285908e-07,
+      "loss": 0.5793,
+      "step": 38933
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.657975246629822e-07,
+      "loss": 0.6279,
+      "step": 38934
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.655417950271891e-07,
+      "loss": 0.6293,
+      "step": 38935
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.65286112822513e-07,
+      "loss": 0.5643,
+      "step": 38936
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.650304780502492e-07,
+      "loss": 0.612,
+      "step": 38937
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.647748907117024e-07,
+      "loss": 0.6021,
+      "step": 38938
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.645193508081682e-07,
+      "loss": 0.5901,
+      "step": 38939
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.642638583409444e-07,
+      "loss": 0.6665,
+      "step": 38940
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.640084133113301e-07,
+      "loss": 0.6344,
+      "step": 38941
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.637530157206229e-07,
+      "loss": 0.5655,
+      "step": 38942
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.63497665570122e-07,
+      "loss": 0.5806,
+      "step": 38943
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.63242362861124e-07,
+      "loss": 0.5847,
+      "step": 38944
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.629871075949235e-07,
+      "loss": 0.6598,
+      "step": 38945
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.627318997728215e-07,
+      "loss": 0.622,
+      "step": 38946
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.62476739396114e-07,
+      "loss": 0.6093,
+      "step": 38947
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.622216264660952e-07,
+      "loss": 0.5816,
+      "step": 38948
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.619665609840642e-07,
+      "loss": 0.6227,
+      "step": 38949
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.617115429513154e-07,
+      "loss": 0.6389,
+      "step": 38950
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.614565723691457e-07,
+      "loss": 0.6217,
+      "step": 38951
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.612016492388484e-07,
+      "loss": 0.6586,
+      "step": 38952
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.609467735617225e-07,
+      "loss": 0.5969,
+      "step": 38953
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.606919453390592e-07,
+      "loss": 0.6281,
+      "step": 38954
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.604371645721586e-07,
+      "loss": 0.6396,
+      "step": 38955
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.601824312623084e-07,
+      "loss": 0.5297,
+      "step": 38956
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.599277454108088e-07,
+      "loss": 0.5173,
+      "step": 38957
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.596731070189488e-07,
+      "loss": 0.6169,
+      "step": 38958
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.594185160880285e-07,
+      "loss": 0.6496,
+      "step": 38959
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.591639726193367e-07,
+      "loss": 0.6236,
+      "step": 38960
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.58909476614169e-07,
+      "loss": 0.5966,
+      "step": 38961
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.586550280738169e-07,
+      "loss": 0.5999,
+      "step": 38962
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.584006269995746e-07,
+      "loss": 0.6427,
+      "step": 38963
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.581462733927358e-07,
+      "loss": 0.5766,
+      "step": 38964
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.578919672545892e-07,
+      "loss": 0.5769,
+      "step": 38965
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.576377085864316e-07,
+      "loss": 0.583,
+      "step": 38966
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.573834973895532e-07,
+      "loss": 0.6924,
+      "step": 38967
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.571293336652451e-07,
+      "loss": 0.5966,
+      "step": 38968
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.568752174147963e-07,
+      "loss": 0.6018,
+      "step": 38969
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.566211486395036e-07,
+      "loss": 0.6184,
+      "step": 38970
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.563671273406547e-07,
+      "loss": 0.5853,
+      "step": 38971
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.56113153519542e-07,
+      "loss": 0.5249,
+      "step": 38972
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.558592271774522e-07,
+      "loss": 0.5279,
+      "step": 38973
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.556053483156799e-07,
+      "loss": 0.614,
+      "step": 38974
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.553515169355118e-07,
+      "loss": 0.5948,
+      "step": 38975
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.550977330382402e-07,
+      "loss": 0.5624,
+      "step": 38976
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.548439966251552e-07,
+      "loss": 0.5566,
+      "step": 38977
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.545903076975435e-07,
+      "loss": 0.6135,
+      "step": 38978
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.543366662566952e-07,
+      "loss": 0.5913,
+      "step": 38979
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.540830723038993e-07,
+      "loss": 0.5531,
+      "step": 38980
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.538295258404437e-07,
+      "loss": 0.5753,
+      "step": 38981
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.535760268676206e-07,
+      "loss": 0.6652,
+      "step": 38982
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.533225753867123e-07,
+      "loss": 0.6263,
+      "step": 38983
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.5306917139901e-07,
+      "loss": 0.5772,
+      "step": 38984
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.528158149058006e-07,
+      "loss": 0.5929,
+      "step": 38985
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.525625059083707e-07,
+      "loss": 0.6436,
+      "step": 38986
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.523092444080093e-07,
+      "loss": 0.5879,
+      "step": 38987
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.52056030406002e-07,
+      "loss": 0.6549,
+      "step": 38988
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.518028639036356e-07,
+      "loss": 0.5643,
+      "step": 38989
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.515497449021957e-07,
+      "loss": 0.5928,
+      "step": 38990
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.512966734029702e-07,
+      "loss": 0.606,
+      "step": 38991
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.510436494072425e-07,
+      "loss": 0.6661,
+      "step": 38992
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.507906729163039e-07,
+      "loss": 0.6089,
+      "step": 38993
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.505377439314331e-07,
+      "loss": 0.6357,
+      "step": 38994
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.502848624539182e-07,
+      "loss": 0.587,
+      "step": 38995
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.500320284850436e-07,
+      "loss": 0.5577,
+      "step": 38996
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.497792420260962e-07,
+      "loss": 0.6274,
+      "step": 38997
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.495265030783582e-07,
+      "loss": 0.6232,
+      "step": 38998
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.492738116431152e-07,
+      "loss": 0.6236,
+      "step": 38999
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.490211677216484e-07,
+      "loss": 0.6249,
+      "step": 39000
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.487685713152448e-07,
+      "loss": 0.6092,
+      "step": 39001
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.485160224251863e-07,
+      "loss": 0.5877,
+      "step": 39002
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.482635210527577e-07,
+      "loss": 0.5203,
+      "step": 39003
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.480110671992401e-07,
+      "loss": 0.6062,
+      "step": 39004
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.477586608659181e-07,
+      "loss": 0.5809,
+      "step": 39005
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.475063020540728e-07,
+      "loss": 0.5812,
+      "step": 39006
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.472539907649855e-07,
+      "loss": 0.6123,
+      "step": 39007
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.470017269999418e-07,
+      "loss": 0.6108,
+      "step": 39008
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.467495107602207e-07,
+      "loss": 0.607,
+      "step": 39009
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.464973420471043e-07,
+      "loss": 0.5335,
+      "step": 39010
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.462452208618741e-07,
+      "loss": 0.6658,
+      "step": 39011
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.459931472058112e-07,
+      "loss": 0.5782,
+      "step": 39012
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 6.457411210801956e-07,
+      "loss": 0.6105,
+      "step": 39013
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.454891424863107e-07,
+      "loss": 0.521,
+      "step": 39014
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.452372114254357e-07,
+      "loss": 0.6058,
+      "step": 39015
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.449853278988493e-07,
+      "loss": 0.588,
+      "step": 39016
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.447334919078307e-07,
+      "loss": 0.596,
+      "step": 39017
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.444817034536621e-07,
+      "loss": 0.5798,
+      "step": 39018
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.442299625376214e-07,
+      "loss": 0.5564,
+      "step": 39019
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.439782691609908e-07,
+      "loss": 0.571,
+      "step": 39020
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.437266233250439e-07,
+      "loss": 0.6072,
+      "step": 39021
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.434750250310628e-07,
+      "loss": 0.5824,
+      "step": 39022
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.432234742803256e-07,
+      "loss": 0.5997,
+      "step": 39023
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.429719710741089e-07,
+      "loss": 0.5747,
+      "step": 39024
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.42720515413694e-07,
+      "loss": 0.5821,
+      "step": 39025
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.424691073003553e-07,
+      "loss": 0.5497,
+      "step": 39026
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.42217746735373e-07,
+      "loss": 0.5789,
+      "step": 39027
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.419664337200204e-07,
+      "loss": 0.6211,
+      "step": 39028
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.417151682555789e-07,
+      "loss": 0.6486,
+      "step": 39029
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.414639503433207e-07,
+      "loss": 0.5688,
+      "step": 39030
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.412127799845291e-07,
+      "loss": 0.5793,
+      "step": 39031
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.409616571804733e-07,
+      "loss": 0.5685,
+      "step": 39032
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.407105819324333e-07,
+      "loss": 0.5926,
+      "step": 39033
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.404595542416825e-07,
+      "loss": 0.5532,
+      "step": 39034
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.402085741094988e-07,
+      "loss": 0.6274,
+      "step": 39035
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.399576415371578e-07,
+      "loss": 0.6679,
+      "step": 39036
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.397067565259318e-07,
+      "loss": 0.5879,
+      "step": 39037
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.394559190770966e-07,
+      "loss": 0.5844,
+      "step": 39038
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.392051291919276e-07,
+      "loss": 0.6216,
+      "step": 39039
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.389543868716985e-07,
+      "loss": 0.5566,
+      "step": 39040
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.387036921176837e-07,
+      "loss": 0.6381,
+      "step": 39041
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.384530449311588e-07,
+      "loss": 0.6246,
+      "step": 39042
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.382024453133939e-07,
+      "loss": 0.6112,
+      "step": 39043
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.379518932656659e-07,
+      "loss": 0.599,
+      "step": 39044
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.377013887892436e-07,
+      "loss": 0.6537,
+      "step": 39045
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.374509318854039e-07,
+      "loss": 0.5766,
+      "step": 39046
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.372005225554178e-07,
+      "loss": 0.6247,
+      "step": 39047
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.369501608005579e-07,
+      "loss": 0.604,
+      "step": 39048
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.366998466220952e-07,
+      "loss": 0.5951,
+      "step": 39049
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.364495800213044e-07,
+      "loss": 0.629,
+      "step": 39050
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.361993609994532e-07,
+      "loss": 0.6275,
+      "step": 39051
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.359491895578196e-07,
+      "loss": 0.5932,
+      "step": 39052
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.35699065697667e-07,
+      "loss": 0.5924,
+      "step": 39053
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.354489894202709e-07,
+      "loss": 0.5691,
+      "step": 39054
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.351989607268993e-07,
+      "loss": 0.6053,
+      "step": 39055
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.349489796188269e-07,
+      "loss": 0.6098,
+      "step": 39056
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.346990460973191e-07,
+      "loss": 0.5975,
+      "step": 39057
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.344491601636516e-07,
+      "loss": 0.5838,
+      "step": 39058
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.341993218190878e-07,
+      "loss": 0.5607,
+      "step": 39059
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.339495310649024e-07,
+      "loss": 0.6087,
+      "step": 39060
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.336997879023599e-07,
+      "loss": 0.5763,
+      "step": 39061
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.334500923327336e-07,
+      "loss": 0.5934,
+      "step": 39062
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.332004443572903e-07,
+      "loss": 0.5806,
+      "step": 39063
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.329508439772991e-07,
+      "loss": 0.6174,
+      "step": 39064
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.327012911940267e-07,
+      "loss": 0.5965,
+      "step": 39065
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.32451786008742e-07,
+      "loss": 0.6179,
+      "step": 39066
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.322023284227152e-07,
+      "loss": 0.6338,
+      "step": 39067
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.319529184372098e-07,
+      "loss": 0.5994,
+      "step": 39068
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.317035560534979e-07,
+      "loss": 0.5716,
+      "step": 39069
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.314542412728408e-07,
+      "loss": 0.6249,
+      "step": 39070
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.312049740965098e-07,
+      "loss": 0.5321,
+      "step": 39071
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.309557545257694e-07,
+      "loss": 0.6175,
+      "step": 39072
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.307065825618875e-07,
+      "loss": 0.6499,
+      "step": 39073
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.304574582061296e-07,
+      "loss": 0.6148,
+      "step": 39074
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.302083814597615e-07,
+      "loss": 0.5596,
+      "step": 39075
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.299593523240477e-07,
+      "loss": 0.5777,
+      "step": 39076
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.29710370800255e-07,
+      "loss": 0.6104,
+      "step": 39077
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.294614368896468e-07,
+      "loss": 0.6101,
+      "step": 39078
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.292125505934932e-07,
+      "loss": 0.5436,
+      "step": 39079
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.289637119130509e-07,
+      "loss": 0.5645,
+      "step": 39080
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.287149208495902e-07,
+      "loss": 0.6017,
+      "step": 39081
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.284661774043721e-07,
+      "loss": 0.6177,
+      "step": 39082
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.282174815786634e-07,
+      "loss": 0.654,
+      "step": 39083
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.279688333737266e-07,
+      "loss": 0.6089,
+      "step": 39084
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.27720232790825e-07,
+      "loss": 0.6052,
+      "step": 39085
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.274716798312219e-07,
+      "loss": 0.6008,
+      "step": 39086
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.272231744961776e-07,
+      "loss": 0.6236,
+      "step": 39087
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.269747167869589e-07,
+      "loss": 0.6339,
+      "step": 39088
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.267263067048257e-07,
+      "loss": 0.5954,
+      "step": 39089
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.264779442510449e-07,
+      "loss": 0.6165,
+      "step": 39090
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.262296294268699e-07,
+      "loss": 0.5656,
+      "step": 39091
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.259813622335709e-07,
+      "loss": 0.5928,
+      "step": 39092
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.257331426724034e-07,
+      "loss": 0.593,
+      "step": 39093
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.254849707446331e-07,
+      "loss": 0.5718,
+      "step": 39094
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.252368464515168e-07,
+      "loss": 0.5538,
+      "step": 39095
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.249887697943213e-07,
+      "loss": 0.6663,
+      "step": 39096
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.247407407743e-07,
+      "loss": 0.659,
+      "step": 39097
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.244927593927186e-07,
+      "loss": 0.5612,
+      "step": 39098
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.242448256508338e-07,
+      "loss": 0.6166,
+      "step": 39099
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.23996939549908e-07,
+      "loss": 0.5915,
+      "step": 39100
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.237491010912e-07,
+      "loss": 0.6033,
+      "step": 39101
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.235013102759679e-07,
+      "loss": 0.6044,
+      "step": 39102
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.232535671054729e-07,
+      "loss": 0.6159,
+      "step": 39103
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.230058715809706e-07,
+      "loss": 0.601,
+      "step": 39104
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.227582237037233e-07,
+      "loss": 0.5806,
+      "step": 39105
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.225106234749855e-07,
+      "loss": 0.5737,
+      "step": 39106
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.222630708960209e-07,
+      "loss": 0.5984,
+      "step": 39107
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.220155659680815e-07,
+      "loss": 0.5711,
+      "step": 39108
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.217681086924287e-07,
+      "loss": 0.5772,
+      "step": 39109
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.215206990703171e-07,
+      "loss": 0.6585,
+      "step": 39110
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.212733371030077e-07,
+      "loss": 0.5476,
+      "step": 39111
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.210260227917542e-07,
+      "loss": 0.5984,
+      "step": 39112
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.207787561378142e-07,
+      "loss": 0.5777,
+      "step": 39113
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.205315371424436e-07,
+      "loss": 0.5987,
+      "step": 39114
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.202843658069003e-07,
+      "loss": 0.6481,
+      "step": 39115
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.200372421324363e-07,
+      "loss": 0.6315,
+      "step": 39116
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.197901661203142e-07,
+      "loss": 0.5853,
+      "step": 39117
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.195431377717831e-07,
+      "loss": 0.5633,
+      "step": 39118
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.192961570881006e-07,
+      "loss": 0.5975,
+      "step": 39119
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.190492240705215e-07,
+      "loss": 0.5644,
+      "step": 39120
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.188023387203013e-07,
+      "loss": 0.583,
+      "step": 39121
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.185555010386935e-07,
+      "loss": 0.5905,
+      "step": 39122
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.183087110269537e-07,
+      "loss": 0.6213,
+      "step": 39123
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.180619686863342e-07,
+      "loss": 0.6,
+      "step": 39124
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.178152740180876e-07,
+      "loss": 0.6461,
+      "step": 39125
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.175686270234715e-07,
+      "loss": 0.5568,
+      "step": 39126
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.173220277037351e-07,
+      "loss": 0.5754,
+      "step": 39127
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.170754760601372e-07,
+      "loss": 0.5409,
+      "step": 39128
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.168289720939224e-07,
+      "loss": 0.6594,
+      "step": 39129
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.165825158063499e-07,
+      "loss": 0.5585,
+      "step": 39130
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.163361071986684e-07,
+      "loss": 0.6063,
+      "step": 39131
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.160897462721327e-07,
+      "loss": 0.647,
+      "step": 39132
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.158434330279905e-07,
+      "loss": 0.6166,
+      "step": 39133
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.155971674674999e-07,
+      "loss": 0.5661,
+      "step": 39134
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.153509495919063e-07,
+      "loss": 0.6492,
+      "step": 39135
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.151047794024634e-07,
+      "loss": 0.6939,
+      "step": 39136
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.1485865690042e-07,
+      "loss": 0.5721,
+      "step": 39137
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.146125820870307e-07,
+      "loss": 0.5861,
+      "step": 39138
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.143665549635424e-07,
+      "loss": 0.5731,
+      "step": 39139
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.141205755312074e-07,
+      "loss": 0.6413,
+      "step": 39140
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.138746437912723e-07,
+      "loss": 0.552,
+      "step": 39141
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.136287597449909e-07,
+      "loss": 0.5719,
+      "step": 39142
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.133829233936106e-07,
+      "loss": 0.6249,
+      "step": 39143
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.131371347383808e-07,
+      "loss": 0.6166,
+      "step": 39144
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.128913937805503e-07,
+      "loss": 0.6253,
+      "step": 39145
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.12645700521366e-07,
+      "loss": 0.609,
+      "step": 39146
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.124000549620801e-07,
+      "loss": 0.6244,
+      "step": 39147
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.121544571039361e-07,
+      "loss": 0.5875,
+      "step": 39148
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.119089069481876e-07,
+      "loss": 0.6077,
+      "step": 39149
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.11663404496079e-07,
+      "loss": 0.5906,
+      "step": 39150
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.11417949748857e-07,
+      "loss": 0.5508,
+      "step": 39151
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.111725427077697e-07,
+      "loss": 0.6116,
+      "step": 39152
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.10927183374066e-07,
+      "loss": 0.5987,
+      "step": 39153
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.106818717489882e-07,
+      "loss": 0.5654,
+      "step": 39154
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.104366078337897e-07,
+      "loss": 0.6424,
+      "step": 39155
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.101913916297097e-07,
+      "loss": 0.6231,
+      "step": 39156
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.099462231379982e-07,
+      "loss": 0.5787,
+      "step": 39157
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.097011023598986e-07,
+      "loss": 0.5865,
+      "step": 39158
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.094560292966589e-07,
+      "loss": 0.5838,
+      "step": 39159
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 6.092110039495236e-07,
+      "loss": 0.6348,
+      "step": 39160
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.089660263197383e-07,
+      "loss": 0.5801,
+      "step": 39161
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.087210964085444e-07,
+      "loss": 0.5358,
+      "step": 39162
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.084762142171907e-07,
+      "loss": 0.5643,
+      "step": 39163
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.082313797469198e-07,
+      "loss": 0.5891,
+      "step": 39164
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.079865929989736e-07,
+      "loss": 0.5709,
+      "step": 39165
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.077418539746016e-07,
+      "loss": 0.6054,
+      "step": 39166
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.074971626750414e-07,
+      "loss": 0.6092,
+      "step": 39167
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.072525191015399e-07,
+      "loss": 0.6474,
+      "step": 39168
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.070079232553371e-07,
+      "loss": 0.6338,
+      "step": 39169
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.0676337513768e-07,
+      "loss": 0.6006,
+      "step": 39170
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.065188747498097e-07,
+      "loss": 0.618,
+      "step": 39171
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.062744220929673e-07,
+      "loss": 0.6416,
+      "step": 39172
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.060300171683941e-07,
+      "loss": 0.5091,
+      "step": 39173
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.057856599773348e-07,
+      "loss": 0.6154,
+      "step": 39174
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.055413505210295e-07,
+      "loss": 0.5821,
+      "step": 39175
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.052970888007203e-07,
+      "loss": 0.5907,
+      "step": 39176
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.050528748176477e-07,
+      "loss": 0.6142,
+      "step": 39177
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.048087085730536e-07,
+      "loss": 0.6026,
+      "step": 39178
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.045645900681763e-07,
+      "loss": 0.6042,
+      "step": 39179
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.04320519304259e-07,
+      "loss": 0.5826,
+      "step": 39180
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.040764962825419e-07,
+      "loss": 0.625,
+      "step": 39181
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.038325210042629e-07,
+      "loss": 0.6194,
+      "step": 39182
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.035885934706609e-07,
+      "loss": 0.6608,
+      "step": 39183
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.033447136829795e-07,
+      "loss": 0.585,
+      "step": 39184
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.031008816424556e-07,
+      "loss": 0.5818,
+      "step": 39185
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.028570973503256e-07,
+      "loss": 0.6345,
+      "step": 39186
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.026133608078332e-07,
+      "loss": 0.5655,
+      "step": 39187
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.023696720162143e-07,
+      "loss": 0.6351,
+      "step": 39188
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.021260309767063e-07,
+      "loss": 0.6255,
+      "step": 39189
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.018824376905475e-07,
+      "loss": 0.5823,
+      "step": 39190
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.016388921589777e-07,
+      "loss": 0.6054,
+      "step": 39191
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.013953943832318e-07,
+      "loss": 0.5705,
+      "step": 39192
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.011519443645519e-07,
+      "loss": 0.5635,
+      "step": 39193
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.009085421041682e-07,
+      "loss": 0.654,
+      "step": 39194
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.00665187603322e-07,
+      "loss": 0.6157,
+      "step": 39195
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.004218808632478e-07,
+      "loss": 0.5815,
+      "step": 39196
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 6.001786218851835e-07,
+      "loss": 0.5823,
+      "step": 39197
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.999354106703659e-07,
+      "loss": 0.5649,
+      "step": 39198
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.996922472200283e-07,
+      "loss": 0.6774,
+      "step": 39199
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.994491315354057e-07,
+      "loss": 0.5834,
+      "step": 39200
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.992060636177377e-07,
+      "loss": 0.6265,
+      "step": 39201
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.989630434682559e-07,
+      "loss": 0.6105,
+      "step": 39202
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.987200710881946e-07,
+      "loss": 0.5801,
+      "step": 39203
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.984771464787931e-07,
+      "loss": 0.6449,
+      "step": 39204
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.982342696412801e-07,
+      "loss": 0.6274,
+      "step": 39205
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.979914405768928e-07,
+      "loss": 0.5664,
+      "step": 39206
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.977486592868642e-07,
+      "loss": 0.5737,
+      "step": 39207
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.975059257724281e-07,
+      "loss": 0.5428,
+      "step": 39208
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.9726324003482e-07,
+      "loss": 0.5846,
+      "step": 39209
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.9702060207527e-07,
+      "loss": 0.6509,
+      "step": 39210
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.967780118950106e-07,
+      "loss": 0.5101,
+      "step": 39211
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.965354694952774e-07,
+      "loss": 0.5956,
+      "step": 39212
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.962929748772994e-07,
+      "loss": 0.5732,
+      "step": 39213
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.960505280423134e-07,
+      "loss": 0.6221,
+      "step": 39214
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.958081289915485e-07,
+      "loss": 0.5751,
+      "step": 39215
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.955657777262358e-07,
+      "loss": 0.6146,
+      "step": 39216
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.953234742476066e-07,
+      "loss": 0.5787,
+      "step": 39217
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.950812185568954e-07,
+      "loss": 0.5934,
+      "step": 39218
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.948390106553304e-07,
+      "loss": 0.5786,
+      "step": 39219
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.945968505441424e-07,
+      "loss": 0.642,
+      "step": 39220
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.943547382245607e-07,
+      "loss": 0.6015,
+      "step": 39221
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.941126736978198e-07,
+      "loss": 0.6103,
+      "step": 39222
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.938706569651453e-07,
+      "loss": 0.5649,
+      "step": 39223
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.936286880277686e-07,
+      "loss": 0.6063,
+      "step": 39224
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.933867668869209e-07,
+      "loss": 0.6191,
+      "step": 39225
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.931448935438289e-07,
+      "loss": 0.6281,
+      "step": 39226
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.929030679997228e-07,
+      "loss": 0.5685,
+      "step": 39227
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.926612902558305e-07,
+      "loss": 0.6027,
+      "step": 39228
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.92419560313382e-07,
+      "loss": 0.6211,
+      "step": 39229
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.921778781736043e-07,
+      "loss": 0.6118,
+      "step": 39230
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.919362438377274e-07,
+      "loss": 0.5535,
+      "step": 39231
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.916946573069759e-07,
+      "loss": 0.614,
+      "step": 39232
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.9145311858258e-07,
+      "loss": 0.6319,
+      "step": 39233
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.912116276657653e-07,
+      "loss": 0.5825,
+      "step": 39234
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.90970184557762e-07,
+      "loss": 0.6234,
+      "step": 39235
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.907287892597935e-07,
+      "loss": 0.5669,
+      "step": 39236
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.904874417730877e-07,
+      "loss": 0.5804,
+      "step": 39237
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.902461420988703e-07,
+      "loss": 0.5927,
+      "step": 39238
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.900048902383693e-07,
+      "loss": 0.5896,
+      "step": 39239
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.897636861928091e-07,
+      "loss": 0.5449,
+      "step": 39240
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.895225299634133e-07,
+      "loss": 0.6327,
+      "step": 39241
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.89281421551412e-07,
+      "loss": 0.6013,
+      "step": 39242
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.890403609580275e-07,
+      "loss": 0.6089,
+      "step": 39243
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.887993481844845e-07,
+      "loss": 0.5745,
+      "step": 39244
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.885583832320074e-07,
+      "loss": 0.6055,
+      "step": 39245
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.883174661018232e-07,
+      "loss": 0.5821,
+      "step": 39246
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.88076596795154e-07,
+      "loss": 0.6058,
+      "step": 39247
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.878357753132235e-07,
+      "loss": 0.576,
+      "step": 39248
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.875950016572552e-07,
+      "loss": 0.591,
+      "step": 39249
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.873542758284745e-07,
+      "loss": 0.5501,
+      "step": 39250
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.871135978281017e-07,
+      "loss": 0.6024,
+      "step": 39251
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.868729676573626e-07,
+      "loss": 0.6002,
+      "step": 39252
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.866323853174794e-07,
+      "loss": 0.5743,
+      "step": 39253
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.863918508096733e-07,
+      "loss": 0.5876,
+      "step": 39254
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.861513641351657e-07,
+      "loss": 0.603,
+      "step": 39255
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.859109252951811e-07,
+      "loss": 0.6307,
+      "step": 39256
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.856705342909408e-07,
+      "loss": 0.5731,
+      "step": 39257
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.854301911236659e-07,
+      "loss": 0.6079,
+      "step": 39258
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.851898957945757e-07,
+      "loss": 0.6229,
+      "step": 39259
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.849496483048933e-07,
+      "loss": 0.5698,
+      "step": 39260
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.847094486558402e-07,
+      "loss": 0.6973,
+      "step": 39261
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.844692968486342e-07,
+      "loss": 0.5758,
+      "step": 39262
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.84229192884499e-07,
+      "loss": 0.5693,
+      "step": 39263
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.839891367646522e-07,
+      "loss": 0.6302,
+      "step": 39264
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.837491284903141e-07,
+      "loss": 0.5403,
+      "step": 39265
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.835091680627025e-07,
+      "loss": 0.581,
+      "step": 39266
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.83269255483041e-07,
+      "loss": 0.5345,
+      "step": 39267
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.830293907525442e-07,
+      "loss": 0.6021,
+      "step": 39268
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.827895738724366e-07,
+      "loss": 0.6346,
+      "step": 39269
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.825498048439293e-07,
+      "loss": 0.6723,
+      "step": 39270
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.823100836682461e-07,
+      "loss": 0.5586,
+      "step": 39271
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.820704103466024e-07,
+      "loss": 0.5818,
+      "step": 39272
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.818307848802185e-07,
+      "loss": 0.5617,
+      "step": 39273
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.815912072703101e-07,
+      "loss": 0.6546,
+      "step": 39274
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.81351677518095e-07,
+      "loss": 0.5838,
+      "step": 39275
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.8111219562479e-07,
+      "loss": 0.6317,
+      "step": 39276
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.808727615916132e-07,
+      "loss": 0.5857,
+      "step": 39277
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.806333754197802e-07,
+      "loss": 0.615,
+      "step": 39278
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.803940371105066e-07,
+      "loss": 0.5792,
+      "step": 39279
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.801547466650104e-07,
+      "loss": 0.6191,
+      "step": 39280
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.799155040845073e-07,
+      "loss": 0.5417,
+      "step": 39281
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.796763093702118e-07,
+      "loss": 0.593,
+      "step": 39282
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.794371625233397e-07,
+      "loss": 0.5848,
+      "step": 39283
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.791980635451066e-07,
+      "loss": 0.5967,
+      "step": 39284
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.789590124367273e-07,
+      "loss": 0.6105,
+      "step": 39285
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.78720009199416e-07,
+      "loss": 0.6291,
+      "step": 39286
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.784810538343866e-07,
+      "loss": 0.6183,
+      "step": 39287
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.782421463428556e-07,
+      "loss": 0.6033,
+      "step": 39288
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.780032867260332e-07,
+      "loss": 0.546,
+      "step": 39289
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.777644749851386e-07,
+      "loss": 0.554,
+      "step": 39290
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.775257111213805e-07,
+      "loss": 0.6002,
+      "step": 39291
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.77286995135975e-07,
+      "loss": 0.5354,
+      "step": 39292
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.770483270301308e-07,
+      "loss": 0.6044,
+      "step": 39293
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.768097068050671e-07,
+      "loss": 0.6118,
+      "step": 39294
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.765711344619918e-07,
+      "loss": 0.613,
+      "step": 39295
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.763326100021182e-07,
+      "loss": 0.5969,
+      "step": 39296
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.760941334266579e-07,
+      "loss": 0.6463,
+      "step": 39297
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.75855704736824e-07,
+      "loss": 0.6091,
+      "step": 39298
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.756173239338269e-07,
+      "loss": 0.5776,
+      "step": 39299
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.753789910188779e-07,
+      "loss": 0.6445,
+      "step": 39300
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.751407059931891e-07,
+      "loss": 0.6247,
+      "step": 39301
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.749024688579707e-07,
+      "loss": 0.6375,
+      "step": 39302
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.74664279614433e-07,
+      "loss": 0.6188,
+      "step": 39303
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.74426138263785e-07,
+      "loss": 0.6014,
+      "step": 39304
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.741880448072401e-07,
+      "loss": 0.5772,
+      "step": 39305
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.73949999246004e-07,
+      "loss": 0.5978,
+      "step": 39306
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 5.737120015812925e-07,
+      "loss": 0.6214,
+      "step": 39307
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.734740518143068e-07,
+      "loss": 0.6146,
+      "step": 39308
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.732361499462625e-07,
+      "loss": 0.6205,
+      "step": 39309
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.729982959783642e-07,
+      "loss": 0.5448,
+      "step": 39310
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.727604899118244e-07,
+      "loss": 0.5997,
+      "step": 39311
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.725227317478488e-07,
+      "loss": 0.5889,
+      "step": 39312
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.722850214876463e-07,
+      "loss": 0.5865,
+      "step": 39313
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.720473591324227e-07,
+      "loss": 0.5751,
+      "step": 39314
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.718097446833904e-07,
+      "loss": 0.5739,
+      "step": 39315
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.715721781417527e-07,
+      "loss": 0.5633,
+      "step": 39316
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.713346595087165e-07,
+      "loss": 0.6289,
+      "step": 39317
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.710971887854922e-07,
+      "loss": 0.5836,
+      "step": 39318
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.708597659732851e-07,
+      "loss": 0.6009,
+      "step": 39319
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.706223910733e-07,
+      "loss": 0.5998,
+      "step": 39320
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.703850640867425e-07,
+      "loss": 0.5702,
+      "step": 39321
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.701477850148218e-07,
+      "loss": 0.5745,
+      "step": 39322
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.699105538587424e-07,
+      "loss": 0.5393,
+      "step": 39323
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.696733706197089e-07,
+      "loss": 0.6116,
+      "step": 39324
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.694362352989246e-07,
+      "loss": 0.566,
+      "step": 39325
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.691991478975989e-07,
+      "loss": 0.6123,
+      "step": 39326
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.689621084169328e-07,
+      "loss": 0.5552,
+      "step": 39327
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.687251168581353e-07,
+      "loss": 0.6113,
+      "step": 39328
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.684881732224034e-07,
+      "loss": 0.6626,
+      "step": 39329
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.682512775109484e-07,
+      "loss": 0.6174,
+      "step": 39330
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.680144297249679e-07,
+      "loss": 0.5982,
+      "step": 39331
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.677776298656712e-07,
+      "loss": 0.6293,
+      "step": 39332
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.675408779342562e-07,
+      "loss": 0.6094,
+      "step": 39333
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.673041739319319e-07,
+      "loss": 0.611,
+      "step": 39334
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.67067517859895e-07,
+      "loss": 0.6087,
+      "step": 39335
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.668309097193514e-07,
+      "loss": 0.5822,
+      "step": 39336
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.665943495115023e-07,
+      "loss": 0.6075,
+      "step": 39337
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.663578372375488e-07,
+      "loss": 0.6333,
+      "step": 39338
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.661213728986958e-07,
+      "loss": 0.6009,
+      "step": 39339
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.658849564961433e-07,
+      "loss": 0.5898,
+      "step": 39340
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.656485880310914e-07,
+      "loss": 0.5778,
+      "step": 39341
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.654122675047414e-07,
+      "loss": 0.588,
+      "step": 39342
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.651759949182956e-07,
+      "loss": 0.5904,
+      "step": 39343
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.64939770272952e-07,
+      "loss": 0.5478,
+      "step": 39344
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.647035935699152e-07,
+      "loss": 0.6249,
+      "step": 39345
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.64467464810381e-07,
+      "loss": 0.6117,
+      "step": 39346
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.642313839955516e-07,
+      "loss": 0.55,
+      "step": 39347
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.639953511266238e-07,
+      "loss": 0.6103,
+      "step": 39348
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.637593662048024e-07,
+      "loss": 0.5904,
+      "step": 39349
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.635234292312819e-07,
+      "loss": 0.5876,
+      "step": 39350
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.632875402072624e-07,
+      "loss": 0.6064,
+      "step": 39351
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.630516991339418e-07,
+      "loss": 0.5799,
+      "step": 39352
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.628159060125204e-07,
+      "loss": 0.5989,
+      "step": 39353
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.625801608441928e-07,
+      "loss": 0.6452,
+      "step": 39354
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.623444636301633e-07,
+      "loss": 0.5524,
+      "step": 39355
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.621088143716225e-07,
+      "loss": 0.5758,
+      "step": 39356
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.618732130697724e-07,
+      "loss": 0.5652,
+      "step": 39357
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.616376597258078e-07,
+      "loss": 0.6038,
+      "step": 39358
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.614021543409254e-07,
+      "loss": 0.5619,
+      "step": 39359
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.611666969163243e-07,
+      "loss": 0.6263,
+      "step": 39360
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.609312874532002e-07,
+      "loss": 0.584,
+      "step": 39361
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.606959259527478e-07,
+      "loss": 0.5855,
+      "step": 39362
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.604606124161627e-07,
+      "loss": 0.568,
+      "step": 39363
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.602253468446428e-07,
+      "loss": 0.6271,
+      "step": 39364
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.599901292393817e-07,
+      "loss": 0.5951,
+      "step": 39365
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.597549596015772e-07,
+      "loss": 0.6517,
+      "step": 39366
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.595198379324207e-07,
+      "loss": 0.6397,
+      "step": 39367
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.5928476423311e-07,
+      "loss": 0.6292,
+      "step": 39368
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.590497385048366e-07,
+      "loss": 0.6039,
+      "step": 39369
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.58814760748797e-07,
+      "loss": 0.556,
+      "step": 39370
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.585798309661838e-07,
+      "loss": 0.6378,
+      "step": 39371
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.58344949158195e-07,
+      "loss": 0.5839,
+      "step": 39372
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.581101153260171e-07,
+      "loss": 0.5943,
+      "step": 39373
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.578753294708483e-07,
+      "loss": 0.6026,
+      "step": 39374
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.576405915938788e-07,
+      "loss": 0.657,
+      "step": 39375
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.57405901696304e-07,
+      "loss": 0.6255,
+      "step": 39376
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.571712597793155e-07,
+      "loss": 0.6029,
+      "step": 39377
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.569366658441055e-07,
+      "loss": 0.6214,
+      "step": 39378
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.567021198918654e-07,
+      "loss": 0.6476,
+      "step": 39379
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.564676219237864e-07,
+      "loss": 0.5766,
+      "step": 39380
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.562331719410629e-07,
+      "loss": 0.6399,
+      "step": 39381
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.559987699448821e-07,
+      "loss": 0.6292,
+      "step": 39382
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.557644159364406e-07,
+      "loss": 0.5723,
+      "step": 39383
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.55530109916923e-07,
+      "loss": 0.6126,
+      "step": 39384
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.55295851887524e-07,
+      "loss": 0.6455,
+      "step": 39385
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.550616418494315e-07,
+      "loss": 0.6156,
+      "step": 39386
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.548274798038389e-07,
+      "loss": 0.6005,
+      "step": 39387
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.545933657519331e-07,
+      "loss": 0.6339,
+      "step": 39388
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.543592996949054e-07,
+      "loss": 0.6553,
+      "step": 39389
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.541252816339427e-07,
+      "loss": 0.6425,
+      "step": 39390
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.538913115702371e-07,
+      "loss": 0.6824,
+      "step": 39391
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.536573895049746e-07,
+      "loss": 0.6445,
+      "step": 39392
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.534235154393486e-07,
+      "loss": 0.6494,
+      "step": 39393
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.531896893745414e-07,
+      "loss": 0.5828,
+      "step": 39394
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.529559113117455e-07,
+      "loss": 0.5695,
+      "step": 39395
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.527221812521466e-07,
+      "loss": 0.6392,
+      "step": 39396
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.524884991969326e-07,
+      "loss": 0.5897,
+      "step": 39397
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.522548651472925e-07,
+      "loss": 0.5652,
+      "step": 39398
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.520212791044121e-07,
+      "loss": 0.5288,
+      "step": 39399
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.517877410694783e-07,
+      "loss": 0.5768,
+      "step": 39400
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.515542510436778e-07,
+      "loss": 0.625,
+      "step": 39401
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.513208090281974e-07,
+      "loss": 0.589,
+      "step": 39402
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.510874150242218e-07,
+      "loss": 0.5713,
+      "step": 39403
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.508540690329422e-07,
+      "loss": 0.5807,
+      "step": 39404
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.506207710555367e-07,
+      "loss": 0.6278,
+      "step": 39405
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.503875210931964e-07,
+      "loss": 0.57,
+      "step": 39406
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.501543191471026e-07,
+      "loss": 0.5799,
+      "step": 39407
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.499211652184433e-07,
+      "loss": 0.5963,
+      "step": 39408
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.496880593084009e-07,
+      "loss": 0.643,
+      "step": 39409
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.494550014181643e-07,
+      "loss": 0.609,
+      "step": 39410
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.492219915489117e-07,
+      "loss": 0.5913,
+      "step": 39411
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.489890297018319e-07,
+      "loss": 0.5771,
+      "step": 39412
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.487561158781041e-07,
+      "loss": 0.5377,
+      "step": 39413
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.485232500789173e-07,
+      "loss": 0.6084,
+      "step": 39414
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.482904323054506e-07,
+      "loss": 0.6172,
+      "step": 39415
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.480576625588885e-07,
+      "loss": 0.5907,
+      "step": 39416
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.478249408404135e-07,
+      "loss": 0.6187,
+      "step": 39417
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.475922671512069e-07,
+      "loss": 0.6121,
+      "step": 39418
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.473596414924542e-07,
+      "loss": 0.6269,
+      "step": 39419
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.471270638653347e-07,
+      "loss": 0.6596,
+      "step": 39420
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.468945342710319e-07,
+      "loss": 0.6031,
+      "step": 39421
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.466620527107247e-07,
+      "loss": 0.61,
+      "step": 39422
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.464296191855967e-07,
+      "loss": 0.5991,
+      "step": 39423
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.461972336968279e-07,
+      "loss": 0.6261,
+      "step": 39424
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.45964896245601e-07,
+      "loss": 0.5825,
+      "step": 39425
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.457326068330948e-07,
+      "loss": 0.6137,
+      "step": 39426
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.455003654604895e-07,
+      "loss": 0.584,
+      "step": 39427
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.452681721289655e-07,
+      "loss": 0.6119,
+      "step": 39428
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.450360268397037e-07,
+      "loss": 0.5857,
+      "step": 39429
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.448039295938812e-07,
+      "loss": 0.613,
+      "step": 39430
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.445718803926825e-07,
+      "loss": 0.6186,
+      "step": 39431
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.4433987923728e-07,
+      "loss": 0.5884,
+      "step": 39432
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.441079261288573e-07,
+      "loss": 0.5459,
+      "step": 39433
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.4387602106859e-07,
+      "loss": 0.608,
+      "step": 39434
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.436441640576595e-07,
+      "loss": 0.591,
+      "step": 39435
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.434123550972425e-07,
+      "loss": 0.5742,
+      "step": 39436
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.431805941885171e-07,
+      "loss": 0.666,
+      "step": 39437
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.4294888133266e-07,
+      "loss": 0.5495,
+      "step": 39438
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.42717216530848e-07,
+      "loss": 0.6982,
+      "step": 39439
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.424855997842615e-07,
+      "loss": 0.6278,
+      "step": 39440
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.422540310940738e-07,
+      "loss": 0.5335,
+      "step": 39441
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.420225104614651e-07,
+      "loss": 0.572,
+      "step": 39442
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.417910378876079e-07,
+      "loss": 0.6144,
+      "step": 39443
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.415596133736822e-07,
+      "loss": 0.5713,
+      "step": 39444
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.413282369208606e-07,
+      "loss": 0.659,
+      "step": 39445
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.410969085303209e-07,
+      "loss": 0.5842,
+      "step": 39446
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.408656282032387e-07,
+      "loss": 0.59,
+      "step": 39447
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.406343959407878e-07,
+      "loss": 0.651,
+      "step": 39448
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.404032117441426e-07,
+      "loss": 0.5793,
+      "step": 39449
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.401720756144813e-07,
+      "loss": 0.6263,
+      "step": 39450
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.399409875529737e-07,
+      "loss": 0.6091,
+      "step": 39451
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.397099475607992e-07,
+      "loss": 0.6306,
+      "step": 39452
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.394789556391277e-07,
+      "loss": 0.6323,
+      "step": 39453
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 5.392480117891341e-07,
+      "loss": 0.5692,
+      "step": 39454
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.390171160119916e-07,
+      "loss": 0.5483,
+      "step": 39455
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.387862683088751e-07,
+      "loss": 0.6067,
+      "step": 39456
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.385554686809568e-07,
+      "loss": 0.5629,
+      "step": 39457
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.383247171294082e-07,
+      "loss": 0.6054,
+      "step": 39458
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.380940136554025e-07,
+      "loss": 0.6331,
+      "step": 39459
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.378633582601123e-07,
+      "loss": 0.5915,
+      "step": 39460
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.376327509447099e-07,
+      "loss": 0.5978,
+      "step": 39461
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.374021917103644e-07,
+      "loss": 0.5814,
+      "step": 39462
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.371716805582527e-07,
+      "loss": 0.5795,
+      "step": 39463
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.369412174895416e-07,
+      "loss": 0.6054,
+      "step": 39464
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.367108025054046e-07,
+      "loss": 0.6579,
+      "step": 39465
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.364804356070097e-07,
+      "loss": 0.6353,
+      "step": 39466
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.362501167955303e-07,
+      "loss": 0.609,
+      "step": 39467
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.360198460721344e-07,
+      "loss": 0.5589,
+      "step": 39468
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.357896234379967e-07,
+      "loss": 0.5411,
+      "step": 39469
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.355594488942806e-07,
+      "loss": 0.6013,
+      "step": 39470
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.353293224421596e-07,
+      "loss": 0.5973,
+      "step": 39471
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.350992440828006e-07,
+      "loss": 0.6275,
+      "step": 39472
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.348692138173761e-07,
+      "loss": 0.6266,
+      "step": 39473
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.346392316470539e-07,
+      "loss": 0.5873,
+      "step": 39474
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.344092975730008e-07,
+      "loss": 0.6471,
+      "step": 39475
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.341794115963861e-07,
+      "loss": 0.6125,
+      "step": 39476
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.339495737183776e-07,
+      "loss": 0.6163,
+      "step": 39477
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.337197839401443e-07,
+      "loss": 0.6249,
+      "step": 39478
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.334900422628509e-07,
+      "loss": 0.6172,
+      "step": 39479
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.332603486876709e-07,
+      "loss": 0.6215,
+      "step": 39480
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.330307032157644e-07,
+      "loss": 0.5896,
+      "step": 39481
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.32801105848303e-07,
+      "loss": 0.6144,
+      "step": 39482
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.325715565864509e-07,
+      "loss": 0.6185,
+      "step": 39483
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.323420554313752e-07,
+      "loss": 0.5725,
+      "step": 39484
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.321126023842438e-07,
+      "loss": 0.5953,
+      "step": 39485
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.318831974462202e-07,
+      "loss": 0.617,
+      "step": 39486
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.316538406184702e-07,
+      "loss": 0.6168,
+      "step": 39487
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.314245319021604e-07,
+      "loss": 0.6443,
+      "step": 39488
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.311952712984547e-07,
+      "loss": 0.5804,
+      "step": 39489
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.309660588085208e-07,
+      "loss": 0.5653,
+      "step": 39490
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.307368944335212e-07,
+      "loss": 0.6018,
+      "step": 39491
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.305077781746204e-07,
+      "loss": 0.552,
+      "step": 39492
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.302787100329809e-07,
+      "loss": 0.5622,
+      "step": 39493
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.300496900097707e-07,
+      "loss": 0.5787,
+      "step": 39494
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.29820718106152e-07,
+      "loss": 0.6049,
+      "step": 39495
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.295917943232865e-07,
+      "loss": 0.6297,
+      "step": 39496
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.293629186623394e-07,
+      "loss": 0.6047,
+      "step": 39497
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.291340911244713e-07,
+      "loss": 0.6185,
+      "step": 39498
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.289053117108478e-07,
+      "loss": 0.5913,
+      "step": 39499
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.286765804226291e-07,
+      "loss": 0.6271,
+      "step": 39500
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.284478972609796e-07,
+      "loss": 0.6091,
+      "step": 39501
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.28219262227061e-07,
+      "loss": 0.6381,
+      "step": 39502
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.279906753220332e-07,
+      "loss": 0.5863,
+      "step": 39503
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.277621365470575e-07,
+      "loss": 0.6023,
+      "step": 39504
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.275336459032987e-07,
+      "loss": 0.5992,
+      "step": 39505
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.273052033919135e-07,
+      "loss": 0.5708,
+      "step": 39506
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.270768090140677e-07,
+      "loss": 0.6457,
+      "step": 39507
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.268484627709158e-07,
+      "loss": 0.5687,
+      "step": 39508
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.266201646636226e-07,
+      "loss": 0.5857,
+      "step": 39509
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.263919146933461e-07,
+      "loss": 0.6289,
+      "step": 39510
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.261637128612473e-07,
+      "loss": 0.5736,
+      "step": 39511
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.259355591684856e-07,
+      "loss": 0.6306,
+      "step": 39512
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.25707453616221e-07,
+      "loss": 0.6189,
+      "step": 39513
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.254793962056081e-07,
+      "loss": 0.5918,
+      "step": 39514
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.252513869378117e-07,
+      "loss": 0.6084,
+      "step": 39515
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.250234258139874e-07,
+      "loss": 0.5951,
+      "step": 39516
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.247955128352932e-07,
+      "loss": 0.6481,
+      "step": 39517
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.245676480028905e-07,
+      "loss": 0.5412,
+      "step": 39518
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.243398313179316e-07,
+      "loss": 0.6044,
+      "step": 39519
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.241120627815776e-07,
+      "loss": 0.6676,
+      "step": 39520
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.238843423949858e-07,
+      "loss": 0.5836,
+      "step": 39521
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.236566701593126e-07,
+      "loss": 0.5874,
+      "step": 39522
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.234290460757163e-07,
+      "loss": 0.595,
+      "step": 39523
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.232014701453513e-07,
+      "loss": 0.6122,
+      "step": 39524
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.229739423693747e-07,
+      "loss": 0.508,
+      "step": 39525
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.227464627489443e-07,
+      "loss": 0.6212,
+      "step": 39526
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.225190312852124e-07,
+      "loss": 0.6219,
+      "step": 39527
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.222916479793383e-07,
+      "loss": 0.5691,
+      "step": 39528
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.220643128324765e-07,
+      "loss": 0.5663,
+      "step": 39529
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.218370258457816e-07,
+      "loss": 0.6173,
+      "step": 39530
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.216097870204073e-07,
+      "loss": 0.6258,
+      "step": 39531
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.213825963575114e-07,
+      "loss": 0.5812,
+      "step": 39532
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.211554538582464e-07,
+      "loss": 0.6293,
+      "step": 39533
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.209283595237657e-07,
+      "loss": 0.5735,
+      "step": 39534
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.20701313355223e-07,
+      "loss": 0.6124,
+      "step": 39535
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.20474315353775e-07,
+      "loss": 0.6218,
+      "step": 39536
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.202473655205719e-07,
+      "loss": 0.5589,
+      "step": 39537
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.200204638567685e-07,
+      "loss": 0.6292,
+      "step": 39538
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.197936103635193e-07,
+      "loss": 0.5847,
+      "step": 39539
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.195668050419722e-07,
+      "loss": 0.5943,
+      "step": 39540
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.193400478932842e-07,
+      "loss": 0.5656,
+      "step": 39541
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.191133389186054e-07,
+      "loss": 0.5997,
+      "step": 39542
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.188866781190883e-07,
+      "loss": 0.5755,
+      "step": 39543
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.18660065495884e-07,
+      "loss": 0.565,
+      "step": 39544
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.184335010501474e-07,
+      "loss": 0.6205,
+      "step": 39545
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.18206984783024e-07,
+      "loss": 0.563,
+      "step": 39546
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.179805166956698e-07,
+      "loss": 0.6359,
+      "step": 39547
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.177540967892314e-07,
+      "loss": 0.5827,
+      "step": 39548
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.175277250648636e-07,
+      "loss": 0.5757,
+      "step": 39549
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.173014015237143e-07,
+      "loss": 0.6023,
+      "step": 39550
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.170751261669338e-07,
+      "loss": 0.5873,
+      "step": 39551
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.1684889899567e-07,
+      "loss": 0.659,
+      "step": 39552
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.166227200110752e-07,
+      "loss": 0.5978,
+      "step": 39553
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.163965892142985e-07,
+      "loss": 0.5849,
+      "step": 39554
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.161705066064859e-07,
+      "loss": 0.5927,
+      "step": 39555
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.159444721887907e-07,
+      "loss": 0.6046,
+      "step": 39556
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.157184859623587e-07,
+      "loss": 0.5789,
+      "step": 39557
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.15492547928339e-07,
+      "loss": 0.5696,
+      "step": 39558
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.152666580878774e-07,
+      "loss": 0.5711,
+      "step": 39559
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.150408164421261e-07,
+      "loss": 0.5877,
+      "step": 39560
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.148150229922289e-07,
+      "loss": 0.536,
+      "step": 39561
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.145892777393347e-07,
+      "loss": 0.6094,
+      "step": 39562
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.143635806845892e-07,
+      "loss": 0.5595,
+      "step": 39563
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.141379318291417e-07,
+      "loss": 0.6062,
+      "step": 39564
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.139123311741368e-07,
+      "loss": 0.5884,
+      "step": 39565
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.136867787207223e-07,
+      "loss": 0.6215,
+      "step": 39566
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.134612744700429e-07,
+      "loss": 0.5982,
+      "step": 39567
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.132358184232456e-07,
+      "loss": 0.5458,
+      "step": 39568
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.130104105814737e-07,
+      "loss": 0.5889,
+      "step": 39569
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.127850509458765e-07,
+      "loss": 0.5836,
+      "step": 39570
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.125597395175962e-07,
+      "loss": 0.584,
+      "step": 39571
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.123344762977778e-07,
+      "loss": 0.6185,
+      "step": 39572
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.121092612875667e-07,
+      "loss": 0.575,
+      "step": 39573
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.118840944881076e-07,
+      "loss": 0.6116,
+      "step": 39574
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.116589759005441e-07,
+      "loss": 0.6146,
+      "step": 39575
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.114339055260186e-07,
+      "loss": 0.6005,
+      "step": 39576
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.112088833656792e-07,
+      "loss": 0.6328,
+      "step": 39577
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.109839094206637e-07,
+      "loss": 0.5947,
+      "step": 39578
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.10758983692119e-07,
+      "loss": 0.632,
+      "step": 39579
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.105341061811864e-07,
+      "loss": 0.5508,
+      "step": 39580
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.103092768890095e-07,
+      "loss": 0.6543,
+      "step": 39581
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.100844958167295e-07,
+      "loss": 0.5687,
+      "step": 39582
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.098597629654922e-07,
+      "loss": 0.5615,
+      "step": 39583
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.096350783364345e-07,
+      "loss": 0.5507,
+      "step": 39584
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.094104419307011e-07,
+      "loss": 0.6016,
+      "step": 39585
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.09185853749431e-07,
+      "loss": 0.5789,
+      "step": 39586
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.089613137937687e-07,
+      "loss": 0.6224,
+      "step": 39587
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.087368220648537e-07,
+      "loss": 0.5915,
+      "step": 39588
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.085123785638268e-07,
+      "loss": 0.6002,
+      "step": 39589
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.082879832918264e-07,
+      "loss": 0.6246,
+      "step": 39590
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.080636362499958e-07,
+      "loss": 0.5406,
+      "step": 39591
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.078393374394742e-07,
+      "loss": 0.5784,
+      "step": 39592
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.076150868613994e-07,
+      "loss": 0.6306,
+      "step": 39593
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.07390884516914e-07,
+      "loss": 0.5874,
+      "step": 39594
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.071667304071548e-07,
+      "loss": 0.6331,
+      "step": 39595
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.069426245332621e-07,
+      "loss": 0.6338,
+      "step": 39596
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.067185668963726e-07,
+      "loss": 0.5838,
+      "step": 39597
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.064945574976276e-07,
+      "loss": 0.5985,
+      "step": 39598
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.062705963381642e-07,
+      "loss": 0.6001,
+      "step": 39599
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.060466834191191e-07,
+      "loss": 0.5741,
+      "step": 39600
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 5.058228187416314e-07,
+      "loss": 0.6312,
+      "step": 39601
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.055990023068391e-07,
+      "loss": 0.6019,
+      "step": 39602
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.053752341158779e-07,
+      "loss": 0.6009,
+      "step": 39603
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.051515141698882e-07,
+      "loss": 0.6169,
+      "step": 39604
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.04927842470001e-07,
+      "loss": 0.5932,
+      "step": 39605
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.04704219017359e-07,
+      "loss": 0.5565,
+      "step": 39606
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.044806438130934e-07,
+      "loss": 0.576,
+      "step": 39607
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.042571168583443e-07,
+      "loss": 0.634,
+      "step": 39608
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.040336381542454e-07,
+      "loss": 0.6002,
+      "step": 39609
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.038102077019324e-07,
+      "loss": 0.5947,
+      "step": 39610
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.0358682550254e-07,
+      "loss": 0.6429,
+      "step": 39611
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.03363491557205e-07,
+      "loss": 0.5887,
+      "step": 39612
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.03140205867062e-07,
+      "loss": 0.6575,
+      "step": 39613
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.029169684332424e-07,
+      "loss": 0.5672,
+      "step": 39614
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.026937792568854e-07,
+      "loss": 0.616,
+      "step": 39615
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.02470638339122e-07,
+      "loss": 0.5819,
+      "step": 39616
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.022475456810872e-07,
+      "loss": 0.6376,
+      "step": 39617
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.020245012839131e-07,
+      "loss": 0.6609,
+      "step": 39618
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.018015051487346e-07,
+      "loss": 0.5735,
+      "step": 39619
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.015785572766841e-07,
+      "loss": 0.6429,
+      "step": 39620
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.013556576688971e-07,
+      "loss": 0.5921,
+      "step": 39621
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.011328063265009e-07,
+      "loss": 0.6204,
+      "step": 39622
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.00910003250632e-07,
+      "loss": 0.6149,
+      "step": 39623
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.006872484424196e-07,
+      "loss": 0.6215,
+      "step": 39624
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.004645419029996e-07,
+      "loss": 0.6087,
+      "step": 39625
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.002418836335e-07,
+      "loss": 0.664,
+      "step": 39626
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 5.000192736350551e-07,
+      "loss": 0.544,
+      "step": 39627
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.997967119087921e-07,
+      "loss": 0.6056,
+      "step": 39628
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.995741984558456e-07,
+      "loss": 0.6322,
+      "step": 39629
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.993517332773457e-07,
+      "loss": 0.5759,
+      "step": 39630
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.991293163744215e-07,
+      "loss": 0.545,
+      "step": 39631
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.989069477482022e-07,
+      "loss": 0.6096,
+      "step": 39632
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.986846273998202e-07,
+      "loss": 0.5351,
+      "step": 39633
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.984623553304047e-07,
+      "loss": 0.5965,
+      "step": 39634
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.982401315410834e-07,
+      "loss": 0.5854,
+      "step": 39635
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.980179560329879e-07,
+      "loss": 0.5735,
+      "step": 39636
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.977958288072459e-07,
+      "loss": 0.6481,
+      "step": 39637
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.975737498649868e-07,
+      "loss": 0.5947,
+      "step": 39638
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.973517192073363e-07,
+      "loss": 0.6217,
+      "step": 39639
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.971297368354266e-07,
+      "loss": 0.6023,
+      "step": 39640
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.969078027503815e-07,
+      "loss": 0.5977,
+      "step": 39641
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.966859169533344e-07,
+      "loss": 0.6365,
+      "step": 39642
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.964640794454067e-07,
+      "loss": 0.6058,
+      "step": 39643
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.962422902277298e-07,
+      "loss": 0.6193,
+      "step": 39644
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.960205493014269e-07,
+      "loss": 0.6259,
+      "step": 39645
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.957988566676286e-07,
+      "loss": 0.5791,
+      "step": 39646
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.955772123274583e-07,
+      "loss": 0.4868,
+      "step": 39647
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.953556162820472e-07,
+      "loss": 0.5943,
+      "step": 39648
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.951340685325145e-07,
+      "loss": 0.5758,
+      "step": 39649
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.949125690799905e-07,
+      "loss": 0.6074,
+      "step": 39650
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.946911179255998e-07,
+      "loss": 0.6006,
+      "step": 39651
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.944697150704647e-07,
+      "loss": 0.5916,
+      "step": 39652
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.942483605157155e-07,
+      "loss": 0.5579,
+      "step": 39653
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.940270542624736e-07,
+      "loss": 0.587,
+      "step": 39654
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.938057963118647e-07,
+      "loss": 0.6093,
+      "step": 39655
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.935845866650113e-07,
+      "loss": 0.5747,
+      "step": 39656
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.933634253230391e-07,
+      "loss": 0.5555,
+      "step": 39657
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.931423122870716e-07,
+      "loss": 0.5947,
+      "step": 39658
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.929212475582334e-07,
+      "loss": 0.536,
+      "step": 39659
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.92700231137645e-07,
+      "loss": 0.6202,
+      "step": 39660
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.924792630264319e-07,
+      "loss": 0.6277,
+      "step": 39661
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.922583432257144e-07,
+      "loss": 0.5723,
+      "step": 39662
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.920374717366183e-07,
+      "loss": 0.5945,
+      "step": 39663
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.918166485602649e-07,
+      "loss": 0.6556,
+      "step": 39664
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.915958736977755e-07,
+      "loss": 0.5918,
+      "step": 39665
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.913751471502703e-07,
+      "loss": 0.6447,
+      "step": 39666
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.911544689188752e-07,
+      "loss": 0.5373,
+      "step": 39667
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.90933839004708e-07,
+      "loss": 0.6126,
+      "step": 39668
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.907132574088913e-07,
+      "loss": 0.6052,
+      "step": 39669
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.904927241325431e-07,
+      "loss": 0.5105,
+      "step": 39670
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.902722391767889e-07,
+      "loss": 0.5875,
+      "step": 39671
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.900518025427458e-07,
+      "loss": 0.534,
+      "step": 39672
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.898314142315341e-07,
+      "loss": 0.5523,
+      "step": 39673
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.89611074244275e-07,
+      "loss": 0.6532,
+      "step": 39674
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.893907825820865e-07,
+      "loss": 0.5797,
+      "step": 39675
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.891705392460899e-07,
+      "loss": 0.6142,
+      "step": 39676
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.889503442374011e-07,
+      "loss": 0.555,
+      "step": 39677
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.887301975571435e-07,
+      "loss": 0.5264,
+      "step": 39678
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.885100992064307e-07,
+      "loss": 0.6044,
+      "step": 39679
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.882900491863863e-07,
+      "loss": 0.586,
+      "step": 39680
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.880700474981237e-07,
+      "loss": 0.5922,
+      "step": 39681
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.878500941427644e-07,
+      "loss": 0.5576,
+      "step": 39682
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.87630189121423e-07,
+      "loss": 0.5821,
+      "step": 39683
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.874103324352197e-07,
+      "loss": 0.5835,
+      "step": 39684
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.871905240852692e-07,
+      "loss": 0.6488,
+      "step": 39685
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.869707640726917e-07,
+      "loss": 0.5986,
+      "step": 39686
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.867510523985997e-07,
+      "loss": 0.5989,
+      "step": 39687
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.865313890641121e-07,
+      "loss": 0.6309,
+      "step": 39688
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.86311774070346e-07,
+      "loss": 0.6076,
+      "step": 39689
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.860922074184138e-07,
+      "loss": 0.6261,
+      "step": 39690
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.858726891094345e-07,
+      "loss": 0.652,
+      "step": 39691
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.856532191445229e-07,
+      "loss": 0.5866,
+      "step": 39692
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.854337975247936e-07,
+      "loss": 0.5696,
+      "step": 39693
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.852144242513612e-07,
+      "loss": 0.5554,
+      "step": 39694
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.849950993253405e-07,
+      "loss": 0.5789,
+      "step": 39695
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.847758227478472e-07,
+      "loss": 0.6283,
+      "step": 39696
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.845565945199948e-07,
+      "loss": 0.6544,
+      "step": 39697
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.843374146428958e-07,
+      "loss": 0.6012,
+      "step": 39698
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.841182831176661e-07,
+      "loss": 0.5931,
+      "step": 39699
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.838991999454168e-07,
+      "loss": 0.6417,
+      "step": 39700
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.836801651272638e-07,
+      "loss": 0.5707,
+      "step": 39701
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.834611786643195e-07,
+      "loss": 0.6006,
+      "step": 39702
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.832422405576964e-07,
+      "loss": 0.556,
+      "step": 39703
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.830233508085036e-07,
+      "loss": 0.6286,
+      "step": 39704
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.82804509417859e-07,
+      "loss": 0.5294,
+      "step": 39705
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.825857163868708e-07,
+      "loss": 0.5844,
+      "step": 39706
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.823669717166535e-07,
+      "loss": 0.6069,
+      "step": 39707
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.821482754083151e-07,
+      "loss": 0.5726,
+      "step": 39708
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.819296274629693e-07,
+      "loss": 0.5804,
+      "step": 39709
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.817110278817272e-07,
+      "loss": 0.6677,
+      "step": 39710
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.814924766656969e-07,
+      "loss": 0.6246,
+      "step": 39711
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.812739738159933e-07,
+      "loss": 0.6146,
+      "step": 39712
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.81055519333724e-07,
+      "loss": 0.5156,
+      "step": 39713
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.808371132199985e-07,
+      "loss": 0.6127,
+      "step": 39714
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.806187554759267e-07,
+      "loss": 0.5839,
+      "step": 39715
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.804004461026212e-07,
+      "loss": 0.5814,
+      "step": 39716
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.801821851011867e-07,
+      "loss": 0.5969,
+      "step": 39717
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.799639724727367e-07,
+      "loss": 0.5847,
+      "step": 39718
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.797458082183759e-07,
+      "loss": 0.6186,
+      "step": 39719
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.795276923392156e-07,
+      "loss": 0.66,
+      "step": 39720
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.793096248363626e-07,
+      "loss": 0.6295,
+      "step": 39721
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.790916057109263e-07,
+      "loss": 0.6469,
+      "step": 39722
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.788736349640144e-07,
+      "loss": 0.5781,
+      "step": 39723
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.786557125967339e-07,
+      "loss": 0.6147,
+      "step": 39724
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.784378386101907e-07,
+      "loss": 0.6339,
+      "step": 39725
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.782200130054948e-07,
+      "loss": 0.5911,
+      "step": 39726
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.780022357837499e-07,
+      "loss": 0.5866,
+      "step": 39727
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.777845069460663e-07,
+      "loss": 0.6327,
+      "step": 39728
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.775668264935485e-07,
+      "loss": 0.6542,
+      "step": 39729
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.773491944273012e-07,
+      "loss": 0.5983,
+      "step": 39730
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.771316107484325e-07,
+      "loss": 0.6487,
+      "step": 39731
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.769140754580448e-07,
+      "loss": 0.6323,
+      "step": 39732
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.766965885572483e-07,
+      "loss": 0.6473,
+      "step": 39733
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.764791500471444e-07,
+      "loss": 0.5865,
+      "step": 39734
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7626175992883884e-07,
+      "loss": 0.5413,
+      "step": 39735
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.760444182034363e-07,
+      "loss": 0.6183,
+      "step": 39736
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.758271248720414e-07,
+      "loss": 0.645,
+      "step": 39737
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.756098799357578e-07,
+      "loss": 0.6276,
+      "step": 39738
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7539268339569013e-07,
+      "loss": 0.606,
+      "step": 39739
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7517553525294193e-07,
+      "loss": 0.5811,
+      "step": 39740
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7495843550861676e-07,
+      "loss": 0.5894,
+      "step": 39741
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.74741384163816e-07,
+      "loss": 0.5643,
+      "step": 39742
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7452438121964427e-07,
+      "loss": 0.6147,
+      "step": 39743
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.74307426677203e-07,
+      "loss": 0.6433,
+      "step": 39744
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7409052053759896e-07,
+      "loss": 0.5715,
+      "step": 39745
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.7387366280192695e-07,
+      "loss": 0.59,
+      "step": 39746
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.736568534712949e-07,
+      "loss": 0.5971,
+      "step": 39747
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.734400925468008e-07,
+      "loss": 0.6341,
+      "step": 39748
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7322338002954825e-07,
+      "loss": 0.5554,
+      "step": 39749
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7300671592063754e-07,
+      "loss": 0.5757,
+      "step": 39750
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.72790100221171e-07,
+      "loss": 0.5855,
+      "step": 39751
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7257353293224785e-07,
+      "loss": 0.6262,
+      "step": 39752
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7235701405496714e-07,
+      "loss": 0.583,
+      "step": 39753
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7214054359043136e-07,
+      "loss": 0.5999,
+      "step": 39754
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7192412153973964e-07,
+      "loss": 0.6044,
+      "step": 39755
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7170774790399444e-07,
+      "loss": 0.6361,
+      "step": 39756
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7149142268428924e-07,
+      "loss": 0.5669,
+      "step": 39757
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7127514588172883e-07,
+      "loss": 0.6401,
+      "step": 39758
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7105891749740784e-07,
+      "loss": 0.5892,
+      "step": 39759
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.708427375324298e-07,
+      "loss": 0.6281,
+      "step": 39760
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7062660598788944e-07,
+      "loss": 0.5775,
+      "step": 39761
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.70410522864887e-07,
+      "loss": 0.6104,
+      "step": 39762
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.7019448816451707e-07,
+      "loss": 0.5985,
+      "step": 39763
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6997850188788216e-07,
+      "loss": 0.6156,
+      "step": 39764
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.697625640360759e-07,
+      "loss": 0.623,
+      "step": 39765
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6954667461019843e-07,
+      "loss": 0.631,
+      "step": 39766
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.693308336113445e-07,
+      "loss": 0.6371,
+      "step": 39767
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6911504104061313e-07,
+      "loss": 0.6507,
+      "step": 39768
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6889929689909907e-07,
+      "loss": 0.565,
+      "step": 39769
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.68683601187897e-07,
+      "loss": 0.5403,
+      "step": 39770
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.684679539081072e-07,
+      "loss": 0.6207,
+      "step": 39771
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.682523550608231e-07,
+      "loss": 0.5939,
+      "step": 39772
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.680368046471395e-07,
+      "loss": 0.5766,
+      "step": 39773
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6782130266815216e-07,
+      "loss": 0.6504,
+      "step": 39774
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6760584912495686e-07,
+      "loss": 0.6532,
+      "step": 39775
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.673904440186472e-07,
+      "loss": 0.6234,
+      "step": 39776
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.671750873503189e-07,
+      "loss": 0.5969,
+      "step": 39777
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.669597791210667e-07,
+      "loss": 0.5736,
+      "step": 39778
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6674451933198307e-07,
+      "loss": 0.5553,
+      "step": 39779
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.665293079841615e-07,
+      "loss": 0.6329,
+      "step": 39780
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.663141450786979e-07,
+      "loss": 0.5928,
+      "step": 39781
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.660990306166835e-07,
+      "loss": 0.6117,
+      "step": 39782
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6588396459921416e-07,
+      "loss": 0.5971,
+      "step": 39783
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6566894702737786e-07,
+      "loss": 0.5928,
+      "step": 39784
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.654539779022704e-07,
+      "loss": 0.6193,
+      "step": 39785
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.652390572249832e-07,
+      "loss": 0.5559,
+      "step": 39786
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.650241849966108e-07,
+      "loss": 0.5737,
+      "step": 39787
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.648093612182414e-07,
+      "loss": 0.5999,
+      "step": 39788
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.645945858909695e-07,
+      "loss": 0.6286,
+      "step": 39789
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6437985901588435e-07,
+      "loss": 0.5926,
+      "step": 39790
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6416518059407613e-07,
+      "loss": 0.5891,
+      "step": 39791
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6395055062663953e-07,
+      "loss": 0.6058,
+      "step": 39792
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6373596911466035e-07,
+      "loss": 0.5985,
+      "step": 39793
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.635214360592355e-07,
+      "loss": 0.5271,
+      "step": 39794
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6330695146144744e-07,
+      "loss": 0.5692,
+      "step": 39795
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.63092515322392e-07,
+      "loss": 0.6042,
+      "step": 39796
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.628781276431549e-07,
+      "loss": 0.5883,
+      "step": 39797
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.626637884248286e-07,
+      "loss": 0.6244,
+      "step": 39798
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.624494976685001e-07,
+      "loss": 0.5722,
+      "step": 39799
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.622352553752596e-07,
+      "loss": 0.6171,
+      "step": 39800
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.620210615461929e-07,
+      "loss": 0.6088,
+      "step": 39801
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6180691618239236e-07,
+      "loss": 0.5906,
+      "step": 39802
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6159281928494284e-07,
+      "loss": 0.6075,
+      "step": 39803
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6137877085493445e-07,
+      "loss": 0.5695,
+      "step": 39804
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6116477089345525e-07,
+      "loss": 0.5959,
+      "step": 39805
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6095081940158993e-07,
+      "loss": 0.6011,
+      "step": 39806
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.607369163804265e-07,
+      "loss": 0.5246,
+      "step": 39807
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.6052306183105413e-07,
+      "loss": 0.6344,
+      "step": 39808
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.603092557545574e-07,
+      "loss": 0.6435,
+      "step": 39809
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.600954981520245e-07,
+      "loss": 0.5398,
+      "step": 39810
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5988178902453885e-07,
+      "loss": 0.668,
+      "step": 39811
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5966812837318633e-07,
+      "loss": 0.5772,
+      "step": 39812
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.59454516199056e-07,
+      "loss": 0.5895,
+      "step": 39813
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.592409525032304e-07,
+      "loss": 0.6315,
+      "step": 39814
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5902743728679754e-07,
+      "loss": 0.6101,
+      "step": 39815
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5881397055083767e-07,
+      "loss": 0.5837,
+      "step": 39816
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5860055229643986e-07,
+      "loss": 0.5917,
+      "step": 39817
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5838718252468663e-07,
+      "loss": 0.548,
+      "step": 39818
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5817386123666266e-07,
+      "loss": 0.5925,
+      "step": 39819
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.579605884334504e-07,
+      "loss": 0.5732,
+      "step": 39820
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5774736411613785e-07,
+      "loss": 0.6203,
+      "step": 39821
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.575341882858031e-07,
+      "loss": 0.5262,
+      "step": 39822
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5732106094353303e-07,
+      "loss": 0.6258,
+      "step": 39823
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5710798209040783e-07,
+      "loss": 0.6259,
+      "step": 39824
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.568949517275134e-07,
+      "loss": 0.6396,
+      "step": 39825
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5668196985593106e-07,
+      "loss": 0.574,
+      "step": 39826
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.56469036476741e-07,
+      "loss": 0.6332,
+      "step": 39827
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5625615159102687e-07,
+      "loss": 0.5556,
+      "step": 39828
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.560433151998711e-07,
+      "loss": 0.5754,
+      "step": 39829
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5583052730435395e-07,
+      "loss": 0.6267,
+      "step": 39830
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.556177879055557e-07,
+      "loss": 0.5999,
+      "step": 39831
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5540509700456093e-07,
+      "loss": 0.5472,
+      "step": 39832
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.551924546024467e-07,
+      "loss": 0.6181,
+      "step": 39833
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.549798607002953e-07,
+      "loss": 0.5913,
+      "step": 39834
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5476731529918606e-07,
+      "loss": 0.6185,
+      "step": 39835
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.545548184002002e-07,
+      "loss": 0.6002,
+      "step": 39836
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5434237000441696e-07,
+      "loss": 0.6133,
+      "step": 39837
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5412997011291537e-07,
+      "loss": 0.6424,
+      "step": 39838
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5391761872677467e-07,
+      "loss": 0.5773,
+      "step": 39839
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.53705315847075e-07,
+      "loss": 0.5942,
+      "step": 39840
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5349306147489337e-07,
+      "loss": 0.5139,
+      "step": 39841
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5328085561130995e-07,
+      "loss": 0.6135,
+      "step": 39842
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5306869825740284e-07,
+      "loss": 0.5645,
+      "step": 39843
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5285658941425005e-07,
+      "loss": 0.5943,
+      "step": 39844
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5264452908292625e-07,
+      "loss": 0.5526,
+      "step": 39845
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.524325172645139e-07,
+      "loss": 0.6098,
+      "step": 39846
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5222055396008877e-07,
+      "loss": 0.5822,
+      "step": 39847
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.520086391707257e-07,
+      "loss": 0.5689,
+      "step": 39848
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5179677289750256e-07,
+      "loss": 0.5996,
+      "step": 39849
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.515849551414975e-07,
+      "loss": 0.5851,
+      "step": 39850
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5137318590378623e-07,
+      "loss": 0.6512,
+      "step": 39851
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5116146518544234e-07,
+      "loss": 0.573,
+      "step": 39852
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5094979298754617e-07,
+      "loss": 0.5381,
+      "step": 39853
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.507381693111679e-07,
+      "loss": 0.5945,
+      "step": 39854
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.5052659415738775e-07,
+      "loss": 0.5769,
+      "step": 39855
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.503150675272772e-07,
+      "loss": 0.5912,
+      "step": 39856
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.50103589421913e-07,
+      "loss": 0.5931,
+      "step": 39857
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4989215984236776e-07,
+      "loss": 0.6755,
+      "step": 39858
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.496807787897206e-07,
+      "loss": 0.638,
+      "step": 39859
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4946944626504065e-07,
+      "loss": 0.5887,
+      "step": 39860
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.492581622694037e-07,
+      "loss": 0.5986,
+      "step": 39861
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4904692680388216e-07,
+      "loss": 0.5923,
+      "step": 39862
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.488357398695509e-07,
+      "loss": 0.5377,
+      "step": 39863
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.486246014674833e-07,
+      "loss": 0.6305,
+      "step": 39864
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.484135115987509e-07,
+      "loss": 0.6462,
+      "step": 39865
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4820247026442496e-07,
+      "loss": 0.6057,
+      "step": 39866
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4799147746558134e-07,
+      "loss": 0.586,
+      "step": 39867
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.477805332032914e-07,
+      "loss": 0.6267,
+      "step": 39868
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4756963747862315e-07,
+      "loss": 0.6274,
+      "step": 39869
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.473587902926546e-07,
+      "loss": 0.6606,
+      "step": 39870
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.471479916464516e-07,
+      "loss": 0.6332,
+      "step": 39871
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4693724154108777e-07,
+      "loss": 0.5837,
+      "step": 39872
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.467265399776333e-07,
+      "loss": 0.5311,
+      "step": 39873
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.465158869571595e-07,
+      "loss": 0.6183,
+      "step": 39874
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.463052824807357e-07,
+      "loss": 0.6002,
+      "step": 39875
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.460947265494342e-07,
+      "loss": 0.6142,
+      "step": 39876
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.458842191643209e-07,
+      "loss": 0.5563,
+      "step": 39877
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4567376032646934e-07,
+      "loss": 0.6348,
+      "step": 39878
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.454633500369476e-07,
+      "loss": 0.5699,
+      "step": 39879
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4525298829682595e-07,
+      "loss": 0.6372,
+      "step": 39880
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4504267510717014e-07,
+      "loss": 0.5968,
+      "step": 39881
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.448324104690516e-07,
+      "loss": 0.615,
+      "step": 39882
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4462219438353714e-07,
+      "loss": 0.5706,
+      "step": 39883
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4441202685169606e-07,
+      "loss": 0.636,
+      "step": 39884
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4420190787459736e-07,
+      "loss": 0.5996,
+      "step": 39885
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4399183745330587e-07,
+      "loss": 0.5579,
+      "step": 39886
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4378181558889064e-07,
+      "loss": 0.5645,
+      "step": 39887
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.435718422824187e-07,
+      "loss": 0.603,
+      "step": 39888
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4336191753495794e-07,
+      "loss": 0.5952,
+      "step": 39889
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4315204134757207e-07,
+      "loss": 0.5804,
+      "step": 39890
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.429422137213324e-07,
+      "loss": 0.6205,
+      "step": 39891
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.427324346573003e-07,
+      "loss": 0.6143,
+      "step": 39892
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.4252270415654385e-07,
+      "loss": 0.6011,
+      "step": 39893
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.423130222201277e-07,
+      "loss": 0.6592,
+      "step": 39894
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.4210338884911996e-07,
+      "loss": 0.6012,
+      "step": 39895
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.418938040445819e-07,
+      "loss": 0.6267,
+      "step": 39896
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.416842678075839e-07,
+      "loss": 0.5563,
+      "step": 39897
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.414747801391839e-07,
+      "loss": 0.5971,
+      "step": 39898
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.4126534104045214e-07,
+      "loss": 0.6552,
+      "step": 39899
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.41055950512449e-07,
+      "loss": 0.6042,
+      "step": 39900
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.408466085562413e-07,
+      "loss": 0.5292,
+      "step": 39901
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.4063731517289156e-07,
+      "loss": 0.6712,
+      "step": 39902
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.4042807036346334e-07,
+      "loss": 0.6369,
+      "step": 39903
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.402188741290192e-07,
+      "loss": 0.5686,
+      "step": 39904
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.4000972647062267e-07,
+      "loss": 0.5802,
+      "step": 39905
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.398006273893374e-07,
+      "loss": 0.6113,
+      "step": 39906
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3959157688622355e-07,
+      "loss": 0.6,
+      "step": 39907
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.393825749623448e-07,
+      "loss": 0.5979,
+      "step": 39908
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3917362161876364e-07,
+      "loss": 0.5279,
+      "step": 39909
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.389647168565425e-07,
+      "loss": 0.643,
+      "step": 39910
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3875586067673947e-07,
+      "loss": 0.5351,
+      "step": 39911
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3854705308041924e-07,
+      "loss": 0.6264,
+      "step": 39912
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.38338294068642e-07,
+      "loss": 0.6625,
+      "step": 39913
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.381295836424682e-07,
+      "loss": 0.6302,
+      "step": 39914
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3792092180295674e-07,
+      "loss": 0.57,
+      "step": 39915
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.377123085511714e-07,
+      "loss": 0.6036,
+      "step": 39916
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3750374388816907e-07,
+      "loss": 0.6253,
+      "step": 39917
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.372952278150122e-07,
+      "loss": 0.5828,
+      "step": 39918
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3708676033275663e-07,
+      "loss": 0.5788,
+      "step": 39919
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3687834144246597e-07,
+      "loss": 0.6047,
+      "step": 39920
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3666997114519495e-07,
+      "loss": 0.5687,
+      "step": 39921
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3646164944200596e-07,
+      "loss": 0.5819,
+      "step": 39922
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3625337633395605e-07,
+      "loss": 0.5939,
+      "step": 39923
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3604515182210427e-07,
+      "loss": 0.6893,
+      "step": 39924
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.358369759075065e-07,
+      "loss": 0.5914,
+      "step": 39925
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.35628848591223e-07,
+      "loss": 0.6251,
+      "step": 39926
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.354207698743107e-07,
+      "loss": 0.5887,
+      "step": 39927
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.352127397578254e-07,
+      "loss": 0.6266,
+      "step": 39928
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.350047582428263e-07,
+      "loss": 0.651,
+      "step": 39929
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3479682533036913e-07,
+      "loss": 0.6233,
+      "step": 39930
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3458894102151093e-07,
+      "loss": 0.5979,
+      "step": 39931
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3438110531730636e-07,
+      "loss": 0.6356,
+      "step": 39932
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.341733182188135e-07,
+      "loss": 0.6103,
+      "step": 39933
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3396557972708696e-07,
+      "loss": 0.6061,
+      "step": 39934
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.337578898431849e-07,
+      "loss": 0.6086,
+      "step": 39935
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3355024856815864e-07,
+      "loss": 0.6005,
+      "step": 39936
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.333426559030662e-07,
+      "loss": 0.5532,
+      "step": 39937
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.331351118489602e-07,
+      "loss": 0.5822,
+      "step": 39938
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.329276164068974e-07,
+      "loss": 0.4921,
+      "step": 39939
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.327201695779315e-07,
+      "loss": 0.5798,
+      "step": 39940
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.325127713631161e-07,
+      "loss": 0.5748,
+      "step": 39941
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3230542176350475e-07,
+      "loss": 0.5689,
+      "step": 39942
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3209812078015225e-07,
+      "loss": 0.5841,
+      "step": 39943
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.31890868414111e-07,
+      "loss": 0.5976,
+      "step": 39944
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3168366466643574e-07,
+      "loss": 0.5617,
+      "step": 39945
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3147650953817566e-07,
+      "loss": 0.556,
+      "step": 39946
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3126940303038654e-07,
+      "loss": 0.6055,
+      "step": 39947
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3106234514412094e-07,
+      "loss": 0.5955,
+      "step": 39948
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3085533588042796e-07,
+      "loss": 0.6068,
+      "step": 39949
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3064837524036343e-07,
+      "loss": 0.6341,
+      "step": 39950
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.304414632249765e-07,
+      "loss": 0.6065,
+      "step": 39951
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3023459983531855e-07,
+      "loss": 0.5792,
+      "step": 39952
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.30027785072441e-07,
+      "loss": 0.558,
+      "step": 39953
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.298210189373953e-07,
+      "loss": 0.5852,
+      "step": 39954
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.296143014312315e-07,
+      "loss": 0.5499,
+      "step": 39955
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2940763255500226e-07,
+      "loss": 0.6104,
+      "step": 39956
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.292010123097534e-07,
+      "loss": 0.6046,
+      "step": 39957
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2899444069653853e-07,
+      "loss": 0.569,
+      "step": 39958
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.287879177164045e-07,
+      "loss": 0.5487,
+      "step": 39959
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2858144337040273e-07,
+      "loss": 0.5913,
+      "step": 39960
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2837501765958244e-07,
+      "loss": 0.6115,
+      "step": 39961
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2816864058499165e-07,
+      "loss": 0.5755,
+      "step": 39962
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2796231214767834e-07,
+      "loss": 0.5438,
+      "step": 39963
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2775603234869177e-07,
+      "loss": 0.6105,
+      "step": 39964
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2754980118908217e-07,
+      "loss": 0.6199,
+      "step": 39965
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2734361866989317e-07,
+      "loss": 0.5809,
+      "step": 39966
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.27137484792175e-07,
+      "loss": 0.5966,
+      "step": 39967
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.269313995569768e-07,
+      "loss": 0.581,
+      "step": 39968
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.267253629653423e-07,
+      "loss": 0.6119,
+      "step": 39969
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.265193750183194e-07,
+      "loss": 0.5359,
+      "step": 39970
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2631343571695625e-07,
+      "loss": 0.6642,
+      "step": 39971
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2610754506229756e-07,
+      "loss": 0.5533,
+      "step": 39972
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2590170305539135e-07,
+      "loss": 0.6472,
+      "step": 39973
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2569590969728127e-07,
+      "loss": 0.557,
+      "step": 39974
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.254901649890153e-07,
+      "loss": 0.5773,
+      "step": 39975
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.25284468931636e-07,
+      "loss": 0.5719,
+      "step": 39976
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.250788215261925e-07,
+      "loss": 0.626,
+      "step": 39977
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.248732227737262e-07,
+      "loss": 0.5201,
+      "step": 39978
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.246676726752852e-07,
+      "loss": 0.6877,
+      "step": 39979
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.244621712319097e-07,
+      "loss": 0.6149,
+      "step": 39980
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.242567184446478e-07,
+      "loss": 0.6072,
+      "step": 39981
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.240513143145408e-07,
+      "loss": 0.596,
+      "step": 39982
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.238459588426347e-07,
+      "loss": 0.5999,
+      "step": 39983
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.236406520299707e-07,
+      "loss": 0.5709,
+      "step": 39984
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.234353938775937e-07,
+      "loss": 0.6577,
+      "step": 39985
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2323018438654607e-07,
+      "loss": 0.5985,
+      "step": 39986
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2302502355787035e-07,
+      "loss": 0.6251,
+      "step": 39987
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2281991139260905e-07,
+      "loss": 0.6316,
+      "step": 39988
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2261484789180574e-07,
+      "loss": 0.5751,
+      "step": 39989
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2240983305650076e-07,
+      "loss": 0.5905,
+      "step": 39990
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2220486688773434e-07,
+      "loss": 0.6114,
+      "step": 39991
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.219999493865523e-07,
+      "loss": 0.6253,
+      "step": 39992
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2179508055399165e-07,
+      "loss": 0.6227,
+      "step": 39993
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.215902603910993e-07,
+      "loss": 0.587,
+      "step": 39994
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.213854888989077e-07,
+      "loss": 0.5412,
+      "step": 39995
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2118076607846393e-07,
+      "loss": 0.603,
+      "step": 39996
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2097609193080367e-07,
+      "loss": 0.6102,
+      "step": 39997
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2077146645697173e-07,
+      "loss": 0.6224,
+      "step": 39998
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2056688965800506e-07,
+      "loss": 0.6031,
+      "step": 39999
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.203623615349428e-07,
+      "loss": 0.5755,
+      "step": 40000
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.2015788208882413e-07,
+      "loss": 0.5938,
+      "step": 40001
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1995345132069046e-07,
+      "loss": 0.6401,
+      "step": 40002
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1974906923157864e-07,
+      "loss": 0.6164,
+      "step": 40003
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.195447358225258e-07,
+      "loss": 0.5975,
+      "step": 40004
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.193404510945731e-07,
+      "loss": 0.5765,
+      "step": 40005
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.191362150487588e-07,
+      "loss": 0.6646,
+      "step": 40006
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.189320276861175e-07,
+      "loss": 0.5736,
+      "step": 40007
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.187278890076885e-07,
+      "loss": 0.6112,
+      "step": 40008
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1852379901450966e-07,
+      "loss": 0.6202,
+      "step": 40009
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.183197577076181e-07,
+      "loss": 0.6276,
+      "step": 40010
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1811576508804964e-07,
+      "loss": 0.6181,
+      "step": 40011
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1791182115684e-07,
+      "loss": 0.6177,
+      "step": 40012
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.177079259150274e-07,
+      "loss": 0.5866,
+      "step": 40013
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1750407936364644e-07,
+      "loss": 0.5878,
+      "step": 40014
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1730028150373523e-07,
+      "loss": 0.619,
+      "step": 40015
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.170965323363263e-07,
+      "loss": 0.6451,
+      "step": 40016
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1689283186245765e-07,
+      "loss": 0.6227,
+      "step": 40017
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1668918008316185e-07,
+      "loss": 0.5767,
+      "step": 40018
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.164855769994758e-07,
+      "loss": 0.5504,
+      "step": 40019
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1628202261243203e-07,
+      "loss": 0.5453,
+      "step": 40020
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1607851692306746e-07,
+      "loss": 0.5599,
+      "step": 40021
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.158750599324135e-07,
+      "loss": 0.5922,
+      "step": 40022
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1567165164150604e-07,
+      "loss": 0.5558,
+      "step": 40023
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.154682920513775e-07,
+      "loss": 0.596,
+      "step": 40024
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1526498116306156e-07,
+      "loss": 0.5729,
+      "step": 40025
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1506171897759074e-07,
+      "loss": 0.6167,
+      "step": 40026
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.148585054959997e-07,
+      "loss": 0.5674,
+      "step": 40027
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.146553407193199e-07,
+      "loss": 0.6699,
+      "step": 40028
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.144522246485816e-07,
+      "loss": 0.5524,
+      "step": 40029
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.142491572848195e-07,
+      "loss": 0.6453,
+      "step": 40030
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1404613862906395e-07,
+      "loss": 0.5765,
+      "step": 40031
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1384316868235075e-07,
+      "loss": 0.6135,
+      "step": 40032
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1364024744570465e-07,
+      "loss": 0.6333,
+      "step": 40033
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1343737492016146e-07,
+      "loss": 0.5682,
+      "step": 40034
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1323455110674817e-07,
+      "loss": 0.6605,
+      "step": 40035
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.130317760065006e-07,
+      "loss": 0.6214,
+      "step": 40036
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1282904962044456e-07,
+      "loss": 0.5794,
+      "step": 40037
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1262637194961264e-07,
+      "loss": 0.6077,
+      "step": 40038
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1242374299503285e-07,
+      "loss": 0.5931,
+      "step": 40039
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.1222116275773773e-07,
+      "loss": 0.5761,
+      "step": 40040
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.120186312387531e-07,
+      "loss": 0.5767,
+      "step": 40041
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1181614843911034e-07,
+      "loss": 0.6055,
+      "step": 40042
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1161371435983864e-07,
+      "loss": 0.5725,
+      "step": 40043
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.11411329001965e-07,
+      "loss": 0.6137,
+      "step": 40044
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1120899236651855e-07,
+      "loss": 0.6074,
+      "step": 40045
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.110067044545263e-07,
+      "loss": 0.5863,
+      "step": 40046
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1080446526701956e-07,
+      "loss": 0.5677,
+      "step": 40047
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.10602274805022e-07,
+      "loss": 0.5403,
+      "step": 40048
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1040013306956393e-07,
+      "loss": 0.598,
+      "step": 40049
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.1019804006166897e-07,
+      "loss": 0.7221,
+      "step": 40050
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0999599578236846e-07,
+      "loss": 0.5907,
+      "step": 40051
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.09794000232685e-07,
+      "loss": 0.6237,
+      "step": 40052
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0959205341364884e-07,
+      "loss": 0.5896,
+      "step": 40053
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0939015532628357e-07,
+      "loss": 0.5351,
+      "step": 40054
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0918830597161505e-07,
+      "loss": 0.6078,
+      "step": 40055
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.089865053506692e-07,
+      "loss": 0.5868,
+      "step": 40056
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0878475346447287e-07,
+      "loss": 0.6092,
+      "step": 40057
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0858305031404866e-07,
+      "loss": 0.5917,
+      "step": 40058
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.083813959004257e-07,
+      "loss": 0.6132,
+      "step": 40059
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0817979022462316e-07,
+      "loss": 0.5611,
+      "step": 40060
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0797823328766914e-07,
+      "loss": 0.5519,
+      "step": 40061
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.077767250905873e-07,
+      "loss": 0.6067,
+      "step": 40062
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0757526563440007e-07,
+      "loss": 0.6196,
+      "step": 40063
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0737385492013335e-07,
+      "loss": 0.6041,
+      "step": 40064
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0717249294880967e-07,
+      "loss": 0.6114,
+      "step": 40065
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.069711797214515e-07,
+      "loss": 0.617,
+      "step": 40066
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.067699152390814e-07,
+      "loss": 0.629,
+      "step": 40067
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.065686995027229e-07,
+      "loss": 0.6405,
+      "step": 40068
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.063675325133987e-07,
+      "loss": 0.5792,
+      "step": 40069
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.061664142721322e-07,
+      "loss": 0.6245,
+      "step": 40070
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0596534477994166e-07,
+      "loss": 0.5789,
+      "step": 40071
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.057643240378517e-07,
+      "loss": 0.6002,
+      "step": 40072
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0556335204688266e-07,
+      "loss": 0.5962,
+      "step": 40073
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.053624288080571e-07,
+      "loss": 0.5824,
+      "step": 40074
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.051615543223941e-07,
+      "loss": 0.5808,
+      "step": 40075
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0496072859091627e-07,
+      "loss": 0.6181,
+      "step": 40076
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0475995161464164e-07,
+      "loss": 0.607,
+      "step": 40077
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0455922339459274e-07,
+      "loss": 0.5762,
+      "step": 40078
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0435854393178766e-07,
+      "loss": 0.6111,
+      "step": 40079
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0415791322724776e-07,
+      "loss": 0.6191,
+      "step": 40080
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0395733128199223e-07,
+      "loss": 0.5263,
+      "step": 40081
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0375679809703915e-07,
+      "loss": 0.6763,
+      "step": 40082
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0355631367340884e-07,
+      "loss": 0.6129,
+      "step": 40083
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.033558780121172e-07,
+      "loss": 0.5829,
+      "step": 40084
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0315549111418774e-07,
+      "loss": 0.6301,
+      "step": 40085
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.029551529806341e-07,
+      "loss": 0.5899,
+      "step": 40086
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0275486361247675e-07,
+      "loss": 0.5958,
+      "step": 40087
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0255462301073133e-07,
+      "loss": 0.6448,
+      "step": 40088
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0235443117641825e-07,
+      "loss": 0.5953,
+      "step": 40089
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.021542881105511e-07,
+      "loss": 0.6042,
+      "step": 40090
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.019541938141513e-07,
+      "loss": 0.6056,
+      "step": 40091
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0175414828823146e-07,
+      "loss": 0.5954,
+      "step": 40092
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.015541515338106e-07,
+      "loss": 0.6382,
+      "step": 40093
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0135420355190356e-07,
+      "loss": 0.6132,
+      "step": 40094
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.011543043435273e-07,
+      "loss": 0.6258,
+      "step": 40095
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0095445390969545e-07,
+      "loss": 0.5975,
+      "step": 40096
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.007546522514294e-07,
+      "loss": 0.5971,
+      "step": 40097
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0055489936973613e-07,
+      "loss": 0.6122,
+      "step": 40098
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0035519526563706e-07,
+      "loss": 0.5791,
+      "step": 40099
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.0015553994014243e-07,
+      "loss": 0.5495,
+      "step": 40100
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.999559333942704e-07,
+      "loss": 0.6443,
+      "step": 40101
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9975637562903457e-07,
+      "loss": 0.5692,
+      "step": 40102
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9955686664544744e-07,
+      "loss": 0.5465,
+      "step": 40103
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.993574064445238e-07,
+      "loss": 0.5896,
+      "step": 40104
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9915799502727617e-07,
+      "loss": 0.6551,
+      "step": 40105
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9895863239471924e-07,
+      "loss": 0.5807,
+      "step": 40106
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.987593185478633e-07,
+      "loss": 0.5505,
+      "step": 40107
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.985600534877265e-07,
+      "loss": 0.619,
+      "step": 40108
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.983608372153147e-07,
+      "loss": 0.6378,
+      "step": 40109
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.981616697316459e-07,
+      "loss": 0.5942,
+      "step": 40110
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.979625510377272e-07,
+      "loss": 0.6743,
+      "step": 40111
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9776348113457433e-07,
+      "loss": 0.6109,
+      "step": 40112
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.975644600231987e-07,
+      "loss": 0.6192,
+      "step": 40113
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.973654877046085e-07,
+      "loss": 0.5981,
+      "step": 40114
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9716656417981613e-07,
+      "loss": 0.6147,
+      "step": 40115
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9696768944983423e-07,
+      "loss": 0.6232,
+      "step": 40116
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9676886351566966e-07,
+      "loss": 0.6204,
+      "step": 40117
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.965700863783373e-07,
+      "loss": 0.5908,
+      "step": 40118
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.96371358038844e-07,
+      "loss": 0.605,
+      "step": 40119
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.961726784982012e-07,
+      "loss": 0.6152,
+      "step": 40120
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.959740477574159e-07,
+      "loss": 0.594,
+      "step": 40121
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.957754658175006e-07,
+      "loss": 0.6776,
+      "step": 40122
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.955769326794623e-07,
+      "loss": 0.607,
+      "step": 40123
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.953784483443113e-07,
+      "loss": 0.589,
+      "step": 40124
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.951800128130545e-07,
+      "loss": 0.6036,
+      "step": 40125
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.949816260867001e-07,
+      "loss": 0.6088,
+      "step": 40126
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.947832881662572e-07,
+      "loss": 0.583,
+      "step": 40127
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9458499905273395e-07,
+      "loss": 0.6141,
+      "step": 40128
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.943867587471384e-07,
+      "loss": 0.6175,
+      "step": 40129
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9418856725047527e-07,
+      "loss": 0.6018,
+      "step": 40130
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.939904245637538e-07,
+      "loss": 0.5954,
+      "step": 40131
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.937923306879798e-07,
+      "loss": 0.6029,
+      "step": 40132
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.935942856241615e-07,
+      "loss": 0.542,
+      "step": 40133
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.933962893733023e-07,
+      "loss": 0.6238,
+      "step": 40134
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9319834193641273e-07,
+      "loss": 0.5835,
+      "step": 40135
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9300044331449296e-07,
+      "loss": 0.5634,
+      "step": 40136
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9280259350855223e-07,
+      "loss": 0.5838,
+      "step": 40137
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.926047925195953e-07,
+      "loss": 0.6131,
+      "step": 40138
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9240704034862686e-07,
+      "loss": 0.644,
+      "step": 40139
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9220933699665286e-07,
+      "loss": 0.673,
+      "step": 40140
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.920116824646769e-07,
+      "loss": 0.5476,
+      "step": 40141
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.9181407675370376e-07,
+      "loss": 0.5908,
+      "step": 40142
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.916165198647348e-07,
+      "loss": 0.608,
+      "step": 40143
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.914190117987782e-07,
+      "loss": 0.5826,
+      "step": 40144
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.912215525568353e-07,
+      "loss": 0.597,
+      "step": 40145
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.910241421399108e-07,
+      "loss": 0.6036,
+      "step": 40146
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.908267805490051e-07,
+      "loss": 0.5716,
+      "step": 40147
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.90629467785123e-07,
+      "loss": 0.6437,
+      "step": 40148
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.904322038492669e-07,
+      "loss": 0.5831,
+      "step": 40149
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.902349887424395e-07,
+      "loss": 0.5688,
+      "step": 40150
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.90037822465642e-07,
+      "loss": 0.6218,
+      "step": 40151
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8984070501987824e-07,
+      "loss": 0.6152,
+      "step": 40152
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.896436364061451e-07,
+      "loss": 0.6422,
+      "step": 40153
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.894466166254496e-07,
+      "loss": 0.6063,
+      "step": 40154
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8924964567878976e-07,
+      "loss": 0.5946,
+      "step": 40155
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8905272356716816e-07,
+      "loss": 0.616,
+      "step": 40156
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.888558502915829e-07,
+      "loss": 0.6144,
+      "step": 40157
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.886590258530365e-07,
+      "loss": 0.5771,
+      "step": 40158
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8846225025252814e-07,
+      "loss": 0.6364,
+      "step": 40159
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.882655234910582e-07,
+      "loss": 0.6081,
+      "step": 40160
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8806884556962574e-07,
+      "loss": 0.6011,
+      "step": 40161
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8787221648923123e-07,
+      "loss": 0.58,
+      "step": 40162
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8767563625087267e-07,
+      "loss": 0.663,
+      "step": 40163
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.874791048555493e-07,
+      "loss": 0.5895,
+      "step": 40164
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8728262230425915e-07,
+      "loss": 0.6295,
+      "step": 40165
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.870861885980015e-07,
+      "loss": 0.5909,
+      "step": 40166
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.868898037377755e-07,
+      "loss": 0.6574,
+      "step": 40167
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.866934677245759e-07,
+      "loss": 0.6285,
+      "step": 40168
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.864971805594031e-07,
+      "loss": 0.6201,
+      "step": 40169
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8630094224325287e-07,
+      "loss": 0.5771,
+      "step": 40170
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.861047527771233e-07,
+      "loss": 0.6739,
+      "step": 40171
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.859086121620104e-07,
+      "loss": 0.6239,
+      "step": 40172
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8571252039891426e-07,
+      "loss": 0.5769,
+      "step": 40173
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8551647748882537e-07,
+      "loss": 0.6088,
+      "step": 40174
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8532048343274396e-07,
+      "loss": 0.5465,
+      "step": 40175
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8512453823166483e-07,
+      "loss": 0.5894,
+      "step": 40176
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.84928641886585e-07,
+      "loss": 0.6605,
+      "step": 40177
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.84732794398498e-07,
+      "loss": 0.6024,
+      "step": 40178
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8453699576839973e-07,
+      "loss": 0.5864,
+      "step": 40179
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8434124599728286e-07,
+      "loss": 0.6018,
+      "step": 40180
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.841455450861464e-07,
+      "loss": 0.6233,
+      "step": 40181
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8394989303598305e-07,
+      "loss": 0.6192,
+      "step": 40182
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.837542898477853e-07,
+      "loss": 0.5974,
+      "step": 40183
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8355873552254896e-07,
+      "loss": 0.6086,
+      "step": 40184
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.833632300612666e-07,
+      "loss": 0.6086,
+      "step": 40185
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8316777346493197e-07,
+      "loss": 0.6256,
+      "step": 40186
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8297236573453744e-07,
+      "loss": 0.5662,
+      "step": 40187
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.8277700687107787e-07,
+      "loss": 0.5811,
+      "step": 40188
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.825816968755447e-07,
+      "loss": 0.6126,
+      "step": 40189
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.823864357489315e-07,
+      "loss": 0.6271,
+      "step": 40190
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.821912234922276e-07,
+      "loss": 0.6451,
+      "step": 40191
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.8199606010642763e-07,
+      "loss": 0.5635,
+      "step": 40192
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.81800945592522e-07,
+      "loss": 0.5811,
+      "step": 40193
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.816058799515043e-07,
+      "loss": 0.6887,
+      "step": 40194
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.814108631843616e-07,
+      "loss": 0.5709,
+      "step": 40195
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.8121589529208857e-07,
+      "loss": 0.5862,
+      "step": 40196
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.810209762756745e-07,
+      "loss": 0.5783,
+      "step": 40197
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.808261061361096e-07,
+      "loss": 0.6323,
+      "step": 40198
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.806312848743854e-07,
+      "loss": 0.5517,
+      "step": 40199
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.8043651249148995e-07,
+      "loss": 0.612,
+      "step": 40200
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.802417889884125e-07,
+      "loss": 0.5227,
+      "step": 40201
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.800471143661455e-07,
+      "loss": 0.592,
+      "step": 40202
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.798524886256771e-07,
+      "loss": 0.6166,
+      "step": 40203
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.796579117679933e-07,
+      "loss": 0.6323,
+      "step": 40204
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7946338379408754e-07,
+      "loss": 0.629,
+      "step": 40205
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.792689047049447e-07,
+      "loss": 0.5547,
+      "step": 40206
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.79074474501554e-07,
+      "loss": 0.6254,
+      "step": 40207
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7888009318490235e-07,
+      "loss": 0.5274,
+      "step": 40208
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7868576075598017e-07,
+      "loss": 0.6052,
+      "step": 40209
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.784914772157722e-07,
+      "loss": 0.6267,
+      "step": 40210
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7829724256526976e-07,
+      "loss": 0.5733,
+      "step": 40211
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.781030568054533e-07,
+      "loss": 0.5708,
+      "step": 40212
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7790891993731535e-07,
+      "loss": 0.5986,
+      "step": 40213
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7771483196183843e-07,
+      "loss": 0.5911,
+      "step": 40214
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.775207928800118e-07,
+      "loss": 0.608,
+      "step": 40215
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7732680269282006e-07,
+      "loss": 0.5967,
+      "step": 40216
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.771328614012493e-07,
+      "loss": 0.5956,
+      "step": 40217
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7693896900628303e-07,
+      "loss": 0.6734,
+      "step": 40218
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.767451255089094e-07,
+      "loss": 0.5848,
+      "step": 40219
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.765513309101121e-07,
+      "loss": 0.5903,
+      "step": 40220
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7635758521087583e-07,
+      "loss": 0.5939,
+      "step": 40221
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.761638884121832e-07,
+      "loss": 0.6342,
+      "step": 40222
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.759702405150223e-07,
+      "loss": 0.5752,
+      "step": 40223
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7577664152037563e-07,
+      "loss": 0.601,
+      "step": 40224
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7558309142922465e-07,
+      "loss": 0.625,
+      "step": 40225
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7538959024255527e-07,
+      "loss": 0.6212,
+      "step": 40226
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7519613796135115e-07,
+      "loss": 0.6207,
+      "step": 40227
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7500273458659364e-07,
+      "loss": 0.6146,
+      "step": 40228
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.748093801192643e-07,
+      "loss": 0.5852,
+      "step": 40229
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7461607456035e-07,
+      "loss": 0.6144,
+      "step": 40230
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.744228179108278e-07,
+      "loss": 0.5783,
+      "step": 40231
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.742296101716858e-07,
+      "loss": 0.5854,
+      "step": 40232
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.740364513438988e-07,
+      "loss": 0.5924,
+      "step": 40233
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7384334142845257e-07,
+      "loss": 0.6193,
+      "step": 40234
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7365028042632754e-07,
+      "loss": 0.5839,
+      "step": 40235
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.734572683385052e-07,
+      "loss": 0.6233,
+      "step": 40236
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7326430516596567e-07,
+      "loss": 0.5414,
+      "step": 40237
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7307139090968946e-07,
+      "loss": 0.595,
+      "step": 40238
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.728785255706557e-07,
+      "loss": 0.6075,
+      "step": 40239
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7268570914984705e-07,
+      "loss": 0.6251,
+      "step": 40240
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7249294164824146e-07,
+      "loss": 0.6087,
+      "step": 40241
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7230022306681823e-07,
+      "loss": 0.5886,
+      "step": 40242
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.721075534065588e-07,
+      "loss": 0.6191,
+      "step": 40243
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.71914932668439e-07,
+      "loss": 0.5733,
+      "step": 40244
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7172236085343925e-07,
+      "loss": 0.6577,
+      "step": 40245
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.715298379625376e-07,
+      "loss": 0.5999,
+      "step": 40246
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7133736399671437e-07,
+      "loss": 0.5768,
+      "step": 40247
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7114493895694436e-07,
+      "loss": 0.5782,
+      "step": 40248
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.709525628442079e-07,
+      "loss": 0.6441,
+      "step": 40249
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.707602356594797e-07,
+      "loss": 0.6062,
+      "step": 40250
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.705679574037402e-07,
+      "loss": 0.5767,
+      "step": 40251
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.7037572807796406e-07,
+      "loss": 0.623,
+      "step": 40252
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.701835476831306e-07,
+      "loss": 0.5946,
+      "step": 40253
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.699914162202134e-07,
+      "loss": 0.5694,
+      "step": 40254
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.697993336901917e-07,
+      "loss": 0.6445,
+      "step": 40255
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.696073000940381e-07,
+      "loss": 0.5924,
+      "step": 40256
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.694153154327307e-07,
+      "loss": 0.6053,
+      "step": 40257
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.692233797072453e-07,
+      "loss": 0.6268,
+      "step": 40258
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6903149291855564e-07,
+      "loss": 0.5879,
+      "step": 40259
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.688396550676376e-07,
+      "loss": 0.5536,
+      "step": 40260
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.68647866155466e-07,
+      "loss": 0.5907,
+      "step": 40261
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6845612618301553e-07,
+      "loss": 0.5839,
+      "step": 40262
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6826443515125873e-07,
+      "loss": 0.5536,
+      "step": 40263
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.680727930611727e-07,
+      "loss": 0.5264,
+      "step": 40264
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6788119991372995e-07,
+      "loss": 0.619,
+      "step": 40265
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.67689655709903e-07,
+      "loss": 0.6208,
+      "step": 40266
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6749816045066554e-07,
+      "loss": 0.5152,
+      "step": 40267
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6730671413699124e-07,
+      "loss": 0.6004,
+      "step": 40268
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.671153167698527e-07,
+      "loss": 0.5957,
+      "step": 40269
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6692396835022345e-07,
+      "loss": 0.5778,
+      "step": 40270
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6673266887907286e-07,
+      "loss": 0.5539,
+      "step": 40271
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6654141835737565e-07,
+      "loss": 0.6066,
+      "step": 40272
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6635021678610215e-07,
+      "loss": 0.6494,
+      "step": 40273
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.661590641662249e-07,
+      "loss": 0.622,
+      "step": 40274
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.659679604987154e-07,
+      "loss": 0.6045,
+      "step": 40275
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6577690578454397e-07,
+      "loss": 0.5967,
+      "step": 40276
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.655859000246809e-07,
+      "loss": 0.631,
+      "step": 40277
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.653949432200976e-07,
+      "loss": 0.6164,
+      "step": 40278
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.652040353717656e-07,
+      "loss": 0.5652,
+      "step": 40279
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.650131764806519e-07,
+      "loss": 0.5815,
+      "step": 40280
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6482236654772794e-07,
+      "loss": 0.6418,
+      "step": 40281
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6463160557396514e-07,
+      "loss": 0.5774,
+      "step": 40282
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.644408935603294e-07,
+      "loss": 0.5882,
+      "step": 40283
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6425023050779105e-07,
+      "loss": 0.6431,
+      "step": 40284
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6405961641732046e-07,
+      "loss": 0.535,
+      "step": 40285
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.638690512898857e-07,
+      "loss": 0.5896,
+      "step": 40286
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6367853512645267e-07,
+      "loss": 0.647,
+      "step": 40287
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6348806792799173e-07,
+      "loss": 0.5847,
+      "step": 40288
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6329764969546987e-07,
+      "loss": 0.6447,
+      "step": 40289
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.631072804298541e-07,
+      "loss": 0.6066,
+      "step": 40290
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6291696013211364e-07,
+      "loss": 0.5759,
+      "step": 40291
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6272668880321437e-07,
+      "loss": 0.5725,
+      "step": 40292
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.625364664441244e-07,
+      "loss": 0.5831,
+      "step": 40293
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6234629305580635e-07,
+      "loss": 0.5854,
+      "step": 40294
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6215616863923166e-07,
+      "loss": 0.5414,
+      "step": 40295
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.619660931953628e-07,
+      "loss": 0.6019,
+      "step": 40296
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6177606672516796e-07,
+      "loss": 0.6167,
+      "step": 40297
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.615860892296097e-07,
+      "loss": 0.5924,
+      "step": 40298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.613961607096572e-07,
+      "loss": 0.6465,
+      "step": 40299
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.612062811662731e-07,
+      "loss": 0.6209,
+      "step": 40300
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6101645060042214e-07,
+      "loss": 0.6297,
+      "step": 40301
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6082666901306905e-07,
+      "loss": 0.6286,
+      "step": 40302
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6063693640517983e-07,
+      "loss": 0.6018,
+      "step": 40303
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.604472527777181e-07,
+      "loss": 0.6082,
+      "step": 40304
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.602576181316442e-07,
+      "loss": 0.5442,
+      "step": 40305
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6006803246792734e-07,
+      "loss": 0.5815,
+      "step": 40306
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5987849578752566e-07,
+      "loss": 0.5608,
+      "step": 40307
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5968900809140726e-07,
+      "loss": 0.6539,
+      "step": 40308
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.594995693805292e-07,
+      "loss": 0.6197,
+      "step": 40309
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.593101796558596e-07,
+      "loss": 0.5856,
+      "step": 40310
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.591208389183565e-07,
+      "loss": 0.5744,
+      "step": 40311
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.589315471689847e-07,
+      "loss": 0.6183,
+      "step": 40312
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.587423044087057e-07,
+      "loss": 0.6105,
+      "step": 40313
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.58553110638481e-07,
+      "loss": 0.5901,
+      "step": 40314
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.583639658592697e-07,
+      "loss": 0.6054,
+      "step": 40315
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.581748700720367e-07,
+      "loss": 0.5587,
+      "step": 40316
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5798582327774e-07,
+      "loss": 0.5915,
+      "step": 40317
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.577968254773401e-07,
+      "loss": 0.5766,
+      "step": 40318
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5760787667180053e-07,
+      "loss": 0.5805,
+      "step": 40319
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.574189768620784e-07,
+      "loss": 0.6155,
+      "step": 40320
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.57230126049134e-07,
+      "loss": 0.6669,
+      "step": 40321
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.570413242339277e-07,
+      "loss": 0.6664,
+      "step": 40322
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.568525714174187e-07,
+      "loss": 0.6076,
+      "step": 40323
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5666386760056517e-07,
+      "loss": 0.6336,
+      "step": 40324
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.564752127843285e-07,
+      "loss": 0.5946,
+      "step": 40325
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5628660696966245e-07,
+      "loss": 0.6064,
+      "step": 40326
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5609805015753064e-07,
+      "loss": 0.6152,
+      "step": 40327
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5590954234888784e-07,
+      "loss": 0.5871,
+      "step": 40328
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5572108354469446e-07,
+      "loss": 0.578,
+      "step": 40329
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.555326737459064e-07,
+      "loss": 0.6054,
+      "step": 40330
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.553443129534806e-07,
+      "loss": 0.6654,
+      "step": 40331
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.5515600116837523e-07,
+      "loss": 0.6096,
+      "step": 40332
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.549677383915462e-07,
+      "loss": 0.6102,
+      "step": 40333
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.547795246239527e-07,
+      "loss": 0.5913,
+      "step": 40334
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.545913598665474e-07,
+      "loss": 0.6454,
+      "step": 40335
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5440324412028827e-07,
+      "loss": 0.6234,
+      "step": 40336
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5421517738613133e-07,
+      "loss": 0.5816,
+      "step": 40337
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.540271596650324e-07,
+      "loss": 0.605,
+      "step": 40338
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.538391909579464e-07,
+      "loss": 0.5892,
+      "step": 40339
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.53651271265828e-07,
+      "loss": 0.6204,
+      "step": 40340
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.534634005896331e-07,
+      "loss": 0.6307,
+      "step": 40341
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.532755789303155e-07,
+      "loss": 0.5889,
+      "step": 40342
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.530878062888299e-07,
+      "loss": 0.557,
+      "step": 40343
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5290008266612997e-07,
+      "loss": 0.5358,
+      "step": 40344
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5271240806316943e-07,
+      "loss": 0.5726,
+      "step": 40345
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5252478248090525e-07,
+      "loss": 0.6188,
+      "step": 40346
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5233720592028564e-07,
+      "loss": 0.5568,
+      "step": 40347
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5214967838226643e-07,
+      "loss": 0.5915,
+      "step": 40348
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.519621998678002e-07,
+      "loss": 0.6072,
+      "step": 40349
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5177477037784067e-07,
+      "loss": 0.5923,
+      "step": 40350
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.515873899133382e-07,
+      "loss": 0.6332,
+      "step": 40351
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.514000584752464e-07,
+      "loss": 0.6286,
+      "step": 40352
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5121277606451563e-07,
+      "loss": 0.5761,
+      "step": 40353
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5102554268209966e-07,
+      "loss": 0.6698,
+      "step": 40354
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5083835832894764e-07,
+      "loss": 0.5872,
+      "step": 40355
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5065122300601216e-07,
+      "loss": 0.606,
+      "step": 40356
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5046413671424475e-07,
+      "loss": 0.6084,
+      "step": 40357
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5027709945459453e-07,
+      "loss": 0.6661,
+      "step": 40358
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.5009011122801196e-07,
+      "loss": 0.6049,
+      "step": 40359
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.499031720354473e-07,
+      "loss": 0.5996,
+      "step": 40360
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4971628187785213e-07,
+      "loss": 0.5866,
+      "step": 40361
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4952944075617444e-07,
+      "loss": 0.6001,
+      "step": 40362
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.493426486713636e-07,
+      "loss": 0.6097,
+      "step": 40363
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4915590562436763e-07,
+      "loss": 0.5559,
+      "step": 40364
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4896921161613915e-07,
+      "loss": 0.6003,
+      "step": 40365
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.487825666476219e-07,
+      "loss": 0.5614,
+      "step": 40366
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4859597071977057e-07,
+      "loss": 0.5823,
+      "step": 40367
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.484094238335267e-07,
+      "loss": 0.5863,
+      "step": 40368
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4822292598984286e-07,
+      "loss": 0.6232,
+      "step": 40369
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.480364771896638e-07,
+      "loss": 0.6006,
+      "step": 40370
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4785007743393884e-07,
+      "loss": 0.643,
+      "step": 40371
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.476637267236127e-07,
+      "loss": 0.5947,
+      "step": 40372
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.474774250596369e-07,
+      "loss": 0.5884,
+      "step": 40373
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4729117244295286e-07,
+      "loss": 0.6114,
+      "step": 40374
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4710496887451096e-07,
+      "loss": 0.6023,
+      "step": 40375
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4691881435525485e-07,
+      "loss": 0.5612,
+      "step": 40376
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.467327088861305e-07,
+      "loss": 0.6223,
+      "step": 40377
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4654665246808604e-07,
+      "loss": 0.5535,
+      "step": 40378
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.46360645102064e-07,
+      "loss": 0.6159,
+      "step": 40379
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4617468678901144e-07,
+      "loss": 0.5951,
+      "step": 40380
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4598877752987205e-07,
+      "loss": 0.5777,
+      "step": 40381
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4580291732559055e-07,
+      "loss": 0.6367,
+      "step": 40382
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4561710617711187e-07,
+      "loss": 0.5934,
+      "step": 40383
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.454313440853818e-07,
+      "loss": 0.607,
+      "step": 40384
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.452456310513408e-07,
+      "loss": 0.6372,
+      "step": 40385
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.450599670759347e-07,
+      "loss": 0.6356,
+      "step": 40386
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.448743521601061e-07,
+      "loss": 0.6037,
+      "step": 40387
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4468878630479987e-07,
+      "loss": 0.5987,
+      "step": 40388
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4450326951095626e-07,
+      "loss": 0.6475,
+      "step": 40389
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.443178017795201e-07,
+      "loss": 0.6894,
+      "step": 40390
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.441323831114318e-07,
+      "loss": 0.6561,
+      "step": 40391
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4394701350763613e-07,
+      "loss": 0.6612,
+      "step": 40392
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4376169296907234e-07,
+      "loss": 0.6173,
+      "step": 40393
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4357642149668413e-07,
+      "loss": 0.6149,
+      "step": 40394
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.433911990914129e-07,
+      "loss": 0.5623,
+      "step": 40395
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.43206025754198e-07,
+      "loss": 0.5966,
+      "step": 40396
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.43020901485982e-07,
+      "loss": 0.5544,
+      "step": 40397
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.428358262877041e-07,
+      "loss": 0.6031,
+      "step": 40398
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.426508001603057e-07,
+      "loss": 0.5634,
+      "step": 40399
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4246582310472843e-07,
+      "loss": 0.6126,
+      "step": 40400
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.422808951219092e-07,
+      "loss": 0.5933,
+      "step": 40401
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.420960162127884e-07,
+      "loss": 0.5912,
+      "step": 40402
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4191118637830756e-07,
+      "loss": 0.5914,
+      "step": 40403
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.417264056194036e-07,
+      "loss": 0.5749,
+      "step": 40404
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.415416739370181e-07,
+      "loss": 0.5717,
+      "step": 40405
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4135699133208575e-07,
+      "loss": 0.6094,
+      "step": 40406
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4117235780554925e-07,
+      "loss": 0.5885,
+      "step": 40407
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.409877733583422e-07,
+      "loss": 0.6598,
+      "step": 40408
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4080323799140725e-07,
+      "loss": 0.5555,
+      "step": 40409
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4061875170567804e-07,
+      "loss": 0.6107,
+      "step": 40410
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4043431450209605e-07,
+      "loss": 0.5667,
+      "step": 40411
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4024992638159506e-07,
+      "loss": 0.5867,
+      "step": 40412
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.4006558734511307e-07,
+      "loss": 0.6219,
+      "step": 40413
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3988129739358613e-07,
+      "loss": 0.5857,
+      "step": 40414
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.396970565279534e-07,
+      "loss": 0.5951,
+      "step": 40415
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3951286474914745e-07,
+      "loss": 0.6187,
+      "step": 40416
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3932872205810654e-07,
+      "loss": 0.5619,
+      "step": 40417
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3914462845576647e-07,
+      "loss": 0.5363,
+      "step": 40418
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3896058394305986e-07,
+      "loss": 0.5756,
+      "step": 40419
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3877658852092487e-07,
+      "loss": 0.655,
+      "step": 40420
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.385926421902952e-07,
+      "loss": 0.5684,
+      "step": 40421
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3840874495210783e-07,
+      "loss": 0.6314,
+      "step": 40422
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.382248968072921e-07,
+      "loss": 0.6118,
+      "step": 40423
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.380410977567872e-07,
+      "loss": 0.5908,
+      "step": 40424
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3785734780152345e-07,
+      "loss": 0.5808,
+      "step": 40425
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.376736469424369e-07,
+      "loss": 0.6501,
+      "step": 40426
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.374899951804611e-07,
+      "loss": 0.6301,
+      "step": 40427
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.373063925165287e-07,
+      "loss": 0.5972,
+      "step": 40428
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.371228389515713e-07,
+      "loss": 0.6184,
+      "step": 40429
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.369393344865224e-07,
+      "loss": 0.6378,
+      "step": 40430
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3675587912231467e-07,
+      "loss": 0.6146,
+      "step": 40431
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3657247285988294e-07,
+      "loss": 0.5792,
+      "step": 40432
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3638911570015307e-07,
+      "loss": 0.5726,
+      "step": 40433
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.362058076440622e-07,
+      "loss": 0.607,
+      "step": 40434
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.360225486925395e-07,
+      "loss": 0.5992,
+      "step": 40435
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.358393388465153e-07,
+      "loss": 0.6237,
+      "step": 40436
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3565617810692344e-07,
+      "loss": 0.5782,
+      "step": 40437
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.354730664746919e-07,
+      "loss": 0.574,
+      "step": 40438
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.352900039507534e-07,
+      "loss": 0.6575,
+      "step": 40439
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3510699053603493e-07,
+      "loss": 0.6,
+      "step": 40440
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3492402623146904e-07,
+      "loss": 0.6261,
+      "step": 40441
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.347411110379839e-07,
+      "loss": 0.5877,
+      "step": 40442
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3455824495651215e-07,
+      "loss": 0.5797,
+      "step": 40443
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3437542798797853e-07,
+      "loss": 0.5573,
+      "step": 40444
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3419266013331453e-07,
+      "loss": 0.5919,
+      "step": 40445
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3400994139344835e-07,
+      "loss": 0.6602,
+      "step": 40446
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3382727176930806e-07,
+      "loss": 0.6219,
+      "step": 40447
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.33644651261823e-07,
+      "loss": 0.6103,
+      "step": 40448
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3346207987192126e-07,
+      "loss": 0.5897,
+      "step": 40449
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.332795576005277e-07,
+      "loss": 0.6365,
+      "step": 40450
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.330970844485726e-07,
+      "loss": 0.5481,
+      "step": 40451
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3291466041698193e-07,
+      "loss": 0.5826,
+      "step": 40452
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.327322855066839e-07,
+      "loss": 0.5913,
+      "step": 40453
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.325499597186044e-07,
+      "loss": 0.5364,
+      "step": 40454
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.323676830536693e-07,
+      "loss": 0.5976,
+      "step": 40455
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.321854555128057e-07,
+      "loss": 0.6186,
+      "step": 40456
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3200327709693835e-07,
+      "loss": 0.5966,
+      "step": 40457
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.318211478069944e-07,
+      "loss": 0.6606,
+      "step": 40458
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3163906764389854e-07,
+      "loss": 0.5864,
+      "step": 40459
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3145703660857673e-07,
+      "loss": 0.593,
+      "step": 40460
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3127505470195165e-07,
+      "loss": 0.6449,
+      "step": 40461
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3109312192495027e-07,
+      "loss": 0.5921,
+      "step": 40462
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3091123827849626e-07,
+      "loss": 0.6149,
+      "step": 40463
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.307294037635145e-07,
+      "loss": 0.6066,
+      "step": 40464
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3054761838092866e-07,
+      "loss": 0.5655,
+      "step": 40465
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3036588213166243e-07,
+      "loss": 0.6061,
+      "step": 40466
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.3018419501663735e-07,
+      "loss": 0.6178,
+      "step": 40467
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.300025570367793e-07,
+      "loss": 0.5725,
+      "step": 40468
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.298209681930098e-07,
+      "loss": 0.6635,
+      "step": 40469
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2963942848625363e-07,
+      "loss": 0.5569,
+      "step": 40470
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.294579379174301e-07,
+      "loss": 0.6019,
+      "step": 40471
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2927649648746396e-07,
+      "loss": 0.5291,
+      "step": 40472
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2909510419727453e-07,
+      "loss": 0.6127,
+      "step": 40473
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.289137610477866e-07,
+      "loss": 0.6461,
+      "step": 40474
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2873246703991945e-07,
+      "loss": 0.5942,
+      "step": 40475
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2855122217459567e-07,
+      "loss": 0.6275,
+      "step": 40476
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.283700264527356e-07,
+      "loss": 0.5766,
+      "step": 40477
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.281888798752586e-07,
+      "loss": 0.5394,
+      "step": 40478
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2800778244308715e-07,
+      "loss": 0.6122,
+      "step": 40479
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.278267341571406e-07,
+      "loss": 0.6155,
+      "step": 40480
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.276457350183404e-07,
+      "loss": 0.5825,
+      "step": 40481
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.2746478502760246e-07,
+      "loss": 0.5778,
+      "step": 40482
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.272838841858506e-07,
+      "loss": 0.5878,
+      "step": 40483
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2710303249400057e-07,
+      "loss": 0.6624,
+      "step": 40484
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2692222995297396e-07,
+      "loss": 0.6353,
+      "step": 40485
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.267414765636889e-07,
+      "loss": 0.5612,
+      "step": 40486
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2656077232706253e-07,
+      "loss": 0.6072,
+      "step": 40487
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.26380117244014e-07,
+      "loss": 0.6331,
+      "step": 40488
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2619951131546146e-07,
+      "loss": 0.6739,
+      "step": 40489
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.260189545423209e-07,
+      "loss": 0.6096,
+      "step": 40490
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.258384469255127e-07,
+      "loss": 0.6135,
+      "step": 40491
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2565798846595274e-07,
+      "loss": 0.599,
+      "step": 40492
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2547757916455704e-07,
+      "loss": 0.5541,
+      "step": 40493
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2529721902224256e-07,
+      "loss": 0.5774,
+      "step": 40494
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2511690803992744e-07,
+      "loss": 0.644,
+      "step": 40495
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.249366462185266e-07,
+      "loss": 0.6603,
+      "step": 40496
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.247564335589559e-07,
+      "loss": 0.5269,
+      "step": 40497
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2457627006213133e-07,
+      "loss": 0.6062,
+      "step": 40498
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2439615572896767e-07,
+      "loss": 0.6086,
+      "step": 40499
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2421609056038083e-07,
+      "loss": 0.5933,
+      "step": 40500
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.240360745572857e-07,
+      "loss": 0.5604,
+      "step": 40501
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2385610772059814e-07,
+      "loss": 0.5588,
+      "step": 40502
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.236761900512309e-07,
+      "loss": 0.6343,
+      "step": 40503
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2349632155009856e-07,
+      "loss": 0.5431,
+      "step": 40504
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2331650221811504e-07,
+      "loss": 0.6278,
+      "step": 40505
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.231367320561951e-07,
+      "loss": 0.608,
+      "step": 40506
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.229570110652502e-07,
+      "loss": 0.5604,
+      "step": 40507
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2277733924619745e-07,
+      "loss": 0.5738,
+      "step": 40508
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2259771659994387e-07,
+      "loss": 0.6065,
+      "step": 40509
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.224181431274076e-07,
+      "loss": 0.5945,
+      "step": 40510
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.222386188294979e-07,
+      "loss": 0.5963,
+      "step": 40511
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2205914370712854e-07,
+      "loss": 0.5972,
+      "step": 40512
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.21879717761211e-07,
+      "loss": 0.6259,
+      "step": 40513
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.217003409926567e-07,
+      "loss": 0.6824,
+      "step": 40514
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.215210134023772e-07,
+      "loss": 0.5802,
+      "step": 40515
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2134173499128396e-07,
+      "loss": 0.67,
+      "step": 40516
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.211625057602885e-07,
+      "loss": 0.6209,
+      "step": 40517
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.209833257103001e-07,
+      "loss": 0.5546,
+      "step": 40518
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.208041948422325e-07,
+      "loss": 0.5857,
+      "step": 40519
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.206251131569904e-07,
+      "loss": 0.6153,
+      "step": 40520
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.204460806554877e-07,
+      "loss": 0.5785,
+      "step": 40521
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.202670973386335e-07,
+      "loss": 0.5942,
+      "step": 40522
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.2008816320733716e-07,
+      "loss": 0.6108,
+      "step": 40523
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1990927826250794e-07,
+      "loss": 0.6,
+      "step": 40524
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.19730442505054e-07,
+      "loss": 0.6069,
+      "step": 40525
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.195516559358847e-07,
+      "loss": 0.582,
+      "step": 40526
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1937291855590913e-07,
+      "loss": 0.5872,
+      "step": 40527
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1919423036603336e-07,
+      "loss": 0.5486,
+      "step": 40528
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.190155913671689e-07,
+      "loss": 0.5826,
+      "step": 40529
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.188370015602205e-07,
+      "loss": 0.5493,
+      "step": 40530
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.186584609460963e-07,
+      "loss": 0.599,
+      "step": 40531
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.184799695257035e-07,
+      "loss": 0.5805,
+      "step": 40532
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.183015272999501e-07,
+      "loss": 0.6052,
+      "step": 40533
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.181231342697422e-07,
+      "loss": 0.5729,
+      "step": 40534
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.179447904359867e-07,
+      "loss": 0.6079,
+      "step": 40535
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.177664957995885e-07,
+      "loss": 0.5914,
+      "step": 40536
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1758825036145355e-07,
+      "loss": 0.6509,
+      "step": 40537
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.174100541224889e-07,
+      "loss": 0.5832,
+      "step": 40538
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.172319070835983e-07,
+      "loss": 0.6101,
+      "step": 40539
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1705380924568984e-07,
+      "loss": 0.5887,
+      "step": 40540
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.168757606096662e-07,
+      "loss": 0.637,
+      "step": 40541
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1669776117643326e-07,
+      "loss": 0.5919,
+      "step": 40542
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1651981094689366e-07,
+      "loss": 0.5995,
+      "step": 40543
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1634190992195336e-07,
+      "loss": 0.6307,
+      "step": 40544
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.161640581025149e-07,
+      "loss": 0.5975,
+      "step": 40545
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.159862554894855e-07,
+      "loss": 0.6303,
+      "step": 40546
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1580850208376424e-07,
+      "loss": 0.6156,
+      "step": 40547
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1563079788625603e-07,
+      "loss": 0.6265,
+      "step": 40548
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1545314289786355e-07,
+      "loss": 0.5992,
+      "step": 40549
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.152755371194904e-07,
+      "loss": 0.616,
+      "step": 40550
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.150979805520393e-07,
+      "loss": 0.6231,
+      "step": 40551
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1492047319641173e-07,
+      "loss": 0.5639,
+      "step": 40552
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1474301505350803e-07,
+      "loss": 0.6203,
+      "step": 40553
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1456560612423304e-07,
+      "loss": 0.5112,
+      "step": 40554
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1438824640948605e-07,
+      "loss": 0.5912,
+      "step": 40555
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.142109359101686e-07,
+      "loss": 0.5031,
+      "step": 40556
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.140336746271844e-07,
+      "loss": 0.5728,
+      "step": 40557
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.138564625614282e-07,
+      "loss": 0.5653,
+      "step": 40558
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.136792997138061e-07,
+      "loss": 0.6466,
+      "step": 40559
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.13502186085215e-07,
+      "loss": 0.6291,
+      "step": 40560
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.133251216765576e-07,
+      "loss": 0.6193,
+      "step": 40561
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1314810648873206e-07,
+      "loss": 0.5949,
+      "step": 40562
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.129711405226377e-07,
+      "loss": 0.6174,
+      "step": 40563
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.127942237791726e-07,
+      "loss": 0.6547,
+      "step": 40564
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.126173562592383e-07,
+      "loss": 0.5874,
+      "step": 40565
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1244053796373074e-07,
+      "loss": 0.5494,
+      "step": 40566
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.122637688935526e-07,
+      "loss": 0.6502,
+      "step": 40567
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.120870490495975e-07,
+      "loss": 0.5712,
+      "step": 40568
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1191037843276707e-07,
+      "loss": 0.6321,
+      "step": 40569
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1173375704395494e-07,
+      "loss": 0.5878,
+      "step": 40570
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.115571848840626e-07,
+      "loss": 0.6381,
+      "step": 40571
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.11380661953985e-07,
+      "loss": 0.5566,
+      "step": 40572
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.112041882546202e-07,
+      "loss": 0.6072,
+      "step": 40573
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1102776378686416e-07,
+      "loss": 0.6216,
+      "step": 40574
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.108513885516129e-07,
+      "loss": 0.6145,
+      "step": 40575
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.106750625497645e-07,
+      "loss": 0.6641,
+      "step": 40576
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.104987857822128e-07,
+      "loss": 0.5674,
+      "step": 40577
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.1032255824985593e-07,
+      "loss": 0.5855,
+      "step": 40578
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.101463799535864e-07,
+      "loss": 0.5716,
+      "step": 40579
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0997025089430143e-07,
+      "loss": 0.5746,
+      "step": 40580
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0979417107289466e-07,
+      "loss": 0.5572,
+      "step": 40581
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0961814049026316e-07,
+      "loss": 0.5823,
+      "step": 40582
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.094421591472974e-07,
+      "loss": 0.6092,
+      "step": 40583
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.092662270448976e-07,
+      "loss": 0.655,
+      "step": 40584
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.09090344183951e-07,
+      "loss": 0.5733,
+      "step": 40585
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0891451056535683e-07,
+      "loss": 0.5402,
+      "step": 40586
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0873872619000435e-07,
+      "loss": 0.6222,
+      "step": 40587
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.085629910587895e-07,
+      "loss": 0.5956,
+      "step": 40588
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0838730517260386e-07,
+      "loss": 0.6286,
+      "step": 40589
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.082116685323422e-07,
+      "loss": 0.6118,
+      "step": 40590
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.080360811388927e-07,
+      "loss": 0.6264,
+      "step": 40591
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.078605429931525e-07,
+      "loss": 0.6122,
+      "step": 40592
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.076850540960108e-07,
+      "loss": 0.5309,
+      "step": 40593
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0750961444835916e-07,
+      "loss": 0.5667,
+      "step": 40594
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.073342240510901e-07,
+      "loss": 0.5898,
+      "step": 40595
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0715888290509534e-07,
+      "loss": 0.6749,
+      "step": 40596
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.069835910112651e-07,
+      "loss": 0.6074,
+      "step": 40597
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.068083483704887e-07,
+      "loss": 0.5708,
+      "step": 40598
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0663315498365766e-07,
+      "loss": 0.5615,
+      "step": 40599
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.064580108516635e-07,
+      "loss": 0.6536,
+      "step": 40600
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.062829159753955e-07,
+      "loss": 0.6046,
+      "step": 40601
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0610787035574074e-07,
+      "loss": 0.616,
+      "step": 40602
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0593287399359297e-07,
+      "loss": 0.5651,
+      "step": 40603
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.057579268898381e-07,
+      "loss": 0.6067,
+      "step": 40604
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0558302904536655e-07,
+      "loss": 0.5886,
+      "step": 40605
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0540818046106756e-07,
+      "loss": 0.6244,
+      "step": 40606
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.052333811378283e-07,
+      "loss": 0.6275,
+      "step": 40607
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.050586310765369e-07,
+      "loss": 0.6071,
+      "step": 40608
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.048839302780826e-07,
+      "loss": 0.6251,
+      "step": 40609
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.047092787433525e-07,
+      "loss": 0.5425,
+      "step": 40610
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.045346764732349e-07,
+      "loss": 0.599,
+      "step": 40611
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0436012346861333e-07,
+      "loss": 0.5722,
+      "step": 40612
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0418561973037943e-07,
+      "loss": 0.6088,
+      "step": 40613
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.04011165259418e-07,
+      "loss": 0.5764,
+      "step": 40614
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.038367600566139e-07,
+      "loss": 0.5869,
+      "step": 40615
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.036624041228575e-07,
+      "loss": 0.602,
+      "step": 40616
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.034880974590293e-07,
+      "loss": 0.5981,
+      "step": 40617
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.033138400660196e-07,
+      "loss": 0.6043,
+      "step": 40618
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.031396319447111e-07,
+      "loss": 0.6228,
+      "step": 40619
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0296547309599076e-07,
+      "loss": 0.5594,
+      "step": 40620
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0279136352074135e-07,
+      "loss": 0.6296,
+      "step": 40621
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0261730321985206e-07,
+      "loss": 0.586,
+      "step": 40622
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.024432921942011e-07,
+      "loss": 0.6276,
+      "step": 40623
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0226933044467665e-07,
+      "loss": 0.5842,
+      "step": 40624
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0209541797216246e-07,
+      "loss": 0.6221,
+      "step": 40625
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0192155477754117e-07,
+      "loss": 0.5168,
+      "step": 40626
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0174774086169754e-07,
+      "loss": 0.5765,
+      "step": 40627
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.015739762255132e-07,
+      "loss": 0.5833,
+      "step": 40628
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.0140026086987073e-07,
+      "loss": 0.5555,
+      "step": 40629
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.012265947956561e-07,
+      "loss": 0.6152,
+      "step": 40630
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.0105297800374856e-07,
+      "loss": 0.6214,
+      "step": 40631
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.008794104950297e-07,
+      "loss": 0.5884,
+      "step": 40632
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.007058922703854e-07,
+      "loss": 0.6667,
+      "step": 40633
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.005324233306939e-07,
+      "loss": 0.6135,
+      "step": 40634
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.003590036768389e-07,
+      "loss": 0.6176,
+      "step": 40635
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.001856333096986e-07,
+      "loss": 0.5983,
+      "step": 40636
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.0001231223015793e-07,
+      "loss": 0.602,
+      "step": 40637
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9983904043909386e-07,
+      "loss": 0.5787,
+      "step": 40638
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9966581793739014e-07,
+      "loss": 0.6055,
+      "step": 40639
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9949264472592275e-07,
+      "loss": 0.6315,
+      "step": 40640
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.993195208055766e-07,
+      "loss": 0.6298,
+      "step": 40641
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9914644617722645e-07,
+      "loss": 0.5493,
+      "step": 40642
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.989734208417572e-07,
+      "loss": 0.5838,
+      "step": 40643
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9880044480004255e-07,
+      "loss": 0.6176,
+      "step": 40644
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.986275180529652e-07,
+      "loss": 0.6162,
+      "step": 40645
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9845464060140215e-07,
+      "loss": 0.6347,
+      "step": 40646
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.982818124462328e-07,
+      "loss": 0.5904,
+      "step": 40647
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.981090335883352e-07,
+      "loss": 0.6021,
+      "step": 40648
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9793630402858653e-07,
+      "loss": 0.6094,
+      "step": 40649
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9776362376786385e-07,
+      "loss": 0.5539,
+      "step": 40650
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.975909928070464e-07,
+      "loss": 0.6002,
+      "step": 40651
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9741841114701133e-07,
+      "loss": 0.6109,
+      "step": 40652
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9724587878863343e-07,
+      "loss": 0.599,
+      "step": 40653
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9707339573279095e-07,
+      "loss": 0.5826,
+      "step": 40654
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.96900961980362e-07,
+      "loss": 0.6169,
+      "step": 40655
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.967285775322193e-07,
+      "loss": 0.5869,
+      "step": 40656
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9655624238923983e-07,
+      "loss": 0.5884,
+      "step": 40657
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.963839565523019e-07,
+      "loss": 0.5745,
+      "step": 40658
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9621172002227694e-07,
+      "loss": 0.5911,
+      "step": 40659
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9603953280004426e-07,
+      "loss": 0.6427,
+      "step": 40660
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9586739488647433e-07,
+      "loss": 0.5888,
+      "step": 40661
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9569530628244637e-07,
+      "loss": 0.6167,
+      "step": 40662
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9552326698883085e-07,
+      "loss": 0.5997,
+      "step": 40663
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9535127700650477e-07,
+      "loss": 0.6404,
+      "step": 40664
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.951793363363409e-07,
+      "loss": 0.5455,
+      "step": 40665
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9500744497921285e-07,
+      "loss": 0.5648,
+      "step": 40666
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.948356029359933e-07,
+      "loss": 0.5803,
+      "step": 40667
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.946638102075572e-07,
+      "loss": 0.604,
+      "step": 40668
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9449206679477706e-07,
+      "loss": 0.5735,
+      "step": 40669
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9432037269852444e-07,
+      "loss": 0.6657,
+      "step": 40670
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9414872791967196e-07,
+      "loss": 0.5604,
+      "step": 40671
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.939771324590934e-07,
+      "loss": 0.6422,
+      "step": 40672
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.938055863176592e-07,
+      "loss": 0.6623,
+      "step": 40673
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9363408949624084e-07,
+      "loss": 0.5756,
+      "step": 40674
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9346264199571093e-07,
+      "loss": 0.6452,
+      "step": 40675
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9329124381693994e-07,
+      "loss": 0.6295,
+      "step": 40676
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.931198949607994e-07,
+      "loss": 0.5683,
+      "step": 40677
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9294859542815744e-07,
+      "loss": 0.625,
+      "step": 40678
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9277734521988897e-07,
+      "loss": 0.6125,
+      "step": 40679
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9260614433686e-07,
+      "loss": 0.6114,
+      "step": 40680
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9243499277994416e-07,
+      "loss": 0.6456,
+      "step": 40681
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9226389055000637e-07,
+      "loss": 0.5839,
+      "step": 40682
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.920928376479204e-07,
+      "loss": 0.578,
+      "step": 40683
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.919218340745533e-07,
+      "loss": 0.5976,
+      "step": 40684
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9175087983077443e-07,
+      "loss": 0.643,
+      "step": 40685
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.91579974917453e-07,
+      "loss": 0.6254,
+      "step": 40686
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9140911933545736e-07,
+      "loss": 0.6772,
+      "step": 40687
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9123831308565443e-07,
+      "loss": 0.5846,
+      "step": 40688
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9106755616891356e-07,
+      "loss": 0.5933,
+      "step": 40689
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.908968485861019e-07,
+      "loss": 0.5205,
+      "step": 40690
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9072619033808537e-07,
+      "loss": 0.5965,
+      "step": 40691
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9055558142573436e-07,
+      "loss": 0.5752,
+      "step": 40692
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.903850218499138e-07,
+      "loss": 0.6051,
+      "step": 40693
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.902145116114907e-07,
+      "loss": 0.6644,
+      "step": 40694
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.9004405071133e-07,
+      "loss": 0.5844,
+      "step": 40695
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8987363915029986e-07,
+      "loss": 0.5723,
+      "step": 40696
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8970327692926405e-07,
+      "loss": 0.5356,
+      "step": 40697
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8953296404909295e-07,
+      "loss": 0.6044,
+      "step": 40698
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.893627005106459e-07,
+      "loss": 0.6636,
+      "step": 40699
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8919248631479215e-07,
+      "loss": 0.5803,
+      "step": 40700
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.890223214623944e-07,
+      "loss": 0.6236,
+      "step": 40701
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.888522059543186e-07,
+      "loss": 0.6095,
+      "step": 40702
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8868213979142967e-07,
+      "loss": 0.5624,
+      "step": 40703
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.885121229745902e-07,
+      "loss": 0.5874,
+      "step": 40704
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8834215550466396e-07,
+      "loss": 0.6596,
+      "step": 40705
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.881722373825169e-07,
+      "loss": 0.5872,
+      "step": 40706
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8800236860900943e-07,
+      "loss": 0.6313,
+      "step": 40707
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.878325491850087e-07,
+      "loss": 0.6091,
+      "step": 40708
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8766277911137286e-07,
+      "loss": 0.5812,
+      "step": 40709
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.874930583889679e-07,
+      "loss": 0.6304,
+      "step": 40710
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8732338701865427e-07,
+      "loss": 0.6395,
+      "step": 40711
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8715376500129456e-07,
+      "loss": 0.5832,
+      "step": 40712
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8698419233775254e-07,
+      "loss": 0.5893,
+      "step": 40713
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8681466902888865e-07,
+      "loss": 0.6131,
+      "step": 40714
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.866451950755633e-07,
+      "loss": 0.5962,
+      "step": 40715
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8647577047863806e-07,
+      "loss": 0.6087,
+      "step": 40716
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.863063952389744e-07,
+      "loss": 0.5704,
+      "step": 40717
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.861370693574317e-07,
+      "loss": 0.6028,
+      "step": 40718
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8596779283487474e-07,
+      "loss": 0.5584,
+      "step": 40719
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8579856567215734e-07,
+      "loss": 0.5978,
+      "step": 40720
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.856293878701433e-07,
+      "loss": 0.617,
+      "step": 40721
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.854602594296896e-07,
+      "loss": 0.6044,
+      "step": 40722
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8529118035166005e-07,
+      "loss": 0.6097,
+      "step": 40723
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.851221506369084e-07,
+      "loss": 0.6735,
+      "step": 40724
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8495317028629955e-07,
+      "loss": 0.6376,
+      "step": 40725
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.847842393006861e-07,
+      "loss": 0.5785,
+      "step": 40726
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8461535768093075e-07,
+      "loss": 0.6423,
+      "step": 40727
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.844465254278894e-07,
+      "loss": 0.6192,
+      "step": 40728
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8427774254242033e-07,
+      "loss": 0.5699,
+      "step": 40729
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8410900902538174e-07,
+      "loss": 0.6148,
+      "step": 40730
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.839403248776307e-07,
+      "loss": 0.5817,
+      "step": 40731
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.837716901000254e-07,
+      "loss": 0.5908,
+      "step": 40732
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8360310469341956e-07,
+      "loss": 0.6468,
+      "step": 40733
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8343456865867256e-07,
+      "loss": 0.5829,
+      "step": 40734
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.832660819966415e-07,
+      "loss": 0.5674,
+      "step": 40735
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.830976447081801e-07,
+      "loss": 0.6102,
+      "step": 40736
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.829292567941444e-07,
+      "loss": 0.6061,
+      "step": 40737
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.827609182553925e-07,
+      "loss": 0.6335,
+      "step": 40738
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8259262909277716e-07,
+      "loss": 0.652,
+      "step": 40739
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.824243893071554e-07,
+      "loss": 0.6534,
+      "step": 40740
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.82256198899381e-07,
+      "loss": 0.6316,
+      "step": 40741
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.820880578703089e-07,
+      "loss": 0.6274,
+      "step": 40742
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8191996622079275e-07,
+      "loss": 0.6123,
+      "step": 40743
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.817519239516875e-07,
+      "loss": 0.568,
+      "step": 40744
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8158393106384683e-07,
+      "loss": 0.623,
+      "step": 40745
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.814159875581257e-07,
+      "loss": 0.5603,
+      "step": 40746
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8124809343537454e-07,
+      "loss": 0.6438,
+      "step": 40747
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8108024869644926e-07,
+      "loss": 0.6276,
+      "step": 40748
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.809124533422003e-07,
+      "loss": 0.6225,
+      "step": 40749
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.807447073734815e-07,
+      "loss": 0.5742,
+      "step": 40750
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.805770107911465e-07,
+      "loss": 0.6116,
+      "step": 40751
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.804093635960459e-07,
+      "loss": 0.6198,
+      "step": 40752
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.802417657890322e-07,
+      "loss": 0.6367,
+      "step": 40753
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8007421737095473e-07,
+      "loss": 0.6032,
+      "step": 40754
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.799067183426685e-07,
+      "loss": 0.5495,
+      "step": 40755
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7973926870502155e-07,
+      "loss": 0.6285,
+      "step": 40756
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7957186845886883e-07,
+      "loss": 0.6686,
+      "step": 40757
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.794045176050553e-07,
+      "loss": 0.5891,
+      "step": 40758
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7923721614443565e-07,
+      "loss": 0.5673,
+      "step": 40759
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7906996407785715e-07,
+      "loss": 0.5877,
+      "step": 40760
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7890276140617236e-07,
+      "loss": 0.632,
+      "step": 40761
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.787356081302295e-07,
+      "loss": 0.6397,
+      "step": 40762
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.785685042508779e-07,
+      "loss": 0.6236,
+      "step": 40763
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7840144976896686e-07,
+      "loss": 0.6662,
+      "step": 40764
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.782344446853447e-07,
+      "loss": 0.6334,
+      "step": 40765
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.780674890008606e-07,
+      "loss": 0.6589,
+      "step": 40766
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.779005827163628e-07,
+      "loss": 0.6348,
+      "step": 40767
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.777337258327006e-07,
+      "loss": 0.6185,
+      "step": 40768
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7756691835071903e-07,
+      "loss": 0.5835,
+      "step": 40769
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7740016027126837e-07,
+      "loss": 0.5422,
+      "step": 40770
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7723345159519356e-07,
+      "loss": 0.6005,
+      "step": 40771
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7706679232334497e-07,
+      "loss": 0.6144,
+      "step": 40772
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7690018245656645e-07,
+      "loss": 0.6237,
+      "step": 40773
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.767336219957062e-07,
+      "loss": 0.5964,
+      "step": 40774
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.7656711094160793e-07,
+      "loss": 0.5706,
+      "step": 40775
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.764006492951221e-07,
+      "loss": 0.6077,
+      "step": 40776
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.762342370570903e-07,
+      "loss": 0.6381,
+      "step": 40777
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.760678742283618e-07,
+      "loss": 0.587,
+      "step": 40778
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7590156080977924e-07,
+      "loss": 0.6648,
+      "step": 40779
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.757352968021898e-07,
+      "loss": 0.595,
+      "step": 40780
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.755690822064361e-07,
+      "loss": 0.6148,
+      "step": 40781
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7540291702336406e-07,
+      "loss": 0.6093,
+      "step": 40782
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7523680125381756e-07,
+      "loss": 0.6531,
+      "step": 40783
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7507073489864256e-07,
+      "loss": 0.565,
+      "step": 40784
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.749047179586795e-07,
+      "loss": 0.5484,
+      "step": 40785
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.747387504347743e-07,
+      "loss": 0.5862,
+      "step": 40786
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7457283232776856e-07,
+      "loss": 0.6614,
+      "step": 40787
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.744069636385072e-07,
+      "loss": 0.6082,
+      "step": 40788
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7424114436783276e-07,
+      "loss": 0.6088,
+      "step": 40789
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7407537451658806e-07,
+      "loss": 0.614,
+      "step": 40790
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7390965408561455e-07,
+      "loss": 0.5838,
+      "step": 40791
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7374398307575266e-07,
+      "loss": 0.6394,
+      "step": 40792
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7357836148784733e-07,
+      "loss": 0.6064,
+      "step": 40793
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7341278932273783e-07,
+      "loss": 0.5746,
+      "step": 40794
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.732472665812691e-07,
+      "loss": 0.6416,
+      "step": 40795
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.730817932642771e-07,
+      "loss": 0.6095,
+      "step": 40796
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7291636937260667e-07,
+      "loss": 0.6243,
+      "step": 40797
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.72750994907095e-07,
+      "loss": 0.624,
+      "step": 40798
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.725856698685858e-07,
+      "loss": 0.6146,
+      "step": 40799
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7242039425791846e-07,
+      "loss": 0.554,
+      "step": 40800
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7225516807593113e-07,
+      "loss": 0.5895,
+      "step": 40801
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7208999132346316e-07,
+      "loss": 0.6135,
+      "step": 40802
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7192486400135497e-07,
+      "loss": 0.6366,
+      "step": 40803
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7175978611044596e-07,
+      "loss": 0.604,
+      "step": 40804
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.715947576515754e-07,
+      "loss": 0.6569,
+      "step": 40805
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.714297786255804e-07,
+      "loss": 0.571,
+      "step": 40806
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7126484903330033e-07,
+      "loss": 0.5869,
+      "step": 40807
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.710999688755722e-07,
+      "loss": 0.6101,
+      "step": 40808
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7093513815323324e-07,
+      "loss": 0.5935,
+      "step": 40809
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7077035686712383e-07,
+      "loss": 0.7035,
+      "step": 40810
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7060562501807883e-07,
+      "loss": 0.577,
+      "step": 40811
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.7044094260693654e-07,
+      "loss": 0.5808,
+      "step": 40812
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.702763096345329e-07,
+      "loss": 0.644,
+      "step": 40813
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.70111726101705e-07,
+      "loss": 0.59,
+      "step": 40814
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6994719200928775e-07,
+      "loss": 0.5529,
+      "step": 40815
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.697827073581205e-07,
+      "loss": 0.5897,
+      "step": 40816
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.696182721490359e-07,
+      "loss": 0.6011,
+      "step": 40817
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6945388638287105e-07,
+      "loss": 0.6045,
+      "step": 40818
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.692895500604586e-07,
+      "loss": 0.5998,
+      "step": 40819
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.69125263182638e-07,
+      "loss": 0.607,
+      "step": 40820
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6896102575024065e-07,
+      "loss": 0.646,
+      "step": 40821
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.687968377641037e-07,
+      "loss": 0.5898,
+      "step": 40822
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6863269922505765e-07,
+      "loss": 0.6413,
+      "step": 40823
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6846861013394064e-07,
+      "loss": 0.6248,
+      "step": 40824
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.683045704915832e-07,
+      "loss": 0.6567,
+      "step": 40825
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6814058029882126e-07,
+      "loss": 0.612,
+      "step": 40826
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6797663955648755e-07,
+      "loss": 0.5993,
+      "step": 40827
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.678127482654147e-07,
+      "loss": 0.6041,
+      "step": 40828
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.676489064264343e-07,
+      "loss": 0.6512,
+      "step": 40829
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6748511404038e-07,
+      "loss": 0.597,
+      "step": 40830
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.673213711080858e-07,
+      "loss": 0.6047,
+      "step": 40831
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.671576776303797e-07,
+      "loss": 0.571,
+      "step": 40832
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6699403360809893e-07,
+      "loss": 0.6183,
+      "step": 40833
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.668304390420695e-07,
+      "loss": 0.5913,
+      "step": 40834
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6666689393312627e-07,
+      "loss": 0.6068,
+      "step": 40835
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.665033982820975e-07,
+      "loss": 0.6456,
+      "step": 40836
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.663399520898169e-07,
+      "loss": 0.5775,
+      "step": 40837
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.661765553571138e-07,
+      "loss": 0.5827,
+      "step": 40838
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6601320808481767e-07,
+      "loss": 0.5781,
+      "step": 40839
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6584991027375883e-07,
+      "loss": 0.6159,
+      "step": 40840
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6568666192476887e-07,
+      "loss": 0.5543,
+      "step": 40841
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.655234630386738e-07,
+      "loss": 0.6277,
+      "step": 40842
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.653603136163063e-07,
+      "loss": 0.6448,
+      "step": 40843
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.651972136584935e-07,
+      "loss": 0.6289,
+      "step": 40844
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.650341631660658e-07,
+      "loss": 0.6701,
+      "step": 40845
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.648711621398481e-07,
+      "loss": 0.59,
+      "step": 40846
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6470821058067307e-07,
+      "loss": 0.596,
+      "step": 40847
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6454530848936566e-07,
+      "loss": 0.544,
+      "step": 40848
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6438245586675627e-07,
+      "loss": 0.5382,
+      "step": 40849
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.642196527136698e-07,
+      "loss": 0.6133,
+      "step": 40850
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6405689903093335e-07,
+      "loss": 0.6456,
+      "step": 40851
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.638941948193763e-07,
+      "loss": 0.6055,
+      "step": 40852
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.637315400798235e-07,
+      "loss": 0.6195,
+      "step": 40853
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.635689348131043e-07,
+      "loss": 0.6313,
+      "step": 40854
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.634063790200403e-07,
+      "loss": 0.5826,
+      "step": 40855
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.632438727014619e-07,
+      "loss": 0.6071,
+      "step": 40856
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.630814158581907e-07,
+      "loss": 0.5767,
+      "step": 40857
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.62919008491056e-07,
+      "loss": 0.6333,
+      "step": 40858
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.627566506008805e-07,
+      "loss": 0.615,
+      "step": 40859
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6259434218849246e-07,
+      "loss": 0.5922,
+      "step": 40860
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6243208325471116e-07,
+      "loss": 0.5369,
+      "step": 40861
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.622698738003659e-07,
+      "loss": 0.5668,
+      "step": 40862
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6210771382627844e-07,
+      "loss": 0.5935,
+      "step": 40863
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6194560333327456e-07,
+      "loss": 0.5965,
+      "step": 40864
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6178354232217707e-07,
+      "loss": 0.6343,
+      "step": 40865
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.616215307938086e-07,
+      "loss": 0.6182,
+      "step": 40866
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6145956874899294e-07,
+      "loss": 0.6057,
+      "step": 40867
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6129765618855383e-07,
+      "loss": 0.5894,
+      "step": 40868
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6113579311331406e-07,
+      "loss": 0.6312,
+      "step": 40869
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.60973979524094e-07,
+      "loss": 0.6288,
+      "step": 40870
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6081221542171966e-07,
+      "loss": 0.5825,
+      "step": 40871
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6065050080700816e-07,
+      "loss": 0.6488,
+      "step": 40872
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6048883568078445e-07,
+      "loss": 0.6191,
+      "step": 40873
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.603272200438689e-07,
+      "loss": 0.5385,
+      "step": 40874
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6016565389708427e-07,
+      "loss": 0.5784,
+      "step": 40875
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.6000413724125094e-07,
+      "loss": 0.6034,
+      "step": 40876
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5984267007718834e-07,
+      "loss": 0.6186,
+      "step": 40877
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5968125240571687e-07,
+      "loss": 0.5285,
+      "step": 40878
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5951988422765804e-07,
+      "loss": 0.6181,
+      "step": 40879
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.593585655438302e-07,
+      "loss": 0.6053,
+      "step": 40880
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5919729635505597e-07,
+      "loss": 0.6323,
+      "step": 40881
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5903607666215247e-07,
+      "loss": 0.5789,
+      "step": 40882
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5887490646594017e-07,
+      "loss": 0.6201,
+      "step": 40883
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.587137857672362e-07,
+      "loss": 0.5646,
+      "step": 40884
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.585527145668609e-07,
+      "loss": 0.6388,
+      "step": 40885
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.583916928656316e-07,
+      "loss": 0.598,
+      "step": 40886
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.582307206643675e-07,
+      "loss": 0.6386,
+      "step": 40887
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5806979796388573e-07,
+      "loss": 0.5767,
+      "step": 40888
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.579089247650046e-07,
+      "loss": 0.6056,
+      "step": 40889
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.577481010685423e-07,
+      "loss": 0.5808,
+      "step": 40890
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5758732687531263e-07,
+      "loss": 0.6548,
+      "step": 40891
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5742660218613825e-07,
+      "loss": 0.5947,
+      "step": 40892
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.572659270018296e-07,
+      "loss": 0.6192,
+      "step": 40893
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5710530132320835e-07,
+      "loss": 0.5932,
+      "step": 40894
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.569447251510859e-07,
+      "loss": 0.5578,
+      "step": 40895
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5678419848628287e-07,
+      "loss": 0.5586,
+      "step": 40896
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.566237213296108e-07,
+      "loss": 0.6235,
+      "step": 40897
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5646329368189005e-07,
+      "loss": 0.5735,
+      "step": 40898
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5630291554393007e-07,
+      "loss": 0.5773,
+      "step": 40899
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.56142586916549e-07,
+      "loss": 0.6152,
+      "step": 40900
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5598230780056186e-07,
+      "loss": 0.5574,
+      "step": 40901
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.558220781967813e-07,
+      "loss": 0.635,
+      "step": 40902
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.556618981060244e-07,
+      "loss": 0.5789,
+      "step": 40903
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.555017675291016e-07,
+      "loss": 0.6212,
+      "step": 40904
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.553416864668279e-07,
+      "loss": 0.5357,
+      "step": 40905
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5518165492001813e-07,
+      "loss": 0.5701,
+      "step": 40906
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5502167288948385e-07,
+      "loss": 0.6213,
+      "step": 40907
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5486174037603674e-07,
+      "loss": 0.6205,
+      "step": 40908
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.547018573804938e-07,
+      "loss": 0.58,
+      "step": 40909
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5454202390366333e-07,
+      "loss": 0.5989,
+      "step": 40910
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5438223994635916e-07,
+      "loss": 0.5827,
+      "step": 40911
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5422250550939274e-07,
+      "loss": 0.6172,
+      "step": 40912
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.540628205935758e-07,
+      "loss": 0.6001,
+      "step": 40913
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.539031851997209e-07,
+      "loss": 0.6751,
+      "step": 40914
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.537435993286375e-07,
+      "loss": 0.6189,
+      "step": 40915
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5358406298113705e-07,
+      "loss": 0.593,
+      "step": 40916
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5342457615803007e-07,
+      "loss": 0.5453,
+      "step": 40917
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5326513886012705e-07,
+      "loss": 0.5971,
+      "step": 40918
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.531057510882406e-07,
+      "loss": 0.6328,
+      "step": 40919
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.529464128431758e-07,
+      "loss": 0.5884,
+      "step": 40920
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5278712412574625e-07,
+      "loss": 0.6445,
+      "step": 40921
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.5262788493675915e-07,
+      "loss": 0.5703,
+      "step": 40922
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.52468695277025e-07,
+      "loss": 0.6189,
+      "step": 40923
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.52309555147352e-07,
+      "loss": 0.5813,
+      "step": 40924
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.521504645485484e-07,
+      "loss": 0.6141,
+      "step": 40925
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5199142348142245e-07,
+      "loss": 0.5923,
+      "step": 40926
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5183243194678466e-07,
+      "loss": 0.5479,
+      "step": 40927
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5167348994544095e-07,
+      "loss": 0.6012,
+      "step": 40928
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.515145974781974e-07,
+      "loss": 0.578,
+      "step": 40929
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5135575454586556e-07,
+      "loss": 0.6188,
+      "step": 40930
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5119696114924817e-07,
+      "loss": 0.5901,
+      "step": 40931
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5103821728915565e-07,
+      "loss": 0.586,
+      "step": 40932
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5087952296639074e-07,
+      "loss": 0.589,
+      "step": 40933
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.507208781817638e-07,
+      "loss": 0.5924,
+      "step": 40934
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.505622829360788e-07,
+      "loss": 0.5563,
+      "step": 40935
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.504037372301438e-07,
+      "loss": 0.6153,
+      "step": 40936
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5024524106476044e-07,
+      "loss": 0.5659,
+      "step": 40937
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.5008679444073814e-07,
+      "loss": 0.5672,
+      "step": 40938
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.499283973588784e-07,
+      "loss": 0.5923,
+      "step": 40939
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.497700498199884e-07,
+      "loss": 0.6028,
+      "step": 40940
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4961175182487305e-07,
+      "loss": 0.6052,
+      "step": 40941
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4945350337433615e-07,
+      "loss": 0.5653,
+      "step": 40942
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.492953044691804e-07,
+      "loss": 0.6119,
+      "step": 40943
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4913715511021177e-07,
+      "loss": 0.5573,
+      "step": 40944
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4897905529823295e-07,
+      "loss": 0.6325,
+      "step": 40945
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.488210050340467e-07,
+      "loss": 0.6052,
+      "step": 40946
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.486630043184568e-07,
+      "loss": 0.5955,
+      "step": 40947
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.485050531522659e-07,
+      "loss": 0.6416,
+      "step": 40948
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4834715153627785e-07,
+      "loss": 0.5782,
+      "step": 40949
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.48189299471292e-07,
+      "loss": 0.6441,
+      "step": 40950
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.480314969581132e-07,
+      "loss": 0.6262,
+      "step": 40951
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4787374399754317e-07,
+      "loss": 0.6306,
+      "step": 40952
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4771604059038115e-07,
+      "loss": 0.6368,
+      "step": 40953
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4755838673743096e-07,
+      "loss": 0.6044,
+      "step": 40954
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.47400782439492e-07,
+      "loss": 0.6532,
+      "step": 40955
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4724322769736576e-07,
+      "loss": 0.6114,
+      "step": 40956
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4708572251185504e-07,
+      "loss": 0.6498,
+      "step": 40957
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.469282668837558e-07,
+      "loss": 0.5768,
+      "step": 40958
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.467708608138708e-07,
+      "loss": 0.6043,
+      "step": 40959
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4661350430299937e-07,
+      "loss": 0.5804,
+      "step": 40960
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4645619735194194e-07,
+      "loss": 0.6232,
+      "step": 40961
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.462989399614957e-07,
+      "loss": 0.5852,
+      "step": 40962
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.461417321324622e-07,
+      "loss": 0.6226,
+      "step": 40963
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4598457386563856e-07,
+      "loss": 0.582,
+      "step": 40964
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.458274651618242e-07,
+      "loss": 0.5821,
+      "step": 40965
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4567040602181734e-07,
+      "loss": 0.5508,
+      "step": 40966
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.455133964464151e-07,
+      "loss": 0.6406,
+      "step": 40967
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4535643643641695e-07,
+      "loss": 0.6076,
+      "step": 40968
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4519952599261877e-07,
+      "loss": 0.5665,
+      "step": 40969
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4504266511581885e-07,
+      "loss": 0.6011,
+      "step": 40970
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.448858538068133e-07,
+      "loss": 0.5826,
+      "step": 40971
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.447290920664014e-07,
+      "loss": 0.5597,
+      "step": 40972
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4457237989537584e-07,
+      "loss": 0.5397,
+      "step": 40973
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4441571729453717e-07,
+      "loss": 0.6228,
+      "step": 40974
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.44259104264678e-07,
+      "loss": 0.6219,
+      "step": 40975
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4410254080659557e-07,
+      "loss": 0.5818,
+      "step": 40976
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.439460269210847e-07,
+      "loss": 0.6483,
+      "step": 40977
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.437895626089426e-07,
+      "loss": 0.5666,
+      "step": 40978
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4363314787096194e-07,
+      "loss": 0.6296,
+      "step": 40979
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4347678270793985e-07,
+      "loss": 0.5543,
+      "step": 40980
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4332046712066793e-07,
+      "loss": 0.5659,
+      "step": 40981
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4316420110994224e-07,
+      "loss": 0.6104,
+      "step": 40982
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.430079846765576e-07,
+      "loss": 0.5897,
+      "step": 40983
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4285181782130687e-07,
+      "loss": 0.5853,
+      "step": 40984
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.426957005449826e-07,
+      "loss": 0.5563,
+      "step": 40985
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4253963284837977e-07,
+      "loss": 0.6035,
+      "step": 40986
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.423836147322911e-07,
+      "loss": 0.619,
+      "step": 40987
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4222764619750707e-07,
+      "loss": 0.5845,
+      "step": 40988
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4207172724482366e-07,
+      "loss": 0.6071,
+      "step": 40989
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4191585787503135e-07,
+      "loss": 0.5615,
+      "step": 40990
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.417600380889229e-07,
+      "loss": 0.6033,
+      "step": 40991
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.416042678872876e-07,
+      "loss": 0.661,
+      "step": 40992
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.414485472709205e-07,
+      "loss": 0.5432,
+      "step": 40993
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4129287624061084e-07,
+      "loss": 0.5478,
+      "step": 40994
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4113725479715133e-07,
+      "loss": 0.5666,
+      "step": 40995
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.409816829413292e-07,
+      "loss": 0.5812,
+      "step": 40996
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.408261606739382e-07,
+      "loss": 0.626,
+      "step": 40997
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4067068799576765e-07,
+      "loss": 0.5999,
+      "step": 40998
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4051526490760704e-07,
+      "loss": 0.6121,
+      "step": 40999
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.4035989141024785e-07,
+      "loss": 0.6629,
+      "step": 41000
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.402045675044773e-07,
+      "loss": 0.604,
+      "step": 41001
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.400492931910847e-07,
+      "loss": 0.5488,
+      "step": 41002
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.398940684708617e-07,
+      "loss": 0.5827,
+      "step": 41003
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3973889334459546e-07,
+      "loss": 0.6003,
+      "step": 41004
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.395837678130719e-07,
+      "loss": 0.5597,
+      "step": 41005
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3942869187708386e-07,
+      "loss": 0.6483,
+      "step": 41006
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3927366553741726e-07,
+      "loss": 0.5617,
+      "step": 41007
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.391186887948593e-07,
+      "loss": 0.6506,
+      "step": 41008
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.389637616501961e-07,
+      "loss": 0.6056,
+      "step": 41009
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3880888410421907e-07,
+      "loss": 0.5543,
+      "step": 41010
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.386540561577122e-07,
+      "loss": 0.6216,
+      "step": 41011
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3849927781146254e-07,
+      "loss": 0.6278,
+      "step": 41012
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3834454906625615e-07,
+      "loss": 0.6051,
+      "step": 41013
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.381898699228813e-07,
+      "loss": 0.5739,
+      "step": 41014
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3803524038212067e-07,
+      "loss": 0.611,
+      "step": 41015
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3788066044476366e-07,
+      "loss": 0.5774,
+      "step": 41016
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3772613011159407e-07,
+      "loss": 0.6343,
+      "step": 41017
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3757164938339684e-07,
+      "loss": 0.6306,
+      "step": 41018
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3741721826095577e-07,
+      "loss": 0.5517,
+      "step": 41019
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.37262836745058e-07,
+      "loss": 0.6336,
+      "step": 41020
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.371085048364874e-07,
+      "loss": 0.648,
+      "step": 41021
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3695422253602773e-07,
+      "loss": 0.6469,
+      "step": 41022
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3679998984446063e-07,
+      "loss": 0.6229,
+      "step": 41023
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.366458067625743e-07,
+      "loss": 0.6318,
+      "step": 41024
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3649167329114931e-07,
+      "loss": 0.5891,
+      "step": 41025
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.363375894309694e-07,
+      "loss": 0.5896,
+      "step": 41026
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3618355518281734e-07,
+      "loss": 0.6171,
+      "step": 41027
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.360295705474769e-07,
+      "loss": 0.6472,
+      "step": 41028
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3587563552572967e-07,
+      "loss": 0.5727,
+      "step": 41029
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.357217501183573e-07,
+      "loss": 0.5451,
+      "step": 41030
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3556791432614245e-07,
+      "loss": 0.5903,
+      "step": 41031
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3541412814986674e-07,
+      "loss": 0.5876,
+      "step": 41032
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.352603915903129e-07,
+      "loss": 0.6169,
+      "step": 41033
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3510670464825803e-07,
+      "loss": 0.559,
+      "step": 41034
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.349530673244882e-07,
+      "loss": 0.5773,
+      "step": 41035
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3479947961977943e-07,
+      "loss": 0.6325,
+      "step": 41036
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.346459415349156e-07,
+      "loss": 0.6327,
+      "step": 41037
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3449245307067604e-07,
+      "loss": 0.6154,
+      "step": 41038
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.343390142278401e-07,
+      "loss": 0.5859,
+      "step": 41039
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3418562500718723e-07,
+      "loss": 0.681,
+      "step": 41040
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3403228540949786e-07,
+      "loss": 0.6094,
+      "step": 41041
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3387899543555025e-07,
+      "loss": 0.5902,
+      "step": 41042
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.337257550861227e-07,
+      "loss": 0.608,
+      "step": 41043
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.335725643619957e-07,
+      "loss": 0.5634,
+      "step": 41044
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.334194232639464e-07,
+      "loss": 0.647,
+      "step": 41045
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3326633179275415e-07,
+      "loss": 0.5745,
+      "step": 41046
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3311328994919501e-07,
+      "loss": 0.5973,
+      "step": 41047
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3296029773404837e-07,
+      "loss": 0.6492,
+      "step": 41048
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3280735514809028e-07,
+      "loss": 0.61,
+      "step": 41049
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3265446219209897e-07,
+      "loss": 0.6012,
+      "step": 41050
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.325016188668494e-07,
+      "loss": 0.6206,
+      "step": 41051
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3234882517312097e-07,
+      "loss": 0.5917,
+      "step": 41052
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3219608111168857e-07,
+      "loss": 0.5896,
+      "step": 41053
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3204338668332827e-07,
+      "loss": 0.6137,
+      "step": 41054
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.318907418888161e-07,
+      "loss": 0.6129,
+      "step": 41055
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3173814672892924e-07,
+      "loss": 0.6457,
+      "step": 41056
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.315856012044393e-07,
+      "loss": 0.6114,
+      "step": 41057
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3143310531612563e-07,
+      "loss": 0.6233,
+      "step": 41058
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3128065906475982e-07,
+      "loss": 0.6415,
+      "step": 41059
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3112826245111908e-07,
+      "loss": 0.5927,
+      "step": 41060
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3097591547597608e-07,
+      "loss": 0.618,
+      "step": 41061
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3082361814010578e-07,
+      "loss": 0.5716,
+      "step": 41062
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3067137044428088e-07,
+      "loss": 0.605,
+      "step": 41063
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3051917238927635e-07,
+      "loss": 0.5539,
+      "step": 41064
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3036702397586485e-07,
+      "loss": 0.6127,
+      "step": 41065
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3021492520482026e-07,
+      "loss": 0.5461,
+      "step": 41066
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.3006287607691413e-07,
+      "loss": 0.6304,
+      "step": 41067
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.2991087659291922e-07,
+      "loss": 0.6349,
+      "step": 41068
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.2975892675360933e-07,
+      "loss": 0.6225,
+      "step": 41069
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.29607026559755e-07,
+      "loss": 0.6093,
+      "step": 41070
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2945517601213106e-07,
+      "loss": 0.6817,
+      "step": 41071
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.293033751115048e-07,
+      "loss": 0.5695,
+      "step": 41072
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2915162385864997e-07,
+      "loss": 0.5982,
+      "step": 41073
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2899992225433597e-07,
+      "loss": 0.5995,
+      "step": 41074
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2884827029933664e-07,
+      "loss": 0.6004,
+      "step": 41075
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2869666799442137e-07,
+      "loss": 0.5985,
+      "step": 41076
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2854511534035839e-07,
+      "loss": 0.5988,
+      "step": 41077
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2839361233791934e-07,
+      "loss": 0.6161,
+      "step": 41078
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2824215898787582e-07,
+      "loss": 0.6186,
+      "step": 41079
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.280907552909939e-07,
+      "loss": 0.6111,
+      "step": 41080
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2793940124804514e-07,
+      "loss": 0.6166,
+      "step": 41081
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2778809685979896e-07,
+      "loss": 0.6424,
+      "step": 41082
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2763684212702254e-07,
+      "loss": 0.6066,
+      "step": 41083
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2748563705048632e-07,
+      "loss": 0.5705,
+      "step": 41084
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.273344816309564e-07,
+      "loss": 0.6061,
+      "step": 41085
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2718337586920214e-07,
+      "loss": 0.6596,
+      "step": 41086
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2703231976599184e-07,
+      "loss": 0.6344,
+      "step": 41087
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2688131332209262e-07,
+      "loss": 0.6282,
+      "step": 41088
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2673035653827058e-07,
+      "loss": 0.6153,
+      "step": 41089
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.265794494152951e-07,
+      "loss": 0.6792,
+      "step": 41090
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2642859195392996e-07,
+      "loss": 0.572,
+      "step": 41091
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2627778415494572e-07,
+      "loss": 0.6088,
+      "step": 41092
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2612702601910508e-07,
+      "loss": 0.5805,
+      "step": 41093
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.259763175471763e-07,
+      "loss": 0.621,
+      "step": 41094
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2582565873992323e-07,
+      "loss": 0.6247,
+      "step": 41095
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2567504959811303e-07,
+      "loss": 0.6171,
+      "step": 41096
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2552449012250953e-07,
+      "loss": 0.5878,
+      "step": 41097
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2537398031388103e-07,
+      "loss": 0.6261,
+      "step": 41098
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.25223520172988e-07,
+      "loss": 0.5547,
+      "step": 41099
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.250731097005976e-07,
+      "loss": 0.5588,
+      "step": 41100
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2492274889747256e-07,
+      "loss": 0.6312,
+      "step": 41101
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2477243776437785e-07,
+      "loss": 0.6145,
+      "step": 41102
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2462217630207838e-07,
+      "loss": 0.6096,
+      "step": 41103
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.244719645113369e-07,
+      "loss": 0.6646,
+      "step": 41104
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.24321802392915e-07,
+      "loss": 0.6092,
+      "step": 41105
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2417168994757654e-07,
+      "loss": 0.5885,
+      "step": 41106
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2402162717608644e-07,
+      "loss": 0.6802,
+      "step": 41107
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.23871614079203e-07,
+      "loss": 0.6357,
+      "step": 41108
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2372165065769337e-07,
+      "loss": 0.5801,
+      "step": 41109
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2357173691231582e-07,
+      "loss": 0.6426,
+      "step": 41110
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2342187284383422e-07,
+      "loss": 0.5852,
+      "step": 41111
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2327205845300793e-07,
+      "loss": 0.6033,
+      "step": 41112
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.231222937406008e-07,
+      "loss": 0.6348,
+      "step": 41113
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2297257870737333e-07,
+      "loss": 0.6053,
+      "step": 41114
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.228229133540849e-07,
+      "loss": 0.5381,
+      "step": 41115
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2267329768149493e-07,
+      "loss": 0.6029,
+      "step": 41116
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2252373169036723e-07,
+      "loss": 0.5778,
+      "step": 41117
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2237421538145897e-07,
+      "loss": 0.549,
+      "step": 41118
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2222474875553068e-07,
+      "loss": 0.5647,
+      "step": 41119
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2207533181334284e-07,
+      "loss": 0.5908,
+      "step": 41120
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2192596455565375e-07,
+      "loss": 0.5763,
+      "step": 41121
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2177664698322277e-07,
+      "loss": 0.5942,
+      "step": 41122
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.216273790968071e-07,
+      "loss": 0.5834,
+      "step": 41123
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2147816089716723e-07,
+      "loss": 0.5646,
+      "step": 41124
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2132899238506145e-07,
+      "loss": 0.5642,
+      "step": 41125
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2117987356124582e-07,
+      "loss": 0.5766,
+      "step": 41126
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2103080442647862e-07,
+      "loss": 0.6139,
+      "step": 41127
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2088178498151924e-07,
+      "loss": 0.5693,
+      "step": 41128
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2073281522712152e-07,
+      "loss": 0.6057,
+      "step": 41129
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.205838951640471e-07,
+      "loss": 0.5797,
+      "step": 41130
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2043502479304758e-07,
+      "loss": 0.6126,
+      "step": 41131
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2028620411488233e-07,
+      "loss": 0.5683,
+      "step": 41132
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.2013743313030632e-07,
+      "loss": 0.5667,
+      "step": 41133
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1998871184007786e-07,
+      "loss": 0.6042,
+      "step": 41134
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1984004024494965e-07,
+      "loss": 0.607,
+      "step": 41135
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1969141834567998e-07,
+      "loss": 0.6237,
+      "step": 41136
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1954284614302045e-07,
+      "loss": 0.5816,
+      "step": 41137
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1939432363772938e-07,
+      "loss": 0.5999,
+      "step": 41138
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.192458508305595e-07,
+      "loss": 0.6376,
+      "step": 41139
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1909742772226683e-07,
+      "loss": 0.5544,
+      "step": 41140
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1894905431360415e-07,
+      "loss": 0.6019,
+      "step": 41141
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1880073060532524e-07,
+      "loss": 0.5692,
+      "step": 41142
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1865245659818513e-07,
+      "loss": 0.6746,
+      "step": 41143
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.185042322929354e-07,
+      "loss": 0.5405,
+      "step": 41144
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1835605769033209e-07,
+      "loss": 0.5281,
+      "step": 41145
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1820793279112463e-07,
+      "loss": 0.6253,
+      "step": 41146
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1805985759606907e-07,
+      "loss": 0.5894,
+      "step": 41147
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1791183210591482e-07,
+      "loss": 0.5482,
+      "step": 41148
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1776385632141573e-07,
+      "loss": 0.5722,
+      "step": 41149
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1761593024332228e-07,
+      "loss": 0.6366,
+      "step": 41150
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1746805387238833e-07,
+      "loss": 0.6083,
+      "step": 41151
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.173202272093644e-07,
+      "loss": 0.6311,
+      "step": 41152
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.171724502550021e-07,
+      "loss": 0.5754,
+      "step": 41153
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1702472301004973e-07,
+      "loss": 0.5371,
+      "step": 41154
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1687704547526113e-07,
+      "loss": 0.5643,
+      "step": 41155
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1672941765138344e-07,
+      "loss": 0.6023,
+      "step": 41156
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1658183953917056e-07,
+      "loss": 0.6056,
+      "step": 41157
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1643431113937076e-07,
+      "loss": 0.5686,
+      "step": 41158
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162868324527334e-07,
+      "loss": 0.611,
+      "step": 41159
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1613940348000683e-07,
+      "loss": 0.6296,
+      "step": 41160
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1599202422194153e-07,
+      "loss": 0.6391,
+      "step": 41161
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1584469467928693e-07,
+      "loss": 0.5432,
+      "step": 41162
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1569741485279128e-07,
+      "loss": 0.5893,
+      "step": 41163
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1555018474320177e-07,
+      "loss": 0.5693,
+      "step": 41164
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.154030043512667e-07,
+      "loss": 0.6114,
+      "step": 41165
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.152558736777355e-07,
+      "loss": 0.5859,
+      "step": 41166
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1510879272335304e-07,
+      "loss": 0.6615,
+      "step": 41167
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1496176148887104e-07,
+      "loss": 0.6093,
+      "step": 41168
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.148147799750322e-07,
+      "loss": 0.5309,
+      "step": 41169
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.146678481825859e-07,
+      "loss": 0.5647,
+      "step": 41170
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1452096611227713e-07,
+      "loss": 0.6212,
+      "step": 41171
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1437413376485417e-07,
+      "loss": 0.6957,
+      "step": 41172
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1422735114106085e-07,
+      "loss": 0.56,
+      "step": 41173
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.140806182416466e-07,
+      "loss": 0.534,
+      "step": 41174
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1393393506735306e-07,
+      "loss": 0.612,
+      "step": 41175
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1378730161892736e-07,
+      "loss": 0.586,
+      "step": 41176
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1364071789711338e-07,
+      "loss": 0.5716,
+      "step": 41177
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.134941839026583e-07,
+      "loss": 0.587,
+      "step": 41178
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1334769963630598e-07,
+      "loss": 0.5941,
+      "step": 41179
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1320126509879913e-07,
+      "loss": 0.5989,
+      "step": 41180
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1305488029088272e-07,
+      "loss": 0.5956,
+      "step": 41181
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.129085452133006e-07,
+      "loss": 0.6086,
+      "step": 41182
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1276225986679777e-07,
+      "loss": 0.6212,
+      "step": 41183
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1261602425211468e-07,
+      "loss": 0.6246,
+      "step": 41184
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1246983836999746e-07,
+      "loss": 0.624,
+      "step": 41185
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1232370222118547e-07,
+      "loss": 0.6026,
+      "step": 41186
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1217761580642483e-07,
+      "loss": 0.5607,
+      "step": 41187
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1203157912645488e-07,
+      "loss": 0.6416,
+      "step": 41188
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1188559218201954e-07,
+      "loss": 0.6106,
+      "step": 41189
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1173965497386039e-07,
+      "loss": 0.5791,
+      "step": 41190
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.115937675027191e-07,
+      "loss": 0.6384,
+      "step": 41191
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1144792976933505e-07,
+      "loss": 0.6319,
+      "step": 41192
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1130214177445207e-07,
+      "loss": 0.5688,
+      "step": 41193
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.111564035188085e-07,
+      "loss": 0.5996,
+      "step": 41194
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.110107150031482e-07,
+      "loss": 0.6379,
+      "step": 41195
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1086507622820718e-07,
+      "loss": 0.5557,
+      "step": 41196
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1071948719472934e-07,
+      "loss": 0.5901,
+      "step": 41197
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1057394790345075e-07,
+      "loss": 0.6349,
+      "step": 41198
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1042845835511417e-07,
+      "loss": 0.561,
+      "step": 41199
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1028301855045785e-07,
+      "loss": 0.5901,
+      "step": 41200
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1013762849022013e-07,
+      "loss": 0.636,
+      "step": 41201
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.099922881751404e-07,
+      "loss": 0.5914,
+      "step": 41202
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0984699760595583e-07,
+      "loss": 0.5233,
+      "step": 41203
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0970175678340698e-07,
+      "loss": 0.6617,
+      "step": 41204
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.095565657082299e-07,
+      "loss": 0.5589,
+      "step": 41205
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0941142438116514e-07,
+      "loss": 0.624,
+      "step": 41206
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.092663328029454e-07,
+      "loss": 0.5871,
+      "step": 41207
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0912129097431344e-07,
+      "loss": 0.574,
+      "step": 41208
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.089762988960009e-07,
+      "loss": 0.6108,
+      "step": 41209
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0883135656874942e-07,
+      "loss": 0.5935,
+      "step": 41210
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0868646399329172e-07,
+      "loss": 0.6074,
+      "step": 41211
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0854162117036725e-07,
+      "loss": 0.6064,
+      "step": 41212
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0839682810070983e-07,
+      "loss": 0.6417,
+      "step": 41213
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0825208478505554e-07,
+      "loss": 0.5434,
+      "step": 41214
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0810739122413936e-07,
+      "loss": 0.5829,
+      "step": 41215
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0796274741869736e-07,
+      "loss": 0.6147,
+      "step": 41216
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.0781815336946565e-07,
+      "loss": 0.5693,
+      "step": 41217
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0767360907717693e-07,
+      "loss": 0.5595,
+      "step": 41218
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0752911454256618e-07,
+      "loss": 0.5456,
+      "step": 41219
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.073846697663684e-07,
+      "loss": 0.6184,
+      "step": 41220
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.072402747493163e-07,
+      "loss": 0.5539,
+      "step": 41221
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0709592949214485e-07,
+      "loss": 0.594,
+      "step": 41222
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.069516339955868e-07,
+      "loss": 0.5819,
+      "step": 41223
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0680738826037495e-07,
+      "loss": 0.6233,
+      "step": 41224
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0666319228724418e-07,
+      "loss": 0.6021,
+      "step": 41225
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0651904607692396e-07,
+      "loss": 0.6105,
+      "step": 41226
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0637494963015037e-07,
+      "loss": 0.6313,
+      "step": 41227
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.062309029476528e-07,
+      "loss": 0.6146,
+      "step": 41228
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0608690603016513e-07,
+      "loss": 0.5936,
+      "step": 41229
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0594295887841675e-07,
+      "loss": 0.6096,
+      "step": 41230
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0579906149314267e-07,
+      "loss": 0.6419,
+      "step": 41231
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0565521387507003e-07,
+      "loss": 0.5722,
+      "step": 41232
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0551141602493275e-07,
+      "loss": 0.5938,
+      "step": 41233
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0536766794345907e-07,
+      "loss": 0.5186,
+      "step": 41234
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0522396963138292e-07,
+      "loss": 0.6548,
+      "step": 41235
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0508032108943033e-07,
+      "loss": 0.5598,
+      "step": 41236
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0493672231833407e-07,
+      "loss": 0.6705,
+      "step": 41237
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0479317331882243e-07,
+      "loss": 0.65,
+      "step": 41238
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0464967409162594e-07,
+      "loss": 0.5372,
+      "step": 41239
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0450622463747293e-07,
+      "loss": 0.6358,
+      "step": 41240
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0436282495709282e-07,
+      "loss": 0.6228,
+      "step": 41241
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0421947505121387e-07,
+      "loss": 0.6298,
+      "step": 41242
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.040761749205633e-07,
+      "loss": 0.5995,
+      "step": 41243
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0393292456587277e-07,
+      "loss": 0.6159,
+      "step": 41244
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0378972398786724e-07,
+      "loss": 0.5313,
+      "step": 41245
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.03646573187275e-07,
+      "loss": 0.6543,
+      "step": 41246
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0350347216482214e-07,
+      "loss": 0.5257,
+      "step": 41247
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.033604209212381e-07,
+      "loss": 0.6178,
+      "step": 41248
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0321741945724892e-07,
+      "loss": 0.5714,
+      "step": 41249
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0307446777358186e-07,
+      "loss": 0.5965,
+      "step": 41250
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0293156587096186e-07,
+      "loss": 0.5752,
+      "step": 41251
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.027887137501161e-07,
+      "loss": 0.6566,
+      "step": 41252
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.026459114117696e-07,
+      "loss": 0.5781,
+      "step": 41253
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.025031588566484e-07,
+      "loss": 0.5945,
+      "step": 41254
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.023604560854786e-07,
+      "loss": 0.6201,
+      "step": 41255
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.022178030989852e-07,
+      "loss": 0.6111,
+      "step": 41256
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0207519989789092e-07,
+      "loss": 0.6059,
+      "step": 41257
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.01932646482923e-07,
+      "loss": 0.6199,
+      "step": 41258
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0179014285480524e-07,
+      "loss": 0.6333,
+      "step": 41259
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0164768901426156e-07,
+      "loss": 0.6093,
+      "step": 41260
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.015052849620136e-07,
+      "loss": 0.5784,
+      "step": 41261
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0136293069878854e-07,
+      "loss": 0.5973,
+      "step": 41262
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0122062622530803e-07,
+      "loss": 0.5626,
+      "step": 41263
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.010783715422937e-07,
+      "loss": 0.5644,
+      "step": 41264
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0093616665047166e-07,
+      "loss": 0.5224,
+      "step": 41265
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0079401155056245e-07,
+      "loss": 0.6278,
+      "step": 41266
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.006519062432888e-07,
+      "loss": 0.5943,
+      "step": 41267
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0050985072937123e-07,
+      "loss": 0.5554,
+      "step": 41268
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0036784500953476e-07,
+      "loss": 0.5886,
+      "step": 41269
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.0022588908449768e-07,
+      "loss": 0.6035,
+      "step": 41270
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.00083982954985e-07,
+      "loss": 0.5343,
+      "step": 41271
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9994212662171386e-07,
+      "loss": 0.595,
+      "step": 41272
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9980032008540707e-07,
+      "loss": 0.6103,
+      "step": 41273
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9965856334678514e-07,
+      "loss": 0.6477,
+      "step": 41274
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.995168564065686e-07,
+      "loss": 0.5743,
+      "step": 41275
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9937519926547689e-07,
+      "loss": 0.5779,
+      "step": 41276
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9923359192423052e-07,
+      "loss": 0.6354,
+      "step": 41277
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9909203438354786e-07,
+      "loss": 0.6149,
+      "step": 41278
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9895052664414939e-07,
+      "loss": 0.6481,
+      "step": 41279
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9880906870675344e-07,
+      "loss": 0.61,
+      "step": 41280
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9866766057207832e-07,
+      "loss": 0.574,
+      "step": 41281
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9852630224084458e-07,
+      "loss": 0.6594,
+      "step": 41282
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.983849937137683e-07,
+      "loss": 0.5304,
+      "step": 41283
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.982437349915689e-07,
+      "loss": 0.6073,
+      "step": 41284
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.981025260749625e-07,
+      "loss": 0.6361,
+      "step": 41285
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.979613669646685e-07,
+      "loss": 0.5673,
+      "step": 41286
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.978202576614041e-07,
+      "loss": 0.6209,
+      "step": 41287
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9767919816588543e-07,
+      "loss": 0.5988,
+      "step": 41288
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9753818847882854e-07,
+      "loss": 0.6298,
+      "step": 41289
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9739722860095178e-07,
+      "loss": 0.5745,
+      "step": 41290
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9725631853297012e-07,
+      "loss": 0.5827,
+      "step": 41291
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9711545827560075e-07,
+      "loss": 0.5599,
+      "step": 41292
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9697464782955756e-07,
+      "loss": 0.5518,
+      "step": 41293
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9683388719555774e-07,
+      "loss": 0.6169,
+      "step": 41294
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9669317637431628e-07,
+      "loss": 0.6717,
+      "step": 41295
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9655251536654706e-07,
+      "loss": 0.5642,
+      "step": 41296
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.964119041729662e-07,
+      "loss": 0.6132,
+      "step": 41297
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.962713427942875e-07,
+      "loss": 0.5172,
+      "step": 41298
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.961308312312249e-07,
+      "loss": 0.5585,
+      "step": 41299
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9599036948449335e-07,
+      "loss": 0.6409,
+      "step": 41300
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9584995755480563e-07,
+      "loss": 0.5767,
+      "step": 41301
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.957095954428745e-07,
+      "loss": 0.5606,
+      "step": 41302
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9556928314941493e-07,
+      "loss": 0.633,
+      "step": 41303
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.954290206751397e-07,
+      "loss": 0.6003,
+      "step": 41304
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9528880802076156e-07,
+      "loss": 0.5731,
+      "step": 41305
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9514864518699105e-07,
+      "loss": 0.6252,
+      "step": 41306
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9500853217454206e-07,
+      "loss": 0.5895,
+      "step": 41307
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9486846898412626e-07,
+      "loss": 0.5971,
+      "step": 41308
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9472845561645637e-07,
+      "loss": 0.5787,
+      "step": 41309
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9458849207224183e-07,
+      "loss": 0.6168,
+      "step": 41310
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9444857835219543e-07,
+      "loss": 0.6309,
+      "step": 41311
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.943087144570266e-07,
+      "loss": 0.6099,
+      "step": 41312
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9416890038744696e-07,
+      "loss": 0.6505,
+      "step": 41313
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9402913614416817e-07,
+      "loss": 0.6054,
+      "step": 41314
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.938894217278986e-07,
+      "loss": 0.5933,
+      "step": 41315
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9374975713934763e-07,
+      "loss": 0.5975,
+      "step": 41316
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9361014237922693e-07,
+      "loss": 0.599,
+      "step": 41317
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9347057744824483e-07,
+      "loss": 0.6217,
+      "step": 41318
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9333106234711075e-07,
+      "loss": 0.632,
+      "step": 41319
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9319159707653413e-07,
+      "loss": 0.5233,
+      "step": 41320
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9305218163722217e-07,
+      "loss": 0.6077,
+      "step": 41321
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9291281602988544e-07,
+      "loss": 0.6299,
+      "step": 41322
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9277350025522888e-07,
+      "loss": 0.5753,
+      "step": 41323
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9263423431396423e-07,
+      "loss": 0.5375,
+      "step": 41324
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.924950182067964e-07,
+      "loss": 0.6313,
+      "step": 41325
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9235585193443375e-07,
+      "loss": 0.5536,
+      "step": 41326
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9221673549758236e-07,
+      "loss": 0.5633,
+      "step": 41327
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9207766889695166e-07,
+      "loss": 0.6283,
+      "step": 41328
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9193865213324558e-07,
+      "loss": 0.6133,
+      "step": 41329
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.917996852071724e-07,
+      "loss": 0.6085,
+      "step": 41330
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9166076811943712e-07,
+      "loss": 0.6138,
+      "step": 41331
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9152190087074695e-07,
+      "loss": 0.6428,
+      "step": 41332
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9138308346180466e-07,
+      "loss": 0.6377,
+      "step": 41333
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.912443158933186e-07,
+      "loss": 0.6161,
+      "step": 41334
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.911055981659926e-07,
+      "loss": 0.564,
+      "step": 41335
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9096693028053282e-07,
+      "loss": 0.6401,
+      "step": 41336
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9082831223764088e-07,
+      "loss": 0.6479,
+      "step": 41337
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.90689744038024e-07,
+      "loss": 0.5583,
+      "step": 41338
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9055122568238604e-07,
+      "loss": 0.5912,
+      "step": 41339
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.904127571714287e-07,
+      "loss": 0.5395,
+      "step": 41340
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9027433850585698e-07,
+      "loss": 0.6012,
+      "step": 41341
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.9013596968637583e-07,
+      "loss": 0.6055,
+      "step": 41342
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.899976507136858e-07,
+      "loss": 0.5724,
+      "step": 41343
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.898593815884897e-07,
+      "loss": 0.5935,
+      "step": 41344
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8972116231149252e-07,
+      "loss": 0.5852,
+      "step": 41345
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8958299288339476e-07,
+      "loss": 0.5984,
+      "step": 41346
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8944487330490036e-07,
+      "loss": 0.6215,
+      "step": 41347
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.893068035767076e-07,
+      "loss": 0.5574,
+      "step": 41348
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.891687836995204e-07,
+      "loss": 0.6361,
+      "step": 41349
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8903081367404042e-07,
+      "loss": 0.5727,
+      "step": 41350
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8889289350096817e-07,
+      "loss": 0.6374,
+      "step": 41351
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8875502318100425e-07,
+      "loss": 0.5821,
+      "step": 41352
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.886172027148503e-07,
+      "loss": 0.6439,
+      "step": 41353
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8847943210320462e-07,
+      "loss": 0.6251,
+      "step": 41354
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8834171134676893e-07,
+      "loss": 0.5785,
+      "step": 41355
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8820404044624263e-07,
+      "loss": 0.5534,
+      "step": 41356
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8806641940232408e-07,
+      "loss": 0.5937,
+      "step": 41357
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8792884821571377e-07,
+      "loss": 0.5681,
+      "step": 41358
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8779132688711233e-07,
+      "loss": 0.6081,
+      "step": 41359
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8765385541721582e-07,
+      "loss": 0.6411,
+      "step": 41360
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8751643380672258e-07,
+      "loss": 0.5471,
+      "step": 41361
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8737906205633316e-07,
+      "loss": 0.5927,
+      "step": 41362
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.872417401667448e-07,
+      "loss": 0.5961,
+      "step": 41363
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.8710446813865467e-07,
+      "loss": 0.5901,
+      "step": 41364
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8696724597276006e-07,
+      "loss": 0.5975,
+      "step": 41365
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8683007366975926e-07,
+      "loss": 0.6305,
+      "step": 41366
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8669295123034725e-07,
+      "loss": 0.6058,
+      "step": 41367
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8655587865522461e-07,
+      "loss": 0.6611,
+      "step": 41368
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8641885594508413e-07,
+      "loss": 0.6254,
+      "step": 41369
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8628188310062413e-07,
+      "loss": 0.6109,
+      "step": 41370
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8614496012253959e-07,
+      "loss": 0.5908,
+      "step": 41371
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8600808701152773e-07,
+      "loss": 0.5674,
+      "step": 41372
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8587126376828134e-07,
+      "loss": 0.6338,
+      "step": 41373
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8573449039349878e-07,
+      "loss": 0.559,
+      "step": 41374
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8559776688787168e-07,
+      "loss": 0.6656,
+      "step": 41375
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8546109325209838e-07,
+      "loss": 0.6131,
+      "step": 41376
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8532446948687055e-07,
+      "loss": 0.5774,
+      "step": 41377
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.851878955928832e-07,
+      "loss": 0.6039,
+      "step": 41378
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.850513715708313e-07,
+      "loss": 0.616,
+      "step": 41379
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8491489742140766e-07,
+      "loss": 0.6588,
+      "step": 41380
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8477847314530616e-07,
+      "loss": 0.643,
+      "step": 41381
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.846420987432196e-07,
+      "loss": 0.5868,
+      "step": 41382
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.845057742158407e-07,
+      "loss": 0.5556,
+      "step": 41383
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.843694995638623e-07,
+      "loss": 0.6101,
+      "step": 41384
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8423327478797825e-07,
+      "loss": 0.5864,
+      "step": 41385
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8409709988887913e-07,
+      "loss": 0.6081,
+      "step": 41386
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.839609748672577e-07,
+      "loss": 0.6078,
+      "step": 41387
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8382489972380347e-07,
+      "loss": 0.6374,
+      "step": 41388
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8368887445921136e-07,
+      "loss": 0.6448,
+      "step": 41389
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8355289907417084e-07,
+      "loss": 0.6495,
+      "step": 41390
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.834169735693736e-07,
+      "loss": 0.523,
+      "step": 41391
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8328109794550796e-07,
+      "loss": 0.6569,
+      "step": 41392
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8314527220326673e-07,
+      "loss": 0.5288,
+      "step": 41393
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8300949634334043e-07,
+      "loss": 0.62,
+      "step": 41394
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.828737703664163e-07,
+      "loss": 0.561,
+      "step": 41395
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.82738094273186e-07,
+      "loss": 0.5869,
+      "step": 41396
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8260246806433902e-07,
+      "loss": 0.6525,
+      "step": 41397
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8246689174056476e-07,
+      "loss": 0.6022,
+      "step": 41398
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8233136530254936e-07,
+      "loss": 0.5621,
+      "step": 41399
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.821958887509856e-07,
+      "loss": 0.6607,
+      "step": 41400
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8206046208655847e-07,
+      "loss": 0.5956,
+      "step": 41401
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8192508530995855e-07,
+      "loss": 0.5814,
+      "step": 41402
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8178975842187086e-07,
+      "loss": 0.588,
+      "step": 41403
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8165448142298702e-07,
+      "loss": 0.5553,
+      "step": 41404
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8151925431398987e-07,
+      "loss": 0.5892,
+      "step": 41405
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8138407709557104e-07,
+      "loss": 0.6037,
+      "step": 41406
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8124894976841446e-07,
+      "loss": 0.6094,
+      "step": 41407
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8111387233320733e-07,
+      "loss": 0.6243,
+      "step": 41408
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8097884479063576e-07,
+      "loss": 0.644,
+      "step": 41409
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8084386714138703e-07,
+      "loss": 0.5878,
+      "step": 41410
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8070893938614498e-07,
+      "loss": 0.5783,
+      "step": 41411
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8057406152559908e-07,
+      "loss": 0.5876,
+      "step": 41412
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8043923356042992e-07,
+      "loss": 0.609,
+      "step": 41413
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.803044554913258e-07,
+      "loss": 0.6159,
+      "step": 41414
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8016972731897063e-07,
+      "loss": 0.5945,
+      "step": 41415
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8003504904404722e-07,
+      "loss": 0.5948,
+      "step": 41416
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7990042066724388e-07,
+      "loss": 0.5977,
+      "step": 41417
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.797658421892412e-07,
+      "loss": 0.5878,
+      "step": 41418
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7963131361072529e-07,
+      "loss": 0.5804,
+      "step": 41419
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.794968349323778e-07,
+      "loss": 0.5953,
+      "step": 41420
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7936240615488265e-07,
+      "loss": 0.5791,
+      "step": 41421
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7922802727892375e-07,
+      "loss": 0.6405,
+      "step": 41422
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7909369830518387e-07,
+      "loss": 0.7106,
+      "step": 41423
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7895941923434468e-07,
+      "loss": 0.6079,
+      "step": 41424
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.78825190067089e-07,
+      "loss": 0.6233,
+      "step": 41425
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7869101080409844e-07,
+      "loss": 0.5929,
+      "step": 41426
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7855688144605477e-07,
+      "loss": 0.6082,
+      "step": 41427
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.784228019936418e-07,
+      "loss": 0.6161,
+      "step": 41428
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7828877244753794e-07,
+      "loss": 0.5741,
+      "step": 41429
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7815479280842374e-07,
+      "loss": 0.5723,
+      "step": 41430
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7802086307698307e-07,
+      "loss": 0.5201,
+      "step": 41431
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7788698325389432e-07,
+      "loss": 0.6205,
+      "step": 41432
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.777531533398391e-07,
+      "loss": 0.604,
+      "step": 41433
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7761937333549583e-07,
+      "loss": 0.6028,
+      "step": 41434
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7748564324154617e-07,
+      "loss": 0.654,
+      "step": 41435
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7735196305866842e-07,
+      "loss": 0.6453,
+      "step": 41436
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7721833278754096e-07,
+      "loss": 0.6104,
+      "step": 41437
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7708475242884436e-07,
+      "loss": 0.6005,
+      "step": 41438
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7695122198325698e-07,
+      "loss": 0.5627,
+      "step": 41439
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7681774145145824e-07,
+      "loss": 0.6182,
+      "step": 41440
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.766843108341232e-07,
+      "loss": 0.6747,
+      "step": 41441
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7655093013193347e-07,
+      "loss": 0.612,
+      "step": 41442
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7641759934556524e-07,
+      "loss": 0.6546,
+      "step": 41443
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7628431847569683e-07,
+      "loss": 0.5535,
+      "step": 41444
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7615108752300436e-07,
+      "loss": 0.6361,
+      "step": 41445
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7601790648816508e-07,
+      "loss": 0.5942,
+      "step": 41446
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.758847753718551e-07,
+      "loss": 0.6154,
+      "step": 41447
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.757516941747528e-07,
+      "loss": 0.5391,
+      "step": 41448
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7561866289753204e-07,
+      "loss": 0.6243,
+      "step": 41449
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.754856815408723e-07,
+      "loss": 0.602,
+      "step": 41450
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7535275010544527e-07,
+      "loss": 0.6301,
+      "step": 41451
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.752198685919293e-07,
+      "loss": 0.6564,
+      "step": 41452
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7508703700099716e-07,
+      "loss": 0.6848,
+      "step": 41453
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7495425533332612e-07,
+      "loss": 0.6117,
+      "step": 41454
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7482152358959003e-07,
+      "loss": 0.5979,
+      "step": 41455
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7468884177046285e-07,
+      "loss": 0.6256,
+      "step": 41456
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7455620987661958e-07,
+      "loss": 0.5439,
+      "step": 41457
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7442362790873302e-07,
+      "loss": 0.624,
+      "step": 41458
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7429109586747816e-07,
+      "loss": 0.5447,
+      "step": 41459
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7415861375352672e-07,
+      "loss": 0.5773,
+      "step": 41460
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.740261815675548e-07,
+      "loss": 0.6085,
+      "step": 41461
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7389379931023187e-07,
+      "loss": 0.6097,
+      "step": 41462
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7376146698223294e-07,
+      "loss": 0.5402,
+      "step": 41463
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7362918458422973e-07,
+      "loss": 0.6148,
+      "step": 41464
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7349695211689387e-07,
+      "loss": 0.6037,
+      "step": 41465
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7336476958089932e-07,
+      "loss": 0.5902,
+      "step": 41466
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.732326369769155e-07,
+      "loss": 0.6051,
+      "step": 41467
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7310055430561301e-07,
+      "loss": 0.5861,
+      "step": 41468
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7296852156766687e-07,
+      "loss": 0.5689,
+      "step": 41469
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.728365387637443e-07,
+      "loss": 0.5592,
+      "step": 41470
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.727046058945181e-07,
+      "loss": 0.6155,
+      "step": 41471
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7257272296065664e-07,
+      "loss": 0.5828,
+      "step": 41472
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.724408899628327e-07,
+      "loss": 0.6388,
+      "step": 41473
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7230910690171353e-07,
+      "loss": 0.6392,
+      "step": 41474
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7217737377797083e-07,
+      "loss": 0.5787,
+      "step": 41475
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7204569059227293e-07,
+      "loss": 0.5888,
+      "step": 41476
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.719140573452893e-07,
+      "loss": 0.6018,
+      "step": 41477
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7178247403768833e-07,
+      "loss": 0.6151,
+      "step": 41478
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7165094067013832e-07,
+      "loss": 0.6019,
+      "step": 41479
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7151945724330876e-07,
+      "loss": 0.602,
+      "step": 41480
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.713880237578669e-07,
+      "loss": 0.4933,
+      "step": 41481
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.712566402144822e-07,
+      "loss": 0.5647,
+      "step": 41482
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7112530661381965e-07,
+      "loss": 0.5785,
+      "step": 41483
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7099402295654878e-07,
+      "loss": 0.6401,
+      "step": 41484
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7086278924333455e-07,
+      "loss": 0.6299,
+      "step": 41485
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7073160547484645e-07,
+      "loss": 0.5944,
+      "step": 41486
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.706004716517484e-07,
+      "loss": 0.5852,
+      "step": 41487
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.704693877747099e-07,
+      "loss": 0.583,
+      "step": 41488
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7033835384439368e-07,
+      "loss": 0.6477,
+      "step": 41489
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7020736986146813e-07,
+      "loss": 0.5676,
+      "step": 41490
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.7007643582659606e-07,
+      "loss": 0.5897,
+      "step": 41491
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.699455517404458e-07,
+      "loss": 0.5666,
+      "step": 41492
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6981471760368128e-07,
+      "loss": 0.6012,
+      "step": 41493
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6968393341696644e-07,
+      "loss": 0.5879,
+      "step": 41494
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6955319918096736e-07,
+      "loss": 0.5792,
+      "step": 41495
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6942251489634575e-07,
+      "loss": 0.612,
+      "step": 41496
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.692918805637689e-07,
+      "loss": 0.6735,
+      "step": 41497
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6916129618389843e-07,
+      "loss": 0.6177,
+      "step": 41498
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6903076175739829e-07,
+      "loss": 0.5263,
+      "step": 41499
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.689002772849313e-07,
+      "loss": 0.6004,
+      "step": 41500
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6876984276716246e-07,
+      "loss": 0.6279,
+      "step": 41501
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6863945820475235e-07,
+      "loss": 0.5814,
+      "step": 41502
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6850912359836492e-07,
+      "loss": 0.6188,
+      "step": 41503
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6837883894866292e-07,
+      "loss": 0.5222,
+      "step": 41504
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6824860425630585e-07,
+      "loss": 0.5863,
+      "step": 41505
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6811841952195762e-07,
+      "loss": 0.6201,
+      "step": 41506
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.679882847462788e-07,
+      "loss": 0.6017,
+      "step": 41507
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6785819992993114e-07,
+      "loss": 0.6229,
+      "step": 41508
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6772816507357627e-07,
+      "loss": 0.6003,
+      "step": 41509
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6759818017787256e-07,
+      "loss": 0.5901,
+      "step": 41510
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.6746824524348281e-07,
+      "loss": 0.6232,
+      "step": 41511
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6733836027106541e-07,
+      "loss": 0.6394,
+      "step": 41512
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6720852526128316e-07,
+      "loss": 0.634,
+      "step": 41513
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.670787402147933e-07,
+      "loss": 0.6262,
+      "step": 41514
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6694900513225643e-07,
+      "loss": 0.5945,
+      "step": 41515
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6681932001433199e-07,
+      "loss": 0.6368,
+      "step": 41516
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6668968486167615e-07,
+      "loss": 0.574,
+      "step": 41517
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6656009967495168e-07,
+      "loss": 0.5798,
+      "step": 41518
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6643056445481365e-07,
+      "loss": 0.5792,
+      "step": 41519
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6630107920192484e-07,
+      "loss": 0.6259,
+      "step": 41520
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6617164391693698e-07,
+      "loss": 0.5649,
+      "step": 41521
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6604225860051283e-07,
+      "loss": 0.5641,
+      "step": 41522
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.659129232533063e-07,
+      "loss": 0.5705,
+      "step": 41523
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6578363787597805e-07,
+      "loss": 0.5953,
+      "step": 41524
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6565440246918196e-07,
+      "loss": 0.6289,
+      "step": 41525
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6552521703357749e-07,
+      "loss": 0.5957,
+      "step": 41526
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6539608156981747e-07,
+      "loss": 0.5884,
+      "step": 41527
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6526699607856133e-07,
+      "loss": 0.6178,
+      "step": 41528
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6513796056046306e-07,
+      "loss": 0.5599,
+      "step": 41529
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6500897501617985e-07,
+      "loss": 0.5749,
+      "step": 41530
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6488003944636676e-07,
+      "loss": 0.604,
+      "step": 41531
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6475115385167772e-07,
+      "loss": 0.5818,
+      "step": 41532
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6462231823276885e-07,
+      "loss": 0.643,
+      "step": 41533
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.644935325902941e-07,
+      "loss": 0.573,
+      "step": 41534
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6436479692490847e-07,
+      "loss": 0.5827,
+      "step": 41535
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6423611123726591e-07,
+      "loss": 0.5588,
+      "step": 41536
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.641074755280203e-07,
+      "loss": 0.5921,
+      "step": 41537
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.639788897978245e-07,
+      "loss": 0.5892,
+      "step": 41538
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.638503540473335e-07,
+      "loss": 0.555,
+      "step": 41539
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6372186827719795e-07,
+      "loss": 0.6654,
+      "step": 41540
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6359343248807392e-07,
+      "loss": 0.5827,
+      "step": 41541
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6346504668061204e-07,
+      "loss": 0.584,
+      "step": 41542
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.633367108554651e-07,
+      "loss": 0.6215,
+      "step": 41543
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6320842501328372e-07,
+      "loss": 0.5789,
+      "step": 41544
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.630801891547229e-07,
+      "loss": 0.5903,
+      "step": 41545
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6295200328043104e-07,
+      "loss": 0.5771,
+      "step": 41546
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6282386739106203e-07,
+      "loss": 0.4938,
+      "step": 41547
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6269578148726538e-07,
+      "loss": 0.573,
+      "step": 41548
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.625677455696928e-07,
+      "loss": 0.6418,
+      "step": 41549
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6243975963899373e-07,
+      "loss": 0.6231,
+      "step": 41550
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.62311823695821e-07,
+      "loss": 0.5534,
+      "step": 41551
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6218393774082187e-07,
+      "loss": 0.579,
+      "step": 41552
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6205610177464803e-07,
+      "loss": 0.5947,
+      "step": 41553
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6192831579794678e-07,
+      "loss": 0.6516,
+      "step": 41554
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6180057981136975e-07,
+      "loss": 0.6381,
+      "step": 41555
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6167289381556539e-07,
+      "loss": 0.5785,
+      "step": 41556
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6154525781118203e-07,
+      "loss": 0.595,
+      "step": 41557
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6141767179887025e-07,
+      "loss": 0.5321,
+      "step": 41558
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.61290135779274e-07,
+      "loss": 0.6621,
+      "step": 41559
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6116264975304606e-07,
+      "loss": 0.5973,
+      "step": 41560
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6103521372083043e-07,
+      "loss": 0.6007,
+      "step": 41561
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.609078276832776e-07,
+      "loss": 0.5569,
+      "step": 41562
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.607804916410327e-07,
+      "loss": 0.6151,
+      "step": 41563
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6065320559474517e-07,
+      "loss": 0.6318,
+      "step": 41564
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6052596954505782e-07,
+      "loss": 0.5799,
+      "step": 41565
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6039878349262127e-07,
+      "loss": 0.576,
+      "step": 41566
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6027164743807944e-07,
+      "loss": 0.616,
+      "step": 41567
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6014456138208069e-07,
+      "loss": 0.5575,
+      "step": 41568
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.6001752532526782e-07,
+      "loss": 0.6658,
+      "step": 41569
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5989053926828814e-07,
+      "loss": 0.6564,
+      "step": 41570
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5976360321178552e-07,
+      "loss": 0.6147,
+      "step": 41571
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5963671715640728e-07,
+      "loss": 0.5802,
+      "step": 41572
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.595098811027962e-07,
+      "loss": 0.579,
+      "step": 41573
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5938309505159732e-07,
+      "loss": 0.5548,
+      "step": 41574
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.592563590034546e-07,
+      "loss": 0.5998,
+      "step": 41575
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5912967295901194e-07,
+      "loss": 0.5728,
+      "step": 41576
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.590030369189144e-07,
+      "loss": 0.5945,
+      "step": 41577
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5887645088380364e-07,
+      "loss": 0.5892,
+      "step": 41578
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5874991485432478e-07,
+      "loss": 0.6063,
+      "step": 41579
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5862342883111947e-07,
+      "loss": 0.5821,
+      "step": 41580
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5849699281483055e-07,
+      "loss": 0.534,
+      "step": 41581
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5837060680609972e-07,
+      "loss": 0.5565,
+      "step": 41582
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5824427080557092e-07,
+      "loss": 0.6102,
+      "step": 41583
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5811798481388474e-07,
+      "loss": 0.6522,
+      "step": 41584
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5799174883168512e-07,
+      "loss": 0.5702,
+      "step": 41585
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.578655628596104e-07,
+      "loss": 0.6177,
+      "step": 41586
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5773942689830347e-07,
+      "loss": 0.592,
+      "step": 41587
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5761334094840375e-07,
+      "loss": 0.594,
+      "step": 41588
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.574873050105541e-07,
+      "loss": 0.6402,
+      "step": 41589
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.57361319085394e-07,
+      "loss": 0.5828,
+      "step": 41590
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5723538317356402e-07,
+      "loss": 0.6042,
+      "step": 41591
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5710949727570256e-07,
+      "loss": 0.5374,
+      "step": 41592
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5698366139245137e-07,
+      "loss": 0.6282,
+      "step": 41593
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5685787552444874e-07,
+      "loss": 0.5847,
+      "step": 41594
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5673213967233315e-07,
+      "loss": 0.68,
+      "step": 41595
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5660645383674623e-07,
+      "loss": 0.5408,
+      "step": 41596
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5648081801832305e-07,
+      "loss": 0.5588,
+      "step": 41597
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5635523221770422e-07,
+      "loss": 0.5968,
+      "step": 41598
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.562296964355259e-07,
+      "loss": 0.6247,
+      "step": 41599
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5610421067242865e-07,
+      "loss": 0.5842,
+      "step": 41600
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5597877492904868e-07,
+      "loss": 0.5886,
+      "step": 41601
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5585338920602432e-07,
+      "loss": 0.6134,
+      "step": 41602
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5572805350399068e-07,
+      "loss": 0.6353,
+      "step": 41603
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5560276782358607e-07,
+      "loss": 0.6071,
+      "step": 41604
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.554775321654467e-07,
+      "loss": 0.5808,
+      "step": 41605
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5535234653020981e-07,
+      "loss": 0.5922,
+      "step": 41606
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5522721091851044e-07,
+      "loss": 0.6027,
+      "step": 41607
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5510212533098478e-07,
+      "loss": 0.6088,
+      "step": 41608
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5497708976826786e-07,
+      "loss": 0.6237,
+      "step": 41609
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5485210423099583e-07,
+      "loss": 0.6161,
+      "step": 41610
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5472716871980376e-07,
+      "loss": 0.5879,
+      "step": 41611
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5460228323532666e-07,
+      "loss": 0.5762,
+      "step": 41612
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5447744777819852e-07,
+      "loss": 0.5743,
+      "step": 41613
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5435266234905322e-07,
+      "loss": 0.5646,
+      "step": 41614
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5422792694852694e-07,
+      "loss": 0.5677,
+      "step": 41615
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5410324157725032e-07,
+      "loss": 0.5699,
+      "step": 41616
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5397860623585948e-07,
+      "loss": 0.5387,
+      "step": 41617
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5385402092498726e-07,
+      "loss": 0.5582,
+      "step": 41618
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5372948564526536e-07,
+      "loss": 0.5903,
+      "step": 41619
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5360500039732773e-07,
+      "loss": 0.5908,
+      "step": 41620
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.534805651818072e-07,
+      "loss": 0.5778,
+      "step": 41621
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.533561799993355e-07,
+      "loss": 0.5736,
+      "step": 41622
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5323184485054543e-07,
+      "loss": 0.6248,
+      "step": 41623
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.531075597360665e-07,
+      "loss": 0.5812,
+      "step": 41624
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5298332465653375e-07,
+      "loss": 0.6067,
+      "step": 41625
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5285913961257447e-07,
+      "loss": 0.5935,
+      "step": 41626
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.527350046048237e-07,
+      "loss": 0.5698,
+      "step": 41627
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5261091963390872e-07,
+      "loss": 0.5884,
+      "step": 41628
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5248688470046346e-07,
+      "loss": 0.5254,
+      "step": 41629
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5236289980511408e-07,
+      "loss": 0.5729,
+      "step": 41630
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5223896494849454e-07,
+      "loss": 0.6142,
+      "step": 41631
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5211508013123212e-07,
+      "loss": 0.5584,
+      "step": 41632
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5199124535395738e-07,
+      "loss": 0.5708,
+      "step": 41633
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5186746061729873e-07,
+      "loss": 0.5962,
+      "step": 41634
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.517437259218868e-07,
+      "loss": 0.6033,
+      "step": 41635
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5162004126834773e-07,
+      "loss": 0.6347,
+      "step": 41636
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5149640665731214e-07,
+      "loss": 0.6072,
+      "step": 41637
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.513728220894073e-07,
+      "loss": 0.5899,
+      "step": 41638
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.512492875652627e-07,
+      "loss": 0.5926,
+      "step": 41639
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.511258030855034e-07,
+      "loss": 0.5685,
+      "step": 41640
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.510023686507589e-07,
+      "loss": 0.6021,
+      "step": 41641
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5087898426165536e-07,
+      "loss": 0.5947,
+      "step": 41642
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5075564991882008e-07,
+      "loss": 0.6483,
+      "step": 41643
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.506323656228803e-07,
+      "loss": 0.6221,
+      "step": 41644
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5050913137446221e-07,
+      "loss": 0.6125,
+      "step": 41645
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5038594717419198e-07,
+      "loss": 0.621,
+      "step": 41646
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5026281302269353e-07,
+      "loss": 0.5825,
+      "step": 41647
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5013972892059637e-07,
+      "loss": 0.5176,
+      "step": 41648
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.5001669486852333e-07,
+      "loss": 0.5951,
+      "step": 41649
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4989371086709946e-07,
+      "loss": 0.5835,
+      "step": 41650
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4977077691695097e-07,
+      "loss": 0.5462,
+      "step": 41651
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4964789301870175e-07,
+      "loss": 0.6406,
+      "step": 41652
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4952505917297687e-07,
+      "loss": 0.5706,
+      "step": 41653
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.494022753803992e-07,
+      "loss": 0.6487,
+      "step": 41654
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4927954164159376e-07,
+      "loss": 0.6298,
+      "step": 41655
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4915685795718447e-07,
+      "loss": 0.616,
+      "step": 41656
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4903422432779423e-07,
+      "loss": 0.5568,
+      "step": 41657
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4891164075404474e-07,
+      "loss": 0.6315,
+      "step": 41658
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4878910723656104e-07,
+      "loss": 0.571,
+      "step": 41659
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4866662377596487e-07,
+      "loss": 0.5516,
+      "step": 41660
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4854419037287903e-07,
+      "loss": 0.6588,
+      "step": 41661
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.484218070279253e-07,
+      "loss": 0.6129,
+      "step": 41662
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4829947374172538e-07,
+      "loss": 0.6121,
+      "step": 41663
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4817719051489986e-07,
+      "loss": 0.5791,
+      "step": 41664
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4805495734807275e-07,
+      "loss": 0.5878,
+      "step": 41665
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.479327742418635e-07,
+      "loss": 0.6173,
+      "step": 41666
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.478106411968927e-07,
+      "loss": 0.5983,
+      "step": 41667
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4768855821378104e-07,
+      "loss": 0.5846,
+      "step": 41668
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.475665252931502e-07,
+      "loss": 0.6033,
+      "step": 41669
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4744454243561858e-07,
+      "loss": 0.594,
+      "step": 41670
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4732260964180678e-07,
+      "loss": 0.6269,
+      "step": 41671
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4720072691233432e-07,
+      "loss": 0.6133,
+      "step": 41672
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4707889424782075e-07,
+      "loss": 0.5938,
+      "step": 41673
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.469571116488855e-07,
+      "loss": 0.6176,
+      "step": 41674
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4683537911614476e-07,
+      "loss": 0.6144,
+      "step": 41675
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4671369665022138e-07,
+      "loss": 0.6632,
+      "step": 41676
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4659206425173045e-07,
+      "loss": 0.5431,
+      "step": 41677
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4647048192129033e-07,
+      "loss": 0.6308,
+      "step": 41678
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.463489496595194e-07,
+      "loss": 0.6204,
+      "step": 41679
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4622746746703608e-07,
+      "loss": 0.5686,
+      "step": 41680
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4610603534445656e-07,
+      "loss": 0.6422,
+      "step": 41681
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.459846532923992e-07,
+      "loss": 0.5409,
+      "step": 41682
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4586332131147796e-07,
+      "loss": 0.5503,
+      "step": 41683
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4574203940231123e-07,
+      "loss": 0.5936,
+      "step": 41684
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.456208075655152e-07,
+      "loss": 0.5818,
+      "step": 41685
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4549962580170606e-07,
+      "loss": 0.665,
+      "step": 41686
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4537849411149996e-07,
+      "loss": 0.6203,
+      "step": 41687
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4525741249551195e-07,
+      "loss": 0.5892,
+      "step": 41688
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.45136380954356e-07,
+      "loss": 0.5879,
+      "step": 41689
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.450153994886483e-07,
+      "loss": 0.612,
+      "step": 41690
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4489446809900497e-07,
+      "loss": 0.6431,
+      "step": 41691
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.447735867860367e-07,
+      "loss": 0.5763,
+      "step": 41692
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4465275555036186e-07,
+      "loss": 0.625,
+      "step": 41693
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4453197439259215e-07,
+      "loss": 0.6307,
+      "step": 41694
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4441124331334268e-07,
+      "loss": 0.6145,
+      "step": 41695
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4429056231322403e-07,
+      "loss": 0.5834,
+      "step": 41696
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4416993139285352e-07,
+      "loss": 0.6065,
+      "step": 41697
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4404935055284064e-07,
+      "loss": 0.5777,
+      "step": 41698
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4392881979380047e-07,
+      "loss": 0.6195,
+      "step": 41699
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4380833911634363e-07,
+      "loss": 0.6924,
+      "step": 41700
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4368790852108294e-07,
+      "loss": 0.5497,
+      "step": 41701
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4356752800863017e-07,
+      "loss": 0.6318,
+      "step": 41702
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4344719757959924e-07,
+      "loss": 0.5768,
+      "step": 41703
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4332691723459857e-07,
+      "loss": 0.5815,
+      "step": 41704
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.43206686974241e-07,
+      "loss": 0.5859,
+      "step": 41705
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4308650679913604e-07,
+      "loss": 0.5603,
+      "step": 41706
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4296637670989543e-07,
+      "loss": 0.5963,
+      "step": 41707
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4284629670712978e-07,
+      "loss": 0.593,
+      "step": 41708
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4272626679144862e-07,
+      "loss": 0.6196,
+      "step": 41709
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4260628696346258e-07,
+      "loss": 0.5918,
+      "step": 41710
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4248635722378112e-07,
+      "loss": 0.6217,
+      "step": 41711
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.423664775730127e-07,
+      "loss": 0.687,
+      "step": 41712
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.422466480117668e-07,
+      "loss": 0.5664,
+      "step": 41713
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4212686854065405e-07,
+      "loss": 0.5733,
+      "step": 41714
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4200713916028064e-07,
+      "loss": 0.5814,
+      "step": 41715
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.418874598712572e-07,
+      "loss": 0.5849,
+      "step": 41716
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4176783067418875e-07,
+      "loss": 0.6857,
+      "step": 41717
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4164825156968597e-07,
+      "loss": 0.645,
+      "step": 41718
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4152872255835503e-07,
+      "loss": 0.5704,
+      "step": 41719
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4140924364080544e-07,
+      "loss": 0.6131,
+      "step": 41720
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4128981481764115e-07,
+      "loss": 0.6406,
+      "step": 41721
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4117043608947167e-07,
+      "loss": 0.5417,
+      "step": 41722
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4105110745690098e-07,
+      "loss": 0.5844,
+      "step": 41723
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.409318289205386e-07,
+      "loss": 0.6398,
+      "step": 41724
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4081260048098732e-07,
+      "loss": 0.6646,
+      "step": 41725
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4069342213885673e-07,
+      "loss": 0.6011,
+      "step": 41726
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.405742938947485e-07,
+      "loss": 0.6289,
+      "step": 41727
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.404552157492711e-07,
+      "loss": 0.6447,
+      "step": 41728
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4033618770302738e-07,
+      "loss": 0.6152,
+      "step": 41729
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4021720975662235e-07,
+      "loss": 0.59,
+      "step": 41730
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4009828191066222e-07,
+      "loss": 0.6195,
+      "step": 41731
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3997940416575095e-07,
+      "loss": 0.6895,
+      "step": 41732
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.398605765224914e-07,
+      "loss": 0.6126,
+      "step": 41733
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3974179898148753e-07,
+      "loss": 0.5693,
+      "step": 41734
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.396230715433433e-07,
+      "loss": 0.6288,
+      "step": 41735
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3950439420866157e-07,
+      "loss": 0.5955,
+      "step": 41736
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.393857669780474e-07,
+      "loss": 0.5603,
+      "step": 41737
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.392671898521003e-07,
+      "loss": 0.5784,
+      "step": 41738
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3914866283142537e-07,
+      "loss": 0.5937,
+      "step": 41739
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.390301859166232e-07,
+      "loss": 0.5599,
+      "step": 41740
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.389117591082978e-07,
+      "loss": 0.5648,
+      "step": 41741
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3879338240704864e-07,
+      "loss": 0.5644,
+      "step": 41742
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3867505581347861e-07,
+      "loss": 0.6149,
+      "step": 41743
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3855677932818835e-07,
+      "loss": 0.6368,
+      "step": 41744
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3843855295177956e-07,
+      "loss": 0.5818,
+      "step": 41745
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3832037668485287e-07,
+      "loss": 0.6393,
+      "step": 41746
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3820225052800896e-07,
+      "loss": 0.6099,
+      "step": 41747
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3808417448184618e-07,
+      "loss": 0.6123,
+      "step": 41748
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3796614854696632e-07,
+      "loss": 0.6415,
+      "step": 41749
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.378481727239689e-07,
+      "loss": 0.5926,
+      "step": 41750
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3773024701345229e-07,
+      "loss": 0.5657,
+      "step": 41751
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3761237141601824e-07,
+      "loss": 0.6355,
+      "step": 41752
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3749454593226296e-07,
+      "loss": 0.6124,
+      "step": 41753
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.373767705627871e-07,
+      "loss": 0.6371,
+      "step": 41754
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.372590453081868e-07,
+      "loss": 0.5372,
+      "step": 41755
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3714137016906382e-07,
+      "loss": 0.5984,
+      "step": 41756
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3702374514601212e-07,
+      "loss": 0.5971,
+      "step": 41757
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3690617023963237e-07,
+      "loss": 0.5673,
+      "step": 41758
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3678864545052074e-07,
+      "loss": 0.5999,
+      "step": 41759
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.366711707792745e-07,
+      "loss": 0.6512,
+      "step": 41760
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3655374622648986e-07,
+      "loss": 0.6321,
+      "step": 41761
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3643637179276525e-07,
+      "loss": 0.5816,
+      "step": 41762
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.363190474786946e-07,
+      "loss": 0.6196,
+      "step": 41763
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.362017732848775e-07,
+      "loss": 0.588,
+      "step": 41764
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3608454921190673e-07,
+      "loss": 0.5908,
+      "step": 41765
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3596737526038073e-07,
+      "loss": 0.5758,
+      "step": 41766
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3585025143089236e-07,
+      "loss": 0.6315,
+      "step": 41767
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3573317772403672e-07,
+      "loss": 0.5635,
+      "step": 41768
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3561615414041108e-07,
+      "loss": 0.5559,
+      "step": 41769
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.354991806806083e-07,
+      "loss": 0.5331,
+      "step": 41770
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3538225734522238e-07,
+      "loss": 0.555,
+      "step": 41771
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3526538413484836e-07,
+      "loss": 0.6683,
+      "step": 41772
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3514856105008134e-07,
+      "loss": 0.6272,
+      "step": 41773
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3503178809151197e-07,
+      "loss": 0.5774,
+      "step": 41774
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3491506525973642e-07,
+      "loss": 0.6339,
+      "step": 41775
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3479839255534533e-07,
+      "loss": 0.627,
+      "step": 41776
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.346817699789338e-07,
+      "loss": 0.5607,
+      "step": 41777
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3456519753109242e-07,
+      "loss": 0.6228,
+      "step": 41778
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3444867521241522e-07,
+      "loss": 0.5252,
+      "step": 41779
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.343322030234928e-07,
+      "loss": 0.6475,
+      "step": 41780
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3421578096491917e-07,
+      "loss": 0.5171,
+      "step": 41781
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.340994090372827e-07,
+      "loss": 0.5906,
+      "step": 41782
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3398308724117737e-07,
+      "loss": 0.5735,
+      "step": 41783
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3386681557719383e-07,
+      "loss": 0.6009,
+      "step": 41784
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.337505940459227e-07,
+      "loss": 0.5941,
+      "step": 41785
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3363442264795245e-07,
+      "loss": 0.5801,
+      "step": 41786
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.33518301383877e-07,
+      "loss": 0.5615,
+      "step": 41787
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3340223025428366e-07,
+      "loss": 0.6279,
+      "step": 41788
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.332862092597631e-07,
+      "loss": 0.6023,
+      "step": 41789
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3317023840090592e-07,
+      "loss": 0.61,
+      "step": 41790
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3305431767829946e-07,
+      "loss": 0.5938,
+      "step": 41791
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3293844709253433e-07,
+      "loss": 0.649,
+      "step": 41792
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3282262664419788e-07,
+      "loss": 0.571,
+      "step": 41793
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.327068563338796e-07,
+      "loss": 0.6261,
+      "step": 41794
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.325911361621679e-07,
+      "loss": 0.5864,
+      "step": 41795
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3247546612965123e-07,
+      "loss": 0.6612,
+      "step": 41796
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3235984623691466e-07,
+      "loss": 0.587,
+      "step": 41797
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3224427648454885e-07,
+      "loss": 0.5875,
+      "step": 41798
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3212875687313887e-07,
+      "loss": 0.6001,
+      "step": 41799
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3201328740327312e-07,
+      "loss": 0.6052,
+      "step": 41800
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3189786807553785e-07,
+      "loss": 0.6072,
+      "step": 41801
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.317824988905203e-07,
+      "loss": 0.5612,
+      "step": 41802
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3166717984880451e-07,
+      "loss": 0.6492,
+      "step": 41803
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3155191095097885e-07,
+      "loss": 0.6261,
+      "step": 41804
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.3143669219762733e-07,
+      "loss": 0.575,
+      "step": 41805
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3132152358933615e-07,
+      "loss": 0.6129,
+      "step": 41806
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3120640512669146e-07,
+      "loss": 0.5958,
+      "step": 41807
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.310913368102773e-07,
+      "loss": 0.6131,
+      "step": 41808
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3097631864067763e-07,
+      "loss": 0.5855,
+      "step": 41809
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3086135061847637e-07,
+      "loss": 0.6323,
+      "step": 41810
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.307464327442609e-07,
+      "loss": 0.5863,
+      "step": 41811
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3063156501861184e-07,
+      "loss": 0.5899,
+      "step": 41812
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.305167474421154e-07,
+      "loss": 0.6213,
+      "step": 41813
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.304019800153522e-07,
+      "loss": 0.5792,
+      "step": 41814
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3028726273890846e-07,
+      "loss": 0.6024,
+      "step": 41815
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3017259561336482e-07,
+      "loss": 0.6411,
+      "step": 41816
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.3005797863930414e-07,
+      "loss": 0.6076,
+      "step": 41817
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.299434118173104e-07,
+      "loss": 0.6014,
+      "step": 41818
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2982889514796536e-07,
+      "loss": 0.5764,
+      "step": 41819
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2971442863184858e-07,
+      "loss": 0.6397,
+      "step": 41820
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2960001226954511e-07,
+      "loss": 0.5924,
+      "step": 41821
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2948564606163338e-07,
+      "loss": 0.6234,
+      "step": 41822
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.293713300086974e-07,
+      "loss": 0.6389,
+      "step": 41823
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2925706411131445e-07,
+      "loss": 0.6815,
+      "step": 41824
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2914284837006852e-07,
+      "loss": 0.5688,
+      "step": 41825
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2902868278553692e-07,
+      "loss": 0.6415,
+      "step": 41826
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2891456735830365e-07,
+      "loss": 0.5741,
+      "step": 41827
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.288005020889449e-07,
+      "loss": 0.6038,
+      "step": 41828
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2868648697804243e-07,
+      "loss": 0.6305,
+      "step": 41829
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2857252202617465e-07,
+      "loss": 0.627,
+      "step": 41830
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2845860723392112e-07,
+      "loss": 0.6004,
+      "step": 41831
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2834474260186026e-07,
+      "loss": 0.6152,
+      "step": 41832
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2823092813056937e-07,
+      "loss": 0.5843,
+      "step": 41833
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.281171638206302e-07,
+      "loss": 0.5896,
+      "step": 41834
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.280034496726179e-07,
+      "loss": 0.6244,
+      "step": 41835
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2788978568711197e-07,
+      "loss": 0.6329,
+      "step": 41836
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.277761718646886e-07,
+      "loss": 0.5944,
+      "step": 41837
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2766260820592512e-07,
+      "loss": 0.6319,
+      "step": 41838
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2754909471139997e-07,
+      "loss": 0.6155,
+      "step": 41839
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2743563138168934e-07,
+      "loss": 0.5948,
+      "step": 41840
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2732221821736946e-07,
+      "loss": 0.6355,
+      "step": 41841
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.272088552190165e-07,
+      "loss": 0.6163,
+      "step": 41842
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2709554238720668e-07,
+      "loss": 0.6463,
+      "step": 41843
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.269822797225162e-07,
+      "loss": 0.6136,
+      "step": 41844
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2686906722552017e-07,
+      "loss": 0.5858,
+      "step": 41845
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2675590489679367e-07,
+      "loss": 0.664,
+      "step": 41846
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2664279273691071e-07,
+      "loss": 0.6123,
+      "step": 41847
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.265297307464497e-07,
+      "loss": 0.545,
+      "step": 41848
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2641671892598128e-07,
+      "loss": 0.5659,
+      "step": 41849
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.263037572760817e-07,
+      "loss": 0.5696,
+      "step": 41850
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2619084579732377e-07,
+      "loss": 0.5909,
+      "step": 41851
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2607798449028154e-07,
+      "loss": 0.5656,
+      "step": 41852
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2596517335553006e-07,
+      "loss": 0.6341,
+      "step": 41853
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2585241239364e-07,
+      "loss": 0.5699,
+      "step": 41854
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2573970160518645e-07,
+      "loss": 0.6248,
+      "step": 41855
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.256270409907412e-07,
+      "loss": 0.5633,
+      "step": 41856
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2551443055087597e-07,
+      "loss": 0.5873,
+      "step": 41857
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.254018702861648e-07,
+      "loss": 0.5709,
+      "step": 41858
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.252893601971783e-07,
+      "loss": 0.6423,
+      "step": 41859
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2517690028448715e-07,
+      "loss": 0.6025,
+      "step": 41860
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2506449054866643e-07,
+      "loss": 0.5913,
+      "step": 41861
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2495213099028458e-07,
+      "loss": 0.5978,
+      "step": 41862
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2483982160991225e-07,
+      "loss": 0.6699,
+      "step": 41863
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2472756240812012e-07,
+      "loss": 0.6102,
+      "step": 41864
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2461535338548103e-07,
+      "loss": 0.6222,
+      "step": 41865
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2450319454256232e-07,
+      "loss": 0.6001,
+      "step": 41866
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2439108587993577e-07,
+      "loss": 0.5577,
+      "step": 41867
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2427902739817088e-07,
+      "loss": 0.641,
+      "step": 41868
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2416701909783502e-07,
+      "loss": 0.7134,
+      "step": 41869
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2405506097949994e-07,
+      "loss": 0.6556,
+      "step": 41870
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2394315304373185e-07,
+      "loss": 0.5648,
+      "step": 41871
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.238312952911036e-07,
+      "loss": 0.629,
+      "step": 41872
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.237194877221781e-07,
+      "loss": 0.5523,
+      "step": 41873
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2360773033752826e-07,
+      "loss": 0.6068,
+      "step": 41874
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2349602313771802e-07,
+      "loss": 0.5974,
+      "step": 41875
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2338436612331806e-07,
+      "loss": 0.6401,
+      "step": 41876
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2327275929489568e-07,
+      "loss": 0.5517,
+      "step": 41877
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2316120265301602e-07,
+      "loss": 0.5792,
+      "step": 41878
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2304969619824636e-07,
+      "loss": 0.6239,
+      "step": 41879
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2293823993115404e-07,
+      "loss": 0.5615,
+      "step": 41880
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2282683385230533e-07,
+      "loss": 0.5848,
+      "step": 41881
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2271547796226636e-07,
+      "loss": 0.6267,
+      "step": 41882
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2260417226160225e-07,
+      "loss": 0.527,
+      "step": 41883
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2249291675088038e-07,
+      "loss": 0.5742,
+      "step": 41884
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2238171143066247e-07,
+      "loss": 0.5989,
+      "step": 41885
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2227055630151808e-07,
+      "loss": 0.6383,
+      "step": 41886
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2215945136400897e-07,
+      "loss": 0.562,
+      "step": 41887
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2204839661870028e-07,
+      "loss": 0.5685,
+      "step": 41888
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2193739206615707e-07,
+      "loss": 0.562,
+      "step": 41889
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2182643770694224e-07,
+      "loss": 0.6156,
+      "step": 41890
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.217155335416209e-07,
+      "loss": 0.5248,
+      "step": 41891
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2160467957075484e-07,
+      "loss": 0.5751,
+      "step": 41892
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2149387579490912e-07,
+      "loss": 0.6247,
+      "step": 41893
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2138312221464666e-07,
+      "loss": 0.5759,
+      "step": 41894
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.212724188305292e-07,
+      "loss": 0.5832,
+      "step": 41895
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2116176564311964e-07,
+      "loss": 0.5801,
+      "step": 41896
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2105116265297979e-07,
+      "loss": 0.5794,
+      "step": 41897
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2094060986067246e-07,
+      "loss": 0.61,
+      "step": 41898
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2083010726676058e-07,
+      "loss": 0.6184,
+      "step": 41899
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.207196548718026e-07,
+      "loss": 0.5852,
+      "step": 41900
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2060925267636137e-07,
+      "loss": 0.5206,
+      "step": 41901
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2049890068099756e-07,
+      "loss": 0.6243,
+      "step": 41902
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2038859888627297e-07,
+      "loss": 0.6136,
+      "step": 41903
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2027834729274713e-07,
+      "loss": 0.6081,
+      "step": 41904
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.201681459009807e-07,
+      "loss": 0.5873,
+      "step": 41905
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.2005799471153323e-07,
+      "loss": 0.5575,
+      "step": 41906
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.199478937249643e-07,
+      "loss": 0.6002,
+      "step": 41907
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1983784294183343e-07,
+      "loss": 0.6109,
+      "step": 41908
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1972784236270018e-07,
+      "loss": 0.6679,
+      "step": 41909
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1961789198812413e-07,
+      "loss": 0.5817,
+      "step": 41910
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1950799181866257e-07,
+      "loss": 0.6253,
+      "step": 41911
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1939814185487396e-07,
+      "loss": 0.5784,
+      "step": 41912
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1928834209731678e-07,
+      "loss": 0.6066,
+      "step": 41913
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1917859254655051e-07,
+      "loss": 0.571,
+      "step": 41914
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1906889320313142e-07,
+      "loss": 0.62,
+      "step": 41915
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1895924406761683e-07,
+      "loss": 0.5998,
+      "step": 41916
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1884964514056297e-07,
+      "loss": 0.5904,
+      "step": 41917
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1874009642252937e-07,
+      "loss": 0.5977,
+      "step": 41918
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1863059791407005e-07,
+      "loss": 0.5486,
+      "step": 41919
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1852114961574346e-07,
+      "loss": 0.6536,
+      "step": 41920
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1841175152810469e-07,
+      "loss": 0.6451,
+      "step": 41921
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1830240365170997e-07,
+      "loss": 0.5661,
+      "step": 41922
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1819310598711442e-07,
+      "loss": 0.5893,
+      "step": 41923
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1808385853487426e-07,
+      "loss": 0.6144,
+      "step": 41924
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1797466129554346e-07,
+      "loss": 0.6239,
+      "step": 41925
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1786551426967719e-07,
+      "loss": 0.5715,
+      "step": 41926
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1775641745783051e-07,
+      "loss": 0.6586,
+      "step": 41927
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1764737086055855e-07,
+      "loss": 0.6243,
+      "step": 41928
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.175383744784131e-07,
+      "loss": 0.5741,
+      "step": 41929
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1742942831195037e-07,
+      "loss": 0.5906,
+      "step": 41930
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1732053236172325e-07,
+      "loss": 0.5988,
+      "step": 41931
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1721168662828353e-07,
+      "loss": 0.5754,
+      "step": 41932
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1710289111218631e-07,
+      "loss": 0.6111,
+      "step": 41933
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1699414581398338e-07,
+      "loss": 0.6365,
+      "step": 41934
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1688545073422764e-07,
+      "loss": 0.6457,
+      "step": 41935
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1677680587347084e-07,
+      "loss": 0.6243,
+      "step": 41936
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1666821123226702e-07,
+      "loss": 0.5898,
+      "step": 41937
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.165596668111646e-07,
+      "loss": 0.5585,
+      "step": 41938
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.164511726107187e-07,
+      "loss": 0.5874,
+      "step": 41939
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1634272863147777e-07,
+      "loss": 0.6166,
+      "step": 41940
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1623433487399471e-07,
+      "loss": 0.5632,
+      "step": 41941
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1612599133881908e-07,
+      "loss": 0.592,
+      "step": 41942
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1601769802650265e-07,
+      "loss": 0.6294,
+      "step": 41943
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1590945493759387e-07,
+      "loss": 0.6305,
+      "step": 41944
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1580126207264453e-07,
+      "loss": 0.5753,
+      "step": 41945
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1569311943220418e-07,
+      "loss": 0.6651,
+      "step": 41946
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1558502701682017e-07,
+      "loss": 0.6352,
+      "step": 41947
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.154769848270454e-07,
+      "loss": 0.5985,
+      "step": 41948
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1536899286342607e-07,
+      "loss": 0.5907,
+      "step": 41949
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1526105112651176e-07,
+      "loss": 0.5927,
+      "step": 41950
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.151531596168498e-07,
+      "loss": 0.5696,
+      "step": 41951
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.1504531833499089e-07,
+      "loss": 0.6627,
+      "step": 41952
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1493752728148122e-07,
+      "loss": 0.6688,
+      "step": 41953
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1482978645686926e-07,
+      "loss": 0.5866,
+      "step": 41954
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1472209586170124e-07,
+      "loss": 0.6021,
+      "step": 41955
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1461445549652672e-07,
+      "loss": 0.6121,
+      "step": 41956
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1450686536188971e-07,
+      "loss": 0.5546,
+      "step": 41957
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1439932545834088e-07,
+      "loss": 0.5596,
+      "step": 41958
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.14291835786422e-07,
+      "loss": 0.5636,
+      "step": 41959
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1418439634668155e-07,
+      "loss": 0.6037,
+      "step": 41960
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1407700713966574e-07,
+      "loss": 0.6033,
+      "step": 41961
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.139696681659197e-07,
+      "loss": 0.5977,
+      "step": 41962
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1386237942598965e-07,
+      "loss": 0.6172,
+      "step": 41963
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1375514092042072e-07,
+      "loss": 0.5569,
+      "step": 41964
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1364795264975581e-07,
+      "loss": 0.5427,
+      "step": 41965
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1354081461454114e-07,
+      "loss": 0.6176,
+      "step": 41966
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1343372681532183e-07,
+      "loss": 0.6133,
+      "step": 41967
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1332668925263968e-07,
+      "loss": 0.6305,
+      "step": 41968
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1321970192704202e-07,
+      "loss": 0.5723,
+      "step": 41969
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1311276483906952e-07,
+      "loss": 0.5379,
+      "step": 41970
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1300587798926621e-07,
+      "loss": 0.5318,
+      "step": 41971
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1289904137817608e-07,
+      "loss": 0.5691,
+      "step": 41972
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1279225500634206e-07,
+      "loss": 0.6417,
+      "step": 41973
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.126855188743048e-07,
+      "loss": 0.5813,
+      "step": 41974
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1257883298260942e-07,
+      "loss": 0.545,
+      "step": 41975
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1247219733179549e-07,
+      "loss": 0.6263,
+      "step": 41976
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1236561192240703e-07,
+      "loss": 0.6282,
+      "step": 41977
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.122590767549836e-07,
+      "loss": 0.5968,
+      "step": 41978
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.121525918300681e-07,
+      "loss": 0.6773,
+      "step": 41979
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1204615714820122e-07,
+      "loss": 0.5855,
+      "step": 41980
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1193977270992362e-07,
+      "loss": 0.6278,
+      "step": 41981
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.11833438515776e-07,
+      "loss": 0.6687,
+      "step": 41982
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.117271545662979e-07,
+      "loss": 0.6102,
+      "step": 41983
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1162092086203114e-07,
+      "loss": 0.6208,
+      "step": 41984
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1151473740351304e-07,
+      "loss": 0.6154,
+      "step": 41985
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.114086041912854e-07,
+      "loss": 0.5751,
+      "step": 41986
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1130252122588669e-07,
+      "loss": 0.619,
+      "step": 41987
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1119648850785647e-07,
+      "loss": 0.6331,
+      "step": 41988
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1109050603773208e-07,
+      "loss": 0.5828,
+      "step": 41989
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.109845738160531e-07,
+      "loss": 0.6226,
+      "step": 41990
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1087869184335798e-07,
+      "loss": 0.6192,
+      "step": 41991
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1077286012018407e-07,
+      "loss": 0.691,
+      "step": 41992
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1066707864706872e-07,
+      "loss": 0.662,
+      "step": 41993
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1056134742455038e-07,
+      "loss": 0.5418,
+      "step": 41994
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1045566645316641e-07,
+      "loss": 0.5942,
+      "step": 41995
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1035003573345527e-07,
+      "loss": 0.617,
+      "step": 41996
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1024445526594984e-07,
+      "loss": 0.6179,
+      "step": 41997
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1013892505118862e-07,
+      "loss": 0.5917,
+      "step": 41998
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.1003344508970782e-07,
+      "loss": 0.5947,
+      "step": 41999
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0992801538204478e-07,
+      "loss": 0.6136,
+      "step": 42000
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0982263592873354e-07,
+      "loss": 0.6097,
+      "step": 42001
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0971730673030923e-07,
+      "loss": 0.5461,
+      "step": 42002
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0961202778730807e-07,
+      "loss": 0.5716,
+      "step": 42003
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0950679910026518e-07,
+      "loss": 0.6111,
+      "step": 42004
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0940162066971461e-07,
+      "loss": 0.6039,
+      "step": 42005
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0929649249619034e-07,
+      "loss": 0.6004,
+      "step": 42006
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0919141458022864e-07,
+      "loss": 0.6309,
+      "step": 42007
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.090863869223613e-07,
+      "loss": 0.6156,
+      "step": 42008
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0898140952312231e-07,
+      "loss": 0.5904,
+      "step": 42009
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.088764823830457e-07,
+      "loss": 0.5966,
+      "step": 42010
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0877160550266441e-07,
+      "loss": 0.642,
+      "step": 42011
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0866677888251131e-07,
+      "loss": 0.5666,
+      "step": 42012
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0856200252311933e-07,
+      "loss": 0.6091,
+      "step": 42013
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0845727642502025e-07,
+      "loss": 0.6205,
+      "step": 42014
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0835260058874697e-07,
+      "loss": 0.5731,
+      "step": 42015
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.082479750148313e-07,
+      "loss": 0.568,
+      "step": 42016
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0814339970380395e-07,
+      "loss": 0.5718,
+      "step": 42017
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0803887465619777e-07,
+      "loss": 0.6442,
+      "step": 42018
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.079343998725435e-07,
+      "loss": 0.5697,
+      "step": 42019
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0782997535336958e-07,
+      "loss": 0.6405,
+      "step": 42020
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0772560109921005e-07,
+      "loss": 0.587,
+      "step": 42021
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0762127711059334e-07,
+      "loss": 0.6017,
+      "step": 42022
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0751700338805126e-07,
+      "loss": 0.5975,
+      "step": 42023
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0741277993211008e-07,
+      "loss": 0.6098,
+      "step": 42024
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.073086067433038e-07,
+      "loss": 0.6265,
+      "step": 42025
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0720448382215865e-07,
+      "loss": 0.6226,
+      "step": 42026
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0710041116920533e-07,
+      "loss": 0.5907,
+      "step": 42027
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.069963887849712e-07,
+      "loss": 0.6263,
+      "step": 42028
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0689241666998696e-07,
+      "loss": 0.5827,
+      "step": 42029
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0678849482477882e-07,
+      "loss": 0.6003,
+      "step": 42030
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0668462324987639e-07,
+      "loss": 0.555,
+      "step": 42031
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.065808019458059e-07,
+      "loss": 0.5787,
+      "step": 42032
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.064770309130958e-07,
+      "loss": 0.5745,
+      "step": 42033
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0637331015227459e-07,
+      "loss": 0.6519,
+      "step": 42034
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0626963966386628e-07,
+      "loss": 0.5608,
+      "step": 42035
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0616601944840155e-07,
+      "loss": 0.6686,
+      "step": 42036
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0606244950640332e-07,
+      "loss": 0.6174,
+      "step": 42037
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0595892983839895e-07,
+      "loss": 0.6397,
+      "step": 42038
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0585546044491579e-07,
+      "loss": 0.6037,
+      "step": 42039
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0575204132647899e-07,
+      "loss": 0.613,
+      "step": 42040
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0564867248361254e-07,
+      "loss": 0.5641,
+      "step": 42041
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0554535391684383e-07,
+      "loss": 0.5381,
+      "step": 42042
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0544208562669689e-07,
+      "loss": 0.587,
+      "step": 42043
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0533886761369572e-07,
+      "loss": 0.595,
+      "step": 42044
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0523569987836657e-07,
+      "loss": 0.5871,
+      "step": 42045
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0513258242123237e-07,
+      "loss": 0.6207,
+      "step": 42046
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0502951524281713e-07,
+      "loss": 0.5649,
+      "step": 42047
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0492649834364488e-07,
+      "loss": 0.6369,
+      "step": 42048
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0482353172423853e-07,
+      "loss": 0.6016,
+      "step": 42049
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0472061538512323e-07,
+      "loss": 0.5869,
+      "step": 42050
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0461774932681856e-07,
+      "loss": 0.6377,
+      "step": 42051
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0451493354984965e-07,
+      "loss": 0.571,
+      "step": 42052
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0441216805473942e-07,
+      "loss": 0.6305,
+      "step": 42053
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0430945284200744e-07,
+      "loss": 0.5836,
+      "step": 42054
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0420678791217887e-07,
+      "loss": 0.5998,
+      "step": 42055
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0410417326577216e-07,
+      "loss": 0.5853,
+      "step": 42056
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0400160890331135e-07,
+      "loss": 0.6025,
+      "step": 42057
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0389909482531602e-07,
+      "loss": 0.6092,
+      "step": 42058
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0379663103230797e-07,
+      "loss": 0.5756,
+      "step": 42059
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0369421752480791e-07,
+      "loss": 0.5618,
+      "step": 42060
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0359185430333541e-07,
+      "loss": 0.6267,
+      "step": 42061
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0348954136841006e-07,
+      "loss": 0.5872,
+      "step": 42062
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0338727872055366e-07,
+      "loss": 0.611,
+      "step": 42063
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0328506636028469e-07,
+      "loss": 0.628,
+      "step": 42064
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.031829042881216e-07,
+      "loss": 0.6144,
+      "step": 42065
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0308079250458624e-07,
+      "loss": 0.6156,
+      "step": 42066
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0297873101019484e-07,
+      "loss": 0.6159,
+      "step": 42067
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0287671980546809e-07,
+      "loss": 0.6363,
+      "step": 42068
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0277475889092225e-07,
+      "loss": 0.5818,
+      "step": 42069
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.026728482670758e-07,
+      "loss": 0.5524,
+      "step": 42070
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.025709879344483e-07,
+      "loss": 0.5874,
+      "step": 42071
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0246917789355604e-07,
+      "loss": 0.6042,
+      "step": 42072
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0236741814491635e-07,
+      "loss": 0.6398,
+      "step": 42073
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0226570868904662e-07,
+      "loss": 0.5889,
+      "step": 42074
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0216404952646308e-07,
+      "loss": 0.6131,
+      "step": 42075
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0206244065768311e-07,
+      "loss": 0.6028,
+      "step": 42076
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0196088208322185e-07,
+      "loss": 0.5894,
+      "step": 42077
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0185937380359779e-07,
+      "loss": 0.5621,
+      "step": 42078
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0175791581932381e-07,
+      "loss": 0.5868,
+      "step": 42079
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.016565081309162e-07,
+      "loss": 0.6109,
+      "step": 42080
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0155515073889233e-07,
+      "loss": 0.5652,
+      "step": 42081
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0145384364376398e-07,
+      "loss": 0.5478,
+      "step": 42082
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0135258684604854e-07,
+      "loss": 0.5854,
+      "step": 42083
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0125138034626003e-07,
+      "loss": 0.6083,
+      "step": 42084
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0115022414491138e-07,
+      "loss": 0.5859,
+      "step": 42085
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0104911824251773e-07,
+      "loss": 0.586,
+      "step": 42086
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0094806263959201e-07,
+      "loss": 0.5955,
+      "step": 42087
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0084705733664935e-07,
+      "loss": 0.588,
+      "step": 42088
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0074610233420157e-07,
+      "loss": 0.6371,
+      "step": 42089
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0064519763276048e-07,
+      "loss": 0.5933,
+      "step": 42090
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0054434323284235e-07,
+      "loss": 0.5872,
+      "step": 42091
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0044353913495674e-07,
+      "loss": 0.5933,
+      "step": 42092
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.003427853396166e-07,
+      "loss": 0.6082,
+      "step": 42093
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0024208184733374e-07,
+      "loss": 0.5965,
+      "step": 42094
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0014142865862108e-07,
+      "loss": 0.585,
+      "step": 42095
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.0004082577398822e-07,
+      "loss": 0.5544,
+      "step": 42096
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 9.994027319394805e-08,
+      "loss": 0.5522,
+      "step": 42097
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 9.98397709190102e-08,
+      "loss": 0.6397,
+      "step": 42098
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 9.973931894968646e-08,
+      "loss": 0.6085,
+      "step": 42099
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.963891728648534e-08,
+      "loss": 0.5755,
+      "step": 42100
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.953856592991973e-08,
+      "loss": 0.6263,
+      "step": 42101
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.943826488049813e-08,
+      "loss": 0.6192,
+      "step": 42102
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.933801413872901e-08,
+      "loss": 0.5997,
+      "step": 42103
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.92378137051242e-08,
+      "loss": 0.6283,
+      "step": 42104
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.913766358019105e-08,
+      "loss": 0.5958,
+      "step": 42105
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.903756376443918e-08,
+      "loss": 0.6126,
+      "step": 42106
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.893751425837595e-08,
+      "loss": 0.5517,
+      "step": 42107
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.883751506251204e-08,
+      "loss": 0.5724,
+      "step": 42108
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.873756617735486e-08,
+      "loss": 0.6113,
+      "step": 42109
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.863766760341176e-08,
+      "loss": 0.6099,
+      "step": 42110
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.8537819341189e-08,
+      "loss": 0.6267,
+      "step": 42111
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.843802139119618e-08,
+      "loss": 0.5936,
+      "step": 42112
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.833827375393957e-08,
+      "loss": 0.588,
+      "step": 42113
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.823857642992651e-08,
+      "loss": 0.5819,
+      "step": 42114
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.81389294196633e-08,
+      "loss": 0.5937,
+      "step": 42115
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.803933272365506e-08,
+      "loss": 0.5641,
+      "step": 42116
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.79397863424092e-08,
+      "loss": 0.6416,
+      "step": 42117
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.784029027643194e-08,
+      "loss": 0.5694,
+      "step": 42118
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.774084452622622e-08,
+      "loss": 0.5723,
+      "step": 42119
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.764144909230166e-08,
+      "loss": 0.6225,
+      "step": 42120
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.754210397516006e-08,
+      "loss": 0.6429,
+      "step": 42121
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.744280917530657e-08,
+      "loss": 0.6008,
+      "step": 42122
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.734356469324747e-08,
+      "loss": 0.6256,
+      "step": 42123
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.724437052948454e-08,
+      "loss": 0.6241,
+      "step": 42124
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.714522668452298e-08,
+      "loss": 0.5515,
+      "step": 42125
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.704613315886791e-08,
+      "loss": 0.5965,
+      "step": 42126
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.694708995302115e-08,
+      "loss": 0.6415,
+      "step": 42127
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.684809706748566e-08,
+      "loss": 0.6172,
+      "step": 42128
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.674915450276656e-08,
+      "loss": 0.5866,
+      "step": 42129
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.665026225936458e-08,
+      "loss": 0.6067,
+      "step": 42130
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.655142033778265e-08,
+      "loss": 0.6673,
+      "step": 42131
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.64526287385248e-08,
+      "loss": 0.6012,
+      "step": 42132
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.635388746209062e-08,
+      "loss": 0.6271,
+      "step": 42133
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.625519650898196e-08,
+      "loss": 0.6232,
+      "step": 42134
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.615655587970174e-08,
+      "loss": 0.5854,
+      "step": 42135
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.605796557475066e-08,
+      "loss": 0.6217,
+      "step": 42136
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.595942559462945e-08,
+      "loss": 0.6083,
+      "step": 42137
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.58609359398388e-08,
+      "loss": 0.6313,
+      "step": 42138
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.576249661087944e-08,
+      "loss": 0.6101,
+      "step": 42139
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.566410760825206e-08,
+      "loss": 0.5707,
+      "step": 42140
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.556576893245518e-08,
+      "loss": 0.6162,
+      "step": 42141
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.546748058398947e-08,
+      "loss": 0.6622,
+      "step": 42142
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.536924256335456e-08,
+      "loss": 0.6415,
+      "step": 42143
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.527105487104893e-08,
+      "loss": 0.5776,
+      "step": 42144
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.517291750757108e-08,
+      "loss": 0.6006,
+      "step": 42145
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.507483047342059e-08,
+      "loss": 0.6273,
+      "step": 42146
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.497679376909597e-08,
+      "loss": 0.6309,
+      "step": 42147
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.48788073950957e-08,
+      "loss": 0.6416,
+      "step": 42148
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.478087135191605e-08,
+      "loss": 0.637,
+      "step": 42149
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.468298564005662e-08,
+      "loss": 0.5763,
+      "step": 42150
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.458515026001369e-08,
+      "loss": 0.616,
+      "step": 42151
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.448736521228352e-08,
+      "loss": 0.5637,
+      "step": 42152
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.438963049736571e-08,
+      "loss": 0.6559,
+      "step": 42153
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.429194611575432e-08,
+      "loss": 0.5773,
+      "step": 42154
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.41943120679456e-08,
+      "loss": 0.5586,
+      "step": 42155
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.409672835443806e-08,
+      "loss": 0.5988,
+      "step": 42156
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.399919497572463e-08,
+      "loss": 0.6181,
+      "step": 42157
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.39017119323038e-08,
+      "loss": 0.5572,
+      "step": 42158
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.380427922466962e-08,
+      "loss": 0.59,
+      "step": 42159
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.370689685331613e-08,
+      "loss": 0.5905,
+      "step": 42160
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.360956481873962e-08,
+      "loss": 0.5928,
+      "step": 42161
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.351228312143412e-08,
+      "loss": 0.6204,
+      "step": 42162
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.341505176189369e-08,
+      "loss": 0.5249,
+      "step": 42163
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.331787074061238e-08,
+      "loss": 0.6282,
+      "step": 42164
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.322074005808535e-08,
+      "loss": 0.5732,
+      "step": 42165
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.312365971480442e-08,
+      "loss": 0.5803,
+      "step": 42166
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.302662971126364e-08,
+      "loss": 0.6335,
+      "step": 42167
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.292965004795485e-08,
+      "loss": 0.578,
+      "step": 42168
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.283272072537319e-08,
+      "loss": 0.614,
+      "step": 42169
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.27358417440094e-08,
+      "loss": 0.7062,
+      "step": 42170
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.263901310435752e-08,
+      "loss": 0.6024,
+      "step": 42171
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.254223480690606e-08,
+      "loss": 0.6125,
+      "step": 42172
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.244550685215125e-08,
+      "loss": 0.5895,
+      "step": 42173
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.234882924058053e-08,
+      "loss": 0.6401,
+      "step": 42174
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.225220197268903e-08,
+      "loss": 0.627,
+      "step": 42175
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.215562504896414e-08,
+      "loss": 0.5568,
+      "step": 42176
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.20590984698988e-08,
+      "loss": 0.5823,
+      "step": 42177
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.196262223598152e-08,
+      "loss": 0.6173,
+      "step": 42178
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.186619634770522e-08,
+      "loss": 0.6131,
+      "step": 42179
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.17698208055584e-08,
+      "loss": 0.6364,
+      "step": 42180
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.16734956100318e-08,
+      "loss": 0.6312,
+      "step": 42181
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.157722076161279e-08,
+      "loss": 0.5251,
+      "step": 42182
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.148099626079098e-08,
+      "loss": 0.6242,
+      "step": 42183
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.13848221080571e-08,
+      "loss": 0.5827,
+      "step": 42184
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.128869830389853e-08,
+      "loss": 0.6056,
+      "step": 42185
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.119262484880486e-08,
+      "loss": 0.5754,
+      "step": 42186
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.10966017432624e-08,
+      "loss": 0.5495,
+      "step": 42187
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.100062898776074e-08,
+      "loss": 0.5778,
+      "step": 42188
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.090470658278617e-08,
+      "loss": 0.6213,
+      "step": 42189
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.080883452882716e-08,
+      "loss": 0.597,
+      "step": 42190
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.071301282637002e-08,
+      "loss": 0.6733,
+      "step": 42191
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.061724147590322e-08,
+      "loss": 0.6256,
+      "step": 42192
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.052152047791085e-08,
+      "loss": 0.6157,
+      "step": 42193
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.042584983288138e-08,
+      "loss": 0.5724,
+      "step": 42194
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.033022954129999e-08,
+      "loss": 0.5837,
+      "step": 42195
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.023465960365408e-08,
+      "loss": 0.5729,
+      "step": 42196
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.013914002042767e-08,
+      "loss": 0.6068,
+      "step": 42197
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.004367079210596e-08,
+      "loss": 0.6429,
+      "step": 42198
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.994825191917411e-08,
+      "loss": 0.6127,
+      "step": 42199
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.98528834021184e-08,
+      "loss": 0.6008,
+      "step": 42200
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.975756524142288e-08,
+      "loss": 0.5889,
+      "step": 42201
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.96622974375716e-08,
+      "loss": 0.6132,
+      "step": 42202
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.956707999104752e-08,
+      "loss": 0.6323,
+      "step": 42203
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.947191290233581e-08,
+      "loss": 0.5851,
+      "step": 42204
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.937679617192053e-08,
+      "loss": 0.559,
+      "step": 42205
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.928172980028238e-08,
+      "loss": 0.5723,
+      "step": 42206
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.918671378790877e-08,
+      "loss": 0.5528,
+      "step": 42207
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.909174813527822e-08,
+      "loss": 0.6131,
+      "step": 42208
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.899683284287585e-08,
+      "loss": 0.6023,
+      "step": 42209
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.890196791118244e-08,
+      "loss": 0.532,
+      "step": 42210
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.880715334068091e-08,
+      "loss": 0.5802,
+      "step": 42211
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.87123891318531e-08,
+      "loss": 0.5777,
+      "step": 42212
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.861767528518083e-08,
+      "loss": 0.5982,
+      "step": 42213
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.852301180114375e-08,
+      "loss": 0.6243,
+      "step": 42214
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.84283986802248e-08,
+      "loss": 0.6083,
+      "step": 42215
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.833383592290246e-08,
+      "loss": 0.6221,
+      "step": 42216
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.823932352965969e-08,
+      "loss": 0.5939,
+      "step": 42217
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.814486150097613e-08,
+      "loss": 0.6188,
+      "step": 42218
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.805044983733135e-08,
+      "loss": 0.6087,
+      "step": 42219
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.79560885392039e-08,
+      "loss": 0.6151,
+      "step": 42220
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.786177760707559e-08,
+      "loss": 0.5718,
+      "step": 42221
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.776751704142494e-08,
+      "loss": 0.5941,
+      "step": 42222
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.767330684272824e-08,
+      "loss": 0.6586,
+      "step": 42223
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.757914701146952e-08,
+      "loss": 0.6291,
+      "step": 42224
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.748503754812176e-08,
+      "loss": 0.6341,
+      "step": 42225
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.73909784531668e-08,
+      "loss": 0.614,
+      "step": 42226
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.729696972707979e-08,
+      "loss": 0.6178,
+      "step": 42227
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.720301137034148e-08,
+      "loss": 0.639,
+      "step": 42228
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.710910338342705e-08,
+      "loss": 0.5893,
+      "step": 42229
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.701524576681497e-08,
+      "loss": 0.6351,
+      "step": 42230
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.692143852098045e-08,
+      "loss": 0.5677,
+      "step": 42231
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.682768164640198e-08,
+      "loss": 0.6072,
+      "step": 42232
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.673397514355475e-08,
+      "loss": 0.6374,
+      "step": 42233
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.664031901291725e-08,
+      "loss": 0.5925,
+      "step": 42234
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.654671325496133e-08,
+      "loss": 0.6039,
+      "step": 42235
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.645315787016662e-08,
+      "loss": 0.5767,
+      "step": 42236
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.635965285900493e-08,
+      "loss": 0.579,
+      "step": 42237
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.626619822195481e-08,
+      "loss": 0.6218,
+      "step": 42238
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.617279395948919e-08,
+      "loss": 0.6321,
+      "step": 42239
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.607944007208213e-08,
+      "loss": 0.4837,
+      "step": 42240
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.598613656020993e-08,
+      "loss": 0.5966,
+      "step": 42241
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.589288342434443e-08,
+      "loss": 0.6137,
+      "step": 42242
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.579968066496191e-08,
+      "loss": 0.5903,
+      "step": 42243
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.570652828253312e-08,
+      "loss": 0.6015,
+      "step": 42244
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 8.561342627753544e-08,
+      "loss": 0.6035,
+      "step": 42245
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.552037465043739e-08,
+      "loss": 0.5824,
+      "step": 42246
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.542737340171525e-08,
+      "loss": 0.5906,
+      "step": 42247
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.533442253183866e-08,
+      "loss": 0.6169,
+      "step": 42248
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.524152204128277e-08,
+      "loss": 0.6696,
+      "step": 42249
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.514867193051723e-08,
+      "loss": 0.5869,
+      "step": 42250
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.505587220001721e-08,
+      "loss": 0.6076,
+      "step": 42251
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.496312285025011e-08,
+      "loss": 0.5661,
+      "step": 42252
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.487042388168998e-08,
+      "loss": 0.6846,
+      "step": 42253
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.477777529480647e-08,
+      "loss": 0.6088,
+      "step": 42254
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.46851770900703e-08,
+      "loss": 0.6165,
+      "step": 42255
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.459262926795441e-08,
+      "loss": 0.5595,
+      "step": 42256
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.450013182892625e-08,
+      "loss": 0.5941,
+      "step": 42257
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.44076847734565e-08,
+      "loss": 0.6317,
+      "step": 42258
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.431528810201483e-08,
+      "loss": 0.5776,
+      "step": 42259
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.422294181507196e-08,
+      "loss": 0.5745,
+      "step": 42260
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.413064591309639e-08,
+      "loss": 0.5811,
+      "step": 42261
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.403840039655775e-08,
+      "loss": 0.6057,
+      "step": 42262
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.394620526592235e-08,
+      "loss": 0.6348,
+      "step": 42263
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.38540605216609e-08,
+      "loss": 0.6161,
+      "step": 42264
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.376196616424082e-08,
+      "loss": 0.6321,
+      "step": 42265
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.366992219413061e-08,
+      "loss": 0.5803,
+      "step": 42266
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.35779286117977e-08,
+      "loss": 0.589,
+      "step": 42267
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.348598541770948e-08,
+      "loss": 0.6377,
+      "step": 42268
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.339409261233333e-08,
+      "loss": 0.6033,
+      "step": 42269
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.330225019613558e-08,
+      "loss": 0.6071,
+      "step": 42270
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.321045816958362e-08,
+      "loss": 0.5424,
+      "step": 42271
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.311871653314375e-08,
+      "loss": 0.5813,
+      "step": 42272
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.302702528728113e-08,
+      "loss": 0.6331,
+      "step": 42273
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.293538443246318e-08,
+      "loss": 0.5344,
+      "step": 42274
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.28437939691551e-08,
+      "loss": 0.6144,
+      "step": 42275
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.275225389782204e-08,
+      "loss": 0.5541,
+      "step": 42276
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.266076421892921e-08,
+      "loss": 0.6181,
+      "step": 42277
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.256932493294067e-08,
+      "loss": 0.568,
+      "step": 42278
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.247793604032272e-08,
+      "loss": 0.6393,
+      "step": 42279
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.238659754153833e-08,
+      "loss": 0.6048,
+      "step": 42280
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.229530943705266e-08,
+      "loss": 0.5869,
+      "step": 42281
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.220407172732758e-08,
+      "loss": 0.6035,
+      "step": 42282
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.21128844128305e-08,
+      "loss": 0.5873,
+      "step": 42283
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.202174749402103e-08,
+      "loss": 0.5641,
+      "step": 42284
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.193066097136438e-08,
+      "loss": 0.515,
+      "step": 42285
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.183962484532126e-08,
+      "loss": 0.6175,
+      "step": 42286
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.174863911635689e-08,
+      "loss": 0.6092,
+      "step": 42287
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.165770378493198e-08,
+      "loss": 0.5566,
+      "step": 42288
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.156681885150952e-08,
+      "loss": 0.6204,
+      "step": 42289
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.147598431655024e-08,
+      "loss": 0.6183,
+      "step": 42290
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.13852001805171e-08,
+      "loss": 0.5821,
+      "step": 42291
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.129446644386862e-08,
+      "loss": 0.6183,
+      "step": 42292
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.120378310707e-08,
+      "loss": 0.624,
+      "step": 42293
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.111315017057864e-08,
+      "loss": 0.6272,
+      "step": 42294
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.102256763485638e-08,
+      "loss": 0.6241,
+      "step": 42295
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.093203550036288e-08,
+      "loss": 0.5924,
+      "step": 42296
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.084155376755997e-08,
+      "loss": 0.6654,
+      "step": 42297
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.075112243690509e-08,
+      "loss": 0.5975,
+      "step": 42298
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.066074150885894e-08,
+      "loss": 0.5982,
+      "step": 42299
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.057041098388118e-08,
+      "loss": 0.6487,
+      "step": 42300
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.048013086242923e-08,
+      "loss": 0.6107,
+      "step": 42301
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.03899011449638e-08,
+      "loss": 0.6374,
+      "step": 42302
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.029972183194123e-08,
+      "loss": 0.6444,
+      "step": 42303
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.020959292382113e-08,
+      "loss": 0.5897,
+      "step": 42304
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.011951442106203e-08,
+      "loss": 0.6509,
+      "step": 42305
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 8.002948632412023e-08,
+      "loss": 0.6248,
+      "step": 42306
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.993950863345423e-08,
+      "loss": 0.6018,
+      "step": 42307
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.984958134952037e-08,
+      "loss": 0.5643,
+      "step": 42308
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.975970447277604e-08,
+      "loss": 0.6869,
+      "step": 42309
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.966987800367864e-08,
+      "loss": 0.6437,
+      "step": 42310
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.95801019426834e-08,
+      "loss": 0.547,
+      "step": 42311
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.949037629024658e-08,
+      "loss": 0.5492,
+      "step": 42312
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.940070104682451e-08,
+      "loss": 0.5852,
+      "step": 42313
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.931107621287349e-08,
+      "loss": 0.6557,
+      "step": 42314
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.922150178884758e-08,
+      "loss": 0.6322,
+      "step": 42315
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.91319777752042e-08,
+      "loss": 0.6634,
+      "step": 42316
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.904250417239523e-08,
+      "loss": 0.5691,
+      "step": 42317
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.895308098087696e-08,
+      "loss": 0.5966,
+      "step": 42318
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.886370820110345e-08,
+      "loss": 0.5894,
+      "step": 42319
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.877438583352993e-08,
+      "loss": 0.5968,
+      "step": 42320
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.868511387860822e-08,
+      "loss": 0.6138,
+      "step": 42321
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.859589233679355e-08,
+      "loss": 0.5774,
+      "step": 42322
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.850672120853886e-08,
+      "loss": 0.6206,
+      "step": 42323
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.841760049429714e-08,
+      "loss": 0.616,
+      "step": 42324
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.832853019452136e-08,
+      "loss": 0.6039,
+      "step": 42325
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.823951030966337e-08,
+      "loss": 0.6483,
+      "step": 42326
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.815054084017726e-08,
+      "loss": 0.5505,
+      "step": 42327
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.806162178651267e-08,
+      "loss": 0.6344,
+      "step": 42328
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.797275314912367e-08,
+      "loss": 0.6011,
+      "step": 42329
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.788393492846103e-08,
+      "loss": 0.5509,
+      "step": 42330
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.779516712497548e-08,
+      "loss": 0.6179,
+      "step": 42331
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.77064497391189e-08,
+      "loss": 0.5998,
+      "step": 42332
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.761778277134091e-08,
+      "loss": 0.6266,
+      "step": 42333
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.752916622209339e-08,
+      "loss": 0.5913,
+      "step": 42334
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.744060009182597e-08,
+      "loss": 0.5723,
+      "step": 42335
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.73520843809894e-08,
+      "loss": 0.5655,
+      "step": 42336
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.726361909003332e-08,
+      "loss": 0.5774,
+      "step": 42337
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.717520421940516e-08,
+      "loss": 0.5798,
+      "step": 42338
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.708683976955678e-08,
+      "loss": 0.6106,
+      "step": 42339
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.69985257409367e-08,
+      "loss": 0.6191,
+      "step": 42340
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.691026213399122e-08,
+      "loss": 0.6011,
+      "step": 42341
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.682204894917333e-08,
+      "loss": 0.6124,
+      "step": 42342
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.673388618692712e-08,
+      "loss": 0.5932,
+      "step": 42343
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.664577384770333e-08,
+      "loss": 0.6004,
+      "step": 42344
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.655771193194717e-08,
+      "loss": 0.5563,
+      "step": 42345
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.646970044010827e-08,
+      "loss": 0.6568,
+      "step": 42346
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.638173937263294e-08,
+      "loss": 0.5793,
+      "step": 42347
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.62938287299686e-08,
+      "loss": 0.5859,
+      "step": 42348
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.620596851256157e-08,
+      "loss": 0.5685,
+      "step": 42349
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.611815872085815e-08,
+      "loss": 0.5324,
+      "step": 42350
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.603039935530354e-08,
+      "loss": 0.6233,
+      "step": 42351
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.594269041634628e-08,
+      "loss": 0.6057,
+      "step": 42352
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.585503190443045e-08,
+      "loss": 0.6378,
+      "step": 42353
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.576742382000125e-08,
+      "loss": 0.5982,
+      "step": 42354
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.567986616350387e-08,
+      "loss": 0.6418,
+      "step": 42355
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.559235893538463e-08,
+      "loss": 0.5773,
+      "step": 42356
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.550490213608653e-08,
+      "loss": 0.6422,
+      "step": 42357
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.541749576605473e-08,
+      "loss": 0.6232,
+      "step": 42358
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.533013982573334e-08,
+      "loss": 0.6071,
+      "step": 42359
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.524283431556534e-08,
+      "loss": 0.541,
+      "step": 42360
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.515557923599592e-08,
+      "loss": 0.541,
+      "step": 42361
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.506837458746697e-08,
+      "loss": 0.5881,
+      "step": 42362
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.498122037042255e-08,
+      "loss": 0.6185,
+      "step": 42363
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.489411658530565e-08,
+      "loss": 0.6551,
+      "step": 42364
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.480706323255816e-08,
+      "loss": 0.6093,
+      "step": 42365
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.472006031262192e-08,
+      "loss": 0.5627,
+      "step": 42366
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.463310782594101e-08,
+      "loss": 0.6399,
+      "step": 42367
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.454620577295513e-08,
+      "loss": 0.719,
+      "step": 42368
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.445935415410721e-08,
+      "loss": 0.5661,
+      "step": 42369
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.437255296983802e-08,
+      "loss": 0.565,
+      "step": 42370
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.428580222058946e-08,
+      "loss": 0.5678,
+      "step": 42371
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.419910190680003e-08,
+      "loss": 0.6229,
+      "step": 42372
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.411245202891271e-08,
+      "loss": 0.5868,
+      "step": 42373
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.402585258736716e-08,
+      "loss": 0.6569,
+      "step": 42374
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.393930358260415e-08,
+      "loss": 0.5715,
+      "step": 42375
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.385280501506109e-08,
+      "loss": 0.5597,
+      "step": 42376
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.376635688517875e-08,
+      "loss": 0.6253,
+      "step": 42377
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.367995919339788e-08,
+      "loss": 0.6293,
+      "step": 42378
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.359361194015479e-08,
+      "loss": 0.619,
+      "step": 42379
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.350731512589026e-08,
+      "loss": 0.5899,
+      "step": 42380
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.342106875104282e-08,
+      "loss": 0.5927,
+      "step": 42381
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.333487281604879e-08,
+      "loss": 0.6538,
+      "step": 42382
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.32487273213478e-08,
+      "loss": 0.6613,
+      "step": 42383
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.316263226737841e-08,
+      "loss": 0.6263,
+      "step": 42384
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.307658765457581e-08,
+      "loss": 0.5808,
+      "step": 42385
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.299059348337856e-08,
+      "loss": 0.5344,
+      "step": 42386
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.290464975422295e-08,
+      "loss": 0.6176,
+      "step": 42387
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.281875646754644e-08,
+      "loss": 0.58,
+      "step": 42388
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.273291362378531e-08,
+      "loss": 0.6026,
+      "step": 42389
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.264712122337481e-08,
+      "loss": 0.5962,
+      "step": 42390
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.256137926675234e-08,
+      "loss": 0.5177,
+      "step": 42391
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.24756877543531e-08,
+      "loss": 0.5987,
+      "step": 42392
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.239004668661232e-08,
+      "loss": 0.6227,
+      "step": 42393
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.230445606396407e-08,
+      "loss": 0.5864,
+      "step": 42394
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.221891588684581e-08,
+      "loss": 0.6269,
+      "step": 42395
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.213342615568942e-08,
+      "loss": 0.5768,
+      "step": 42396
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.204798687093117e-08,
+      "loss": 0.5986,
+      "step": 42397
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.19625980330052e-08,
+      "loss": 0.5653,
+      "step": 42398
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.187725964234449e-08,
+      "loss": 0.5813,
+      "step": 42399
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.179197169938201e-08,
+      "loss": 0.6388,
+      "step": 42400
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.170673420455299e-08,
+      "loss": 0.6364,
+      "step": 42401
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.16215471582904e-08,
+      "loss": 0.5673,
+      "step": 42402
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.153641056102611e-08,
+      "loss": 0.642,
+      "step": 42403
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.145132441319202e-08,
+      "loss": 0.6155,
+      "step": 42404
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.136628871522222e-08,
+      "loss": 0.5645,
+      "step": 42405
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.128130346754636e-08,
+      "loss": 0.5952,
+      "step": 42406
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.119636867059965e-08,
+      "loss": 0.607,
+      "step": 42407
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.111148432481174e-08,
+      "loss": 0.6128,
+      "step": 42408
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.102665043061452e-08,
+      "loss": 0.6306,
+      "step": 42409
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.094186698843875e-08,
+      "loss": 0.6281,
+      "step": 42410
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.085713399871408e-08,
+      "loss": 0.6297,
+      "step": 42411
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.077245146187239e-08,
+      "loss": 0.5921,
+      "step": 42412
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.068781937834557e-08,
+      "loss": 0.6771,
+      "step": 42413
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.060323774855993e-08,
+      "loss": 0.62,
+      "step": 42414
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.051870657294845e-08,
+      "loss": 0.6496,
+      "step": 42415
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.04342258519386e-08,
+      "loss": 0.5878,
+      "step": 42416
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.034979558596112e-08,
+      "loss": 0.6197,
+      "step": 42417
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.026541577544344e-08,
+      "loss": 0.6227,
+      "step": 42418
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.018108642081634e-08,
+      "loss": 0.6235,
+      "step": 42419
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.009680752250614e-08,
+      "loss": 0.5731,
+      "step": 42420
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.00125790809425e-08,
+      "loss": 0.6403,
+      "step": 42421
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.992840109655396e-08,
+      "loss": 0.6614,
+      "step": 42422
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.984427356976575e-08,
+      "loss": 0.6464,
+      "step": 42423
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.976019650100863e-08,
+      "loss": 0.5822,
+      "step": 42424
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.967616989070669e-08,
+      "loss": 0.6063,
+      "step": 42425
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.95921937392896e-08,
+      "loss": 0.6216,
+      "step": 42426
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.950826804718146e-08,
+      "loss": 0.6397,
+      "step": 42427
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.942439281481195e-08,
+      "loss": 0.5644,
+      "step": 42428
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.934056804260403e-08,
+      "loss": 0.6069,
+      "step": 42429
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.925679373098514e-08,
+      "loss": 0.6009,
+      "step": 42430
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.917306988038053e-08,
+      "loss": 0.5729,
+      "step": 42431
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.908939649121759e-08,
+      "loss": 0.5114,
+      "step": 42432
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.900577356391825e-08,
+      "loss": 0.6074,
+      "step": 42433
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.892220109890879e-08,
+      "loss": 0.6393,
+      "step": 42434
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.883867909661557e-08,
+      "loss": 0.6127,
+      "step": 42435
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.875520755746046e-08,
+      "loss": 0.5715,
+      "step": 42436
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.867178648186868e-08,
+      "loss": 0.6346,
+      "step": 42437
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.858841587026432e-08,
+      "loss": 0.6131,
+      "step": 42438
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.850509572307041e-08,
+      "loss": 0.6914,
+      "step": 42439
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.842182604071102e-08,
+      "loss": 0.6009,
+      "step": 42440
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.833860682360915e-08,
+      "loss": 0.6277,
+      "step": 42441
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.825543807218671e-08,
+      "loss": 0.6265,
+      "step": 42442
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.817231978686778e-08,
+      "loss": 0.6454,
+      "step": 42443
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.808925196807314e-08,
+      "loss": 0.6649,
+      "step": 42444
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.800623461622691e-08,
+      "loss": 0.6366,
+      "step": 42445
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.792326773174873e-08,
+      "loss": 0.6198,
+      "step": 42446
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.784035131506161e-08,
+      "loss": 0.5503,
+      "step": 42447
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.775748536658633e-08,
+      "loss": 0.6216,
+      "step": 42448
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.767466988674364e-08,
+      "loss": 0.6159,
+      "step": 42449
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.759190487595547e-08,
+      "loss": 0.5916,
+      "step": 42450
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.750919033464142e-08,
+      "loss": 0.5836,
+      "step": 42451
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.742652626322233e-08,
+      "loss": 0.5719,
+      "step": 42452
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.734391266211782e-08,
+      "loss": 0.6289,
+      "step": 42453
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.726134953174757e-08,
+      "loss": 0.6106,
+      "step": 42454
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.717883687253124e-08,
+      "loss": 0.611,
+      "step": 42455
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.709637468488739e-08,
+      "loss": 0.61,
+      "step": 42456
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.70139629692379e-08,
+      "loss": 0.6717,
+      "step": 42457
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.6931601725998e-08,
+      "loss": 0.6156,
+      "step": 42458
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.684929095558734e-08,
+      "loss": 0.6035,
+      "step": 42459
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.67670306584256e-08,
+      "loss": 0.6317,
+      "step": 42460
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.668482083492911e-08,
+      "loss": 0.61,
+      "step": 42461
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.660266148551642e-08,
+      "loss": 0.5745,
+      "step": 42462
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.652055261060498e-08,
+      "loss": 0.5886,
+      "step": 42463
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.643849421061111e-08,
+      "loss": 0.5559,
+      "step": 42464
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.635648628595227e-08,
+      "loss": 0.5737,
+      "step": 42465
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.627452883704588e-08,
+      "loss": 0.6195,
+      "step": 42466
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.619262186430831e-08,
+      "loss": 0.6224,
+      "step": 42467
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.611076536815475e-08,
+      "loss": 0.5602,
+      "step": 42468
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.602895934900044e-08,
+      "loss": 0.6188,
+      "step": 42469
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.594720380726394e-08,
+      "loss": 0.5996,
+      "step": 42470
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.586549874335824e-08,
+      "loss": 0.6206,
+      "step": 42471
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.578384415769967e-08,
+      "loss": 0.6196,
+      "step": 42472
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.570224005070236e-08,
+      "loss": 0.6111,
+      "step": 42473
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.562068642278263e-08,
+      "loss": 0.6229,
+      "step": 42474
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.553918327435237e-08,
+      "loss": 0.599,
+      "step": 42475
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.545773060582683e-08,
+      "loss": 0.6276,
+      "step": 42476
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.537632841762009e-08,
+      "loss": 0.5896,
+      "step": 42477
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.529497671014629e-08,
+      "loss": 0.6052,
+      "step": 42478
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.52136754838184e-08,
+      "loss": 0.6352,
+      "step": 42479
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.513242473904835e-08,
+      "loss": 0.6022,
+      "step": 42480
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.505122447625134e-08,
+      "loss": 0.6005,
+      "step": 42481
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.497007469583705e-08,
+      "loss": 0.5429,
+      "step": 42482
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.48889753982207e-08,
+      "loss": 0.5909,
+      "step": 42483
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.480792658381307e-08,
+      "loss": 0.617,
+      "step": 42484
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.472692825302495e-08,
+      "loss": 0.5745,
+      "step": 42485
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.464598040626935e-08,
+      "loss": 0.5477,
+      "step": 42486
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.456508304395814e-08,
+      "loss": 0.6216,
+      "step": 42487
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.448423616649991e-08,
+      "loss": 0.6153,
+      "step": 42488
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.440343977430874e-08,
+      "loss": 0.5822,
+      "step": 42489
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.432269386779322e-08,
+      "loss": 0.5751,
+      "step": 42490
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.424199844736301e-08,
+      "loss": 0.5975,
+      "step": 42491
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.416135351342889e-08,
+      "loss": 0.5696,
+      "step": 42492
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.408075906640165e-08,
+      "loss": 0.5591,
+      "step": 42493
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.400021510668985e-08,
+      "loss": 0.606,
+      "step": 42494
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.391972163470206e-08,
+      "loss": 0.5963,
+      "step": 42495
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.383927865084905e-08,
+      "loss": 0.596,
+      "step": 42496
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.375888615553827e-08,
+      "loss": 0.6139,
+      "step": 42497
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.367854414917829e-08,
+      "loss": 0.5683,
+      "step": 42498
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.359825263217767e-08,
+      "loss": 0.6476,
+      "step": 42499
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.351801160494498e-08,
+      "loss": 0.5873,
+      "step": 42500
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.343782106788654e-08,
+      "loss": 0.5752,
+      "step": 42501
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.335768102141093e-08,
+      "loss": 0.5939,
+      "step": 42502
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.327759146592449e-08,
+      "loss": 0.5586,
+      "step": 42503
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.319755240183468e-08,
+      "loss": 0.6604,
+      "step": 42504
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.311756382954892e-08,
+      "loss": 0.5806,
+      "step": 42505
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.303762574947247e-08,
+      "loss": 0.6132,
+      "step": 42506
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.295773816201279e-08,
+      "loss": 0.5911,
+      "step": 42507
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.287790106757396e-08,
+      "loss": 0.6304,
+      "step": 42508
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.279811446656348e-08,
+      "loss": 0.6351,
+      "step": 42509
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.271837835938655e-08,
+      "loss": 0.6514,
+      "step": 42510
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.263869274644619e-08,
+      "loss": 0.5704,
+      "step": 42511
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.255905762814985e-08,
+      "loss": 0.5892,
+      "step": 42512
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.247947300490164e-08,
+      "loss": 0.5627,
+      "step": 42513
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.23999388771046e-08,
+      "loss": 0.5786,
+      "step": 42514
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.232045524516506e-08,
+      "loss": 0.6221,
+      "step": 42515
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.22410221094849e-08,
+      "loss": 0.6489,
+      "step": 42516
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.216163947046828e-08,
+      "loss": 0.6336,
+      "step": 42517
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.208230732851928e-08,
+      "loss": 0.6199,
+      "step": 42518
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.200302568404093e-08,
+      "loss": 0.6162,
+      "step": 42519
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.192379453743403e-08,
+      "loss": 0.6006,
+      "step": 42520
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.18446138891049e-08,
+      "loss": 0.6219,
+      "step": 42521
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.176548373945213e-08,
+      "loss": 0.6326,
+      "step": 42522
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.168640408888094e-08,
+      "loss": 0.5565,
+      "step": 42523
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.16073749377899e-08,
+      "loss": 0.5641,
+      "step": 42524
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.152839628658424e-08,
+      "loss": 0.6313,
+      "step": 42525
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.144946813566254e-08,
+      "loss": 0.6007,
+      "step": 42526
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.13705904854267e-08,
+      "loss": 0.6061,
+      "step": 42527
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.129176333627751e-08,
+      "loss": 0.5521,
+      "step": 42528
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.121298668861575e-08,
+      "loss": 0.5412,
+      "step": 42529
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.11342605428411e-08,
+      "loss": 0.6241,
+      "step": 42530
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.105558489935437e-08,
+      "loss": 0.5869,
+      "step": 42531
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.097695975855522e-08,
+      "loss": 0.6377,
+      "step": 42532
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.089838512084335e-08,
+      "loss": 0.6261,
+      "step": 42533
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.08198609866173e-08,
+      "loss": 0.6123,
+      "step": 42534
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.074138735627567e-08,
+      "loss": 0.5772,
+      "step": 42535
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.066296423021811e-08,
+      "loss": 0.5719,
+      "step": 42536
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.058459160884323e-08,
+      "loss": 0.602,
+      "step": 42537
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.050626949254845e-08,
+      "loss": 0.5974,
+      "step": 42538
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 6.042799788173237e-08,
+      "loss": 0.6172,
+      "step": 42539
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.034977677679244e-08,
+      "loss": 0.5998,
+      "step": 42540
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.02716061781261e-08,
+      "loss": 0.714,
+      "step": 42541
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.019348608613195e-08,
+      "loss": 0.652,
+      "step": 42542
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.011541650120412e-08,
+      "loss": 0.6357,
+      "step": 42543
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 6.003739742374226e-08,
+      "loss": 0.5586,
+      "step": 42544
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.995942885414053e-08,
+      "loss": 0.5951,
+      "step": 42545
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.988151079279636e-08,
+      "loss": 0.6122,
+      "step": 42546
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.980364324010501e-08,
+      "loss": 0.608,
+      "step": 42547
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.972582619646284e-08,
+      "loss": 0.5547,
+      "step": 42548
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.964805966226505e-08,
+      "loss": 0.631,
+      "step": 42549
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.9570343637905817e-08,
+      "loss": 0.6187,
+      "step": 42550
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.9492678123781456e-08,
+      "loss": 0.5789,
+      "step": 42551
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.9415063120286106e-08,
+      "loss": 0.671,
+      "step": 42552
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.93374986278139e-08,
+      "loss": 0.631,
+      "step": 42553
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.925998464675897e-08,
+      "loss": 0.5324,
+      "step": 42554
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.918252117751544e-08,
+      "loss": 0.5933,
+      "step": 42555
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.9105108220477435e-08,
+      "loss": 0.5669,
+      "step": 42556
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.902774577603687e-08,
+      "loss": 0.5601,
+      "step": 42557
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.895043384458787e-08,
+      "loss": 0.5471,
+      "step": 42558
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.887317242652346e-08,
+      "loss": 0.5884,
+      "step": 42559
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.8795961522235546e-08,
+      "loss": 0.5955,
+      "step": 42560
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.871880113211714e-08,
+      "loss": 0.6318,
+      "step": 42561
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.8641691256559054e-08,
+      "loss": 0.5809,
+      "step": 42562
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.8564631895954295e-08,
+      "loss": 0.6542,
+      "step": 42563
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.848762305069367e-08,
+      "loss": 0.5984,
+      "step": 42564
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.841066472117018e-08,
+      "loss": 0.6437,
+      "step": 42565
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.833375690777354e-08,
+      "loss": 0.5722,
+      "step": 42566
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.8256899610894536e-08,
+      "loss": 0.6208,
+      "step": 42567
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.818009283092285e-08,
+      "loss": 0.6128,
+      "step": 42568
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.81033365682504e-08,
+      "loss": 0.6005,
+      "step": 42569
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.802663082326687e-08,
+      "loss": 0.5947,
+      "step": 42570
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.794997559636084e-08,
+      "loss": 0.5757,
+      "step": 42571
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.787337088792311e-08,
+      "loss": 0.5979,
+      "step": 42572
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.779681669834225e-08,
+      "loss": 0.5538,
+      "step": 42573
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.772031302800796e-08,
+      "loss": 0.6154,
+      "step": 42574
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.764385987730769e-08,
+      "loss": 0.5835,
+      "step": 42575
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.756745724663115e-08,
+      "loss": 0.5985,
+      "step": 42576
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.74911051363658e-08,
+      "loss": 0.5931,
+      "step": 42577
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.741480354690132e-08,
+      "loss": 0.6104,
+      "step": 42578
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.7338552478621854e-08,
+      "loss": 0.5869,
+      "step": 42579
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.7262351931918205e-08,
+      "loss": 0.5483,
+      "step": 42580
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.718620190717672e-08,
+      "loss": 0.6259,
+      "step": 42581
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.7110102404783764e-08,
+      "loss": 0.6538,
+      "step": 42582
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.70340534251268e-08,
+      "loss": 0.6193,
+      "step": 42583
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.695805496859108e-08,
+      "loss": 0.6808,
+      "step": 42584
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.688210703556296e-08,
+      "loss": 0.5451,
+      "step": 42585
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.680620962642991e-08,
+      "loss": 0.6243,
+      "step": 42586
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6730362741576064e-08,
+      "loss": 0.5925,
+      "step": 42587
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6654566381387775e-08,
+      "loss": 0.5901,
+      "step": 42588
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.657882054624808e-08,
+      "loss": 0.6205,
+      "step": 42589
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.650312523654444e-08,
+      "loss": 0.5854,
+      "step": 42590
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6427480452661e-08,
+      "loss": 0.6068,
+      "step": 42591
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6351886194981886e-08,
+      "loss": 0.5672,
+      "step": 42592
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.627634246389013e-08,
+      "loss": 0.6362,
+      "step": 42593
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6200849259769874e-08,
+      "loss": 0.5939,
+      "step": 42594
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.612540658300636e-08,
+      "loss": 0.57,
+      "step": 42595
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.6050014433981506e-08,
+      "loss": 0.6288,
+      "step": 42596
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.5974672813079444e-08,
+      "loss": 0.5412,
+      "step": 42597
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.589938172068099e-08,
+      "loss": 0.559,
+      "step": 42598
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.582414115717138e-08,
+      "loss": 0.5656,
+      "step": 42599
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.574895112293033e-08,
+      "loss": 0.5713,
+      "step": 42600
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.5673811618341955e-08,
+      "loss": 0.5683,
+      "step": 42601
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.559872264378707e-08,
+      "loss": 0.5732,
+      "step": 42602
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.55236841996476e-08,
+      "loss": 0.62,
+      "step": 42603
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.544869628630434e-08,
+      "loss": 0.6102,
+      "step": 42604
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.5373758904138094e-08,
+      "loss": 0.6122,
+      "step": 42605
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.529887205353079e-08,
+      "loss": 0.6029,
+      "step": 42606
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.522403573486323e-08,
+      "loss": 0.5992,
+      "step": 42607
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.5149249948514e-08,
+      "loss": 0.578,
+      "step": 42608
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.5074514694863914e-08,
+      "loss": 0.5842,
+      "step": 42609
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.499982997429265e-08,
+      "loss": 0.623,
+      "step": 42610
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.492519578717992e-08,
+      "loss": 0.554,
+      "step": 42611
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.48506121339043e-08,
+      "loss": 0.5661,
+      "step": 42612
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.4776079014846604e-08,
+      "loss": 0.6007,
+      "step": 42613
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.47015964303832e-08,
+      "loss": 0.5826,
+      "step": 42614
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.462716438089377e-08,
+      "loss": 0.6219,
+      "step": 42615
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.455278286675581e-08,
+      "loss": 0.5876,
+      "step": 42616
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.447845188834899e-08,
+      "loss": 0.5814,
+      "step": 42617
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.4404171446048596e-08,
+      "loss": 0.5998,
+      "step": 42618
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.43299415402343e-08,
+      "loss": 0.5524,
+      "step": 42619
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.425576217128248e-08,
+      "loss": 0.6156,
+      "step": 42620
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.418163333956838e-08,
+      "loss": 0.6077,
+      "step": 42621
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.410755504547172e-08,
+      "loss": 0.5757,
+      "step": 42622
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.4033527289366616e-08,
+      "loss": 0.6267,
+      "step": 42623
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.395955007163056e-08,
+      "loss": 0.5666,
+      "step": 42624
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3885623392638807e-08,
+      "loss": 0.6031,
+      "step": 42625
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.381174725276661e-08,
+      "loss": 0.6122,
+      "step": 42626
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.373792165238922e-08,
+      "loss": 0.6198,
+      "step": 42627
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3664146591883016e-08,
+      "loss": 0.6227,
+      "step": 42628
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.359042207162213e-08,
+      "loss": 0.5894,
+      "step": 42629
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.351674809198182e-08,
+      "loss": 0.6116,
+      "step": 42630
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.344312465333512e-08,
+      "loss": 0.6079,
+      "step": 42631
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3369551756057294e-08,
+      "loss": 0.564,
+      "step": 42632
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3296029400521364e-08,
+      "loss": 0.551,
+      "step": 42633
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3222557587101486e-08,
+      "loss": 0.6199,
+      "step": 42634
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.31491363161718e-08,
+      "loss": 0.6113,
+      "step": 42635
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.307576558810312e-08,
+      "loss": 0.5644,
+      "step": 42636
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.3002445403270705e-08,
+      "loss": 0.6578,
+      "step": 42637
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.292917576204426e-08,
+      "loss": 0.5907,
+      "step": 42638
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.2855956664799036e-08,
+      "loss": 0.5825,
+      "step": 42639
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.2782788111905846e-08,
+      "loss": 0.4957,
+      "step": 42640
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.270967010373662e-08,
+      "loss": 0.6516,
+      "step": 42641
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.263660264066217e-08,
+      "loss": 0.6171,
+      "step": 42642
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.256358572305553e-08,
+      "loss": 0.6421,
+      "step": 42643
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.24906193512853e-08,
+      "loss": 0.6195,
+      "step": 42644
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.2417703525723394e-08,
+      "loss": 0.6066,
+      "step": 42645
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.234483824674175e-08,
+      "loss": 0.5803,
+      "step": 42646
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.227202351470784e-08,
+      "loss": 0.5846,
+      "step": 42647
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.219925932999359e-08,
+      "loss": 0.5834,
+      "step": 42648
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.212654569296871e-08,
+      "loss": 0.5602,
+      "step": 42649
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.205388260400179e-08,
+      "loss": 0.5535,
+      "step": 42650
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.198127006346254e-08,
+      "loss": 0.601,
+      "step": 42651
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.190870807171955e-08,
+      "loss": 0.6087,
+      "step": 42652
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.183619662914252e-08,
+      "loss": 0.6197,
+      "step": 42653
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.176373573609894e-08,
+      "loss": 0.5681,
+      "step": 42654
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.16913253929574e-08,
+      "loss": 0.5621,
+      "step": 42655
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.1618965600085394e-08,
+      "loss": 0.5895,
+      "step": 42656
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.1546656357851496e-08,
+      "loss": 0.5868,
+      "step": 42657
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.14743976666221e-08,
+      "loss": 0.5976,
+      "step": 42658
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.140218952676468e-08,
+      "loss": 0.6018,
+      "step": 42659
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.133003193864672e-08,
+      "loss": 0.638,
+      "step": 42660
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.1257924902634596e-08,
+      "loss": 0.5582,
+      "step": 42661
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.1185868419094674e-08,
+      "loss": 0.6109,
+      "step": 42662
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.111386248839223e-08,
+      "loss": 0.5596,
+      "step": 42663
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.104190711089474e-08,
+      "loss": 0.5967,
+      "step": 42664
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.097000228696636e-08,
+      "loss": 0.5897,
+      "step": 42665
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.089814801697346e-08,
+      "loss": 0.6376,
+      "step": 42666
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.0826344301280195e-08,
+      "loss": 0.5064,
+      "step": 42667
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.075459114025294e-08,
+      "loss": 0.614,
+      "step": 42668
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.068288853425474e-08,
+      "loss": 0.5992,
+      "step": 42669
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.061123648365085e-08,
+      "loss": 0.5911,
+      "step": 42670
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.053963498880543e-08,
+      "loss": 0.6201,
+      "step": 42671
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.0468084050081524e-08,
+      "loss": 0.6448,
+      "step": 42672
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.039658366784328e-08,
+      "loss": 0.6053,
+      "step": 42673
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.032513384245374e-08,
+      "loss": 0.6113,
+      "step": 42674
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.025373457427596e-08,
+      "loss": 0.5778,
+      "step": 42675
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.018238586367297e-08,
+      "loss": 0.5898,
+      "step": 42676
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.011108771100781e-08,
+      "loss": 0.6071,
+      "step": 42677
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 5.003984011664131e-08,
+      "loss": 0.619,
+      "step": 42678
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.996864308093652e-08,
+      "loss": 0.5762,
+      "step": 42679
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.989749660425536e-08,
+      "loss": 0.5759,
+      "step": 42680
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.982640068695976e-08,
+      "loss": 0.5749,
+      "step": 42681
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.975535532940834e-08,
+      "loss": 0.6061,
+      "step": 42682
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.968436053196524e-08,
+      "loss": 0.621,
+      "step": 42683
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.961341629499017e-08,
+      "loss": 0.5523,
+      "step": 42684
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.954252261884285e-08,
+      "loss": 0.5789,
+      "step": 42685
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.947167950388409e-08,
+      "loss": 0.553,
+      "step": 42686
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.940088695047474e-08,
+      "loss": 0.6414,
+      "step": 42687
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.933014495897337e-08,
+      "loss": 0.6038,
+      "step": 42688
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.925945352973971e-08,
+      "loss": 0.604,
+      "step": 42689
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.918881266313236e-08,
+      "loss": 0.5956,
+      "step": 42690
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.9118222359511026e-08,
+      "loss": 0.6078,
+      "step": 42691
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.9047682619234317e-08,
+      "loss": 0.5534,
+      "step": 42692
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.897719344266083e-08,
+      "loss": 0.6229,
+      "step": 42693
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.8906754830148064e-08,
+      "loss": 0.5851,
+      "step": 42694
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.883636678205572e-08,
+      "loss": 0.6179,
+      "step": 42695
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.8766029298739085e-08,
+      "loss": 0.6267,
+      "step": 42696
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.869574238055674e-08,
+      "loss": 0.608,
+      "step": 42697
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.862550602786509e-08,
+      "loss": 0.618,
+      "step": 42698
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.855532024102161e-08,
+      "loss": 0.6035,
+      "step": 42699
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.84851850203849e-08,
+      "loss": 0.608,
+      "step": 42700
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.841510036630803e-08,
+      "loss": 0.6025,
+      "step": 42701
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.834506627914848e-08,
+      "loss": 0.5804,
+      "step": 42702
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.8275082759261514e-08,
+      "loss": 0.6114,
+      "step": 42703
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.820514980700464e-08,
+      "loss": 0.5908,
+      "step": 42704
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.813526742273089e-08,
+      "loss": 0.5832,
+      "step": 42705
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.806543560679666e-08,
+      "loss": 0.6419,
+      "step": 42706
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7995654359556107e-08,
+      "loss": 0.5625,
+      "step": 42707
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.792592368136562e-08,
+      "loss": 0.5893,
+      "step": 42708
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.785624357257712e-08,
+      "loss": 0.6248,
+      "step": 42709
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.77866140335459e-08,
+      "loss": 0.6124,
+      "step": 42710
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.771703506462499e-08,
+      "loss": 0.6038,
+      "step": 42711
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7647506666168575e-08,
+      "loss": 0.5638,
+      "step": 42712
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.75780288385308e-08,
+      "loss": 0.5651,
+      "step": 42713
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.75086015820625e-08,
+      "loss": 0.6648,
+      "step": 42714
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.743922489711894e-08,
+      "loss": 0.6008,
+      "step": 42715
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.736989878405096e-08,
+      "loss": 0.5721,
+      "step": 42716
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7300623243211604e-08,
+      "loss": 0.6305,
+      "step": 42717
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7231398274951715e-08,
+      "loss": 0.5556,
+      "step": 42718
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7162223879624324e-08,
+      "loss": 0.5964,
+      "step": 42719
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.7093100057580276e-08,
+      "loss": 0.5956,
+      "step": 42720
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.702402680917151e-08,
+      "loss": 0.6091,
+      "step": 42721
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.695500413474774e-08,
+      "loss": 0.6785,
+      "step": 42722
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.688603203466091e-08,
+      "loss": 0.5834,
+      "step": 42723
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.6817110509261854e-08,
+      "loss": 0.6077,
+      "step": 42724
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.6748239558899175e-08,
+      "loss": 0.6513,
+      "step": 42725
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.667941918392371e-08,
+      "loss": 0.622,
+      "step": 42726
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.661064938468518e-08,
+      "loss": 0.6278,
+      "step": 42727
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.654193016153219e-08,
+      "loss": 0.5861,
+      "step": 42728
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.647326151481557e-08,
+      "loss": 0.5556,
+      "step": 42729
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.640464344488393e-08,
+      "loss": 0.6313,
+      "step": 42730
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.633607595208367e-08,
+      "loss": 0.6392,
+      "step": 42731
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.626755903676561e-08,
+      "loss": 0.6205,
+      "step": 42732
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.6199092699277246e-08,
+      "loss": 0.6193,
+      "step": 42733
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.61306769399672e-08,
+      "loss": 0.6047,
+      "step": 42734
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.6062311759180744e-08,
+      "loss": 0.6528,
+      "step": 42735
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.59939971572676e-08,
+      "loss": 0.5731,
+      "step": 42736
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.592573313457416e-08,
+      "loss": 0.5677,
+      "step": 42737
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.585751969144792e-08,
+      "loss": 0.6222,
+      "step": 42738
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.578935682823416e-08,
+      "loss": 0.5528,
+      "step": 42739
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.572124454527926e-08,
+      "loss": 0.5848,
+      "step": 42740
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.5653182842930744e-08,
+      "loss": 0.5834,
+      "step": 42741
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.558517172153387e-08,
+      "loss": 0.5987,
+      "step": 42742
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.551721118143393e-08,
+      "loss": 0.6268,
+      "step": 42743
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.5449301222976194e-08,
+      "loss": 0.5929,
+      "step": 42744
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.5381441846504835e-08,
+      "loss": 0.5358,
+      "step": 42745
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.531363305236736e-08,
+      "loss": 0.6202,
+      "step": 42746
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.52458748409057e-08,
+      "loss": 0.6986,
+      "step": 42747
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.517816721246515e-08,
+      "loss": 0.6201,
+      "step": 42748
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.511051016738988e-08,
+      "loss": 0.5305,
+      "step": 42749
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.5042903706024045e-08,
+      "loss": 0.5621,
+      "step": 42750
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.497534782871072e-08,
+      "loss": 0.573,
+      "step": 42751
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.490784253579183e-08,
+      "loss": 0.5912,
+      "step": 42752
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.484038782761268e-08,
+      "loss": 0.6402,
+      "step": 42753
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.477298370451522e-08,
+      "loss": 0.5598,
+      "step": 42754
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4705630166841375e-08,
+      "loss": 0.6181,
+      "step": 42755
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.463832721493311e-08,
+      "loss": 0.6452,
+      "step": 42756
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.457107484913459e-08,
+      "loss": 0.6516,
+      "step": 42757
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.450387306978443e-08,
+      "loss": 0.6638,
+      "step": 42758
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.443672187722681e-08,
+      "loss": 0.6308,
+      "step": 42759
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4369621271801446e-08,
+      "loss": 0.6115,
+      "step": 42760
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.430257125384918e-08,
+      "loss": 0.5613,
+      "step": 42761
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4235571823711965e-08,
+      "loss": 0.5907,
+      "step": 42762
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4168622981729525e-08,
+      "loss": 0.6546,
+      "step": 42763
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.410172472824159e-08,
+      "loss": 0.6374,
+      "step": 42764
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.4034877063589e-08,
+      "loss": 0.6124,
+      "step": 42765
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.396807998811037e-08,
+      "loss": 0.6098,
+      "step": 42766
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3901333502145425e-08,
+      "loss": 0.622,
+      "step": 42767
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3834637606033904e-08,
+      "loss": 0.6292,
+      "step": 42768
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.376799230011442e-08,
+      "loss": 0.6298,
+      "step": 42769
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.370139758472669e-08,
+      "loss": 0.5939,
+      "step": 42770
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3634853460207125e-08,
+      "loss": 0.6244,
+      "step": 42771
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3568359926894346e-08,
+      "loss": 0.5742,
+      "step": 42772
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.350191698512696e-08,
+      "loss": 0.5627,
+      "step": 42773
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3435524635241365e-08,
+      "loss": 0.5003,
+      "step": 42774
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.336918287757619e-08,
+      "loss": 0.595,
+      "step": 42775
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.330289171247004e-08,
+      "loss": 0.6201,
+      "step": 42776
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3236651140255993e-08,
+      "loss": 0.611,
+      "step": 42777
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.317046116127266e-08,
+      "loss": 0.5606,
+      "step": 42778
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.3104321775855326e-08,
+      "loss": 0.5692,
+      "step": 42779
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.303823298434262e-08,
+      "loss": 0.6258,
+      "step": 42780
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.29721947870676e-08,
+      "loss": 0.586,
+      "step": 42781
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2906207184367776e-08,
+      "loss": 0.59,
+      "step": 42782
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2840270176577324e-08,
+      "loss": 0.6163,
+      "step": 42783
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2774383764031536e-08,
+      "loss": 0.6299,
+      "step": 42784
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.270854794706458e-08,
+      "loss": 0.6072,
+      "step": 42785
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.264276272601287e-08,
+      "loss": 0.5704,
+      "step": 42786
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.257702810120834e-08,
+      "loss": 0.6214,
+      "step": 42787
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2511344072986294e-08,
+      "loss": 0.5597,
+      "step": 42788
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.24457106416809e-08,
+      "loss": 0.5687,
+      "step": 42789
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.238012780762413e-08,
+      "loss": 0.6289,
+      "step": 42790
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.231459557115014e-08,
+      "loss": 0.6559,
+      "step": 42791
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2249113932593125e-08,
+      "loss": 0.6312,
+      "step": 42792
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2183682892283916e-08,
+      "loss": 0.623,
+      "step": 42793
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2118302450555596e-08,
+      "loss": 0.6367,
+      "step": 42794
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.205297260774011e-08,
+      "loss": 0.5877,
+      "step": 42795
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.198769336417052e-08,
+      "loss": 0.6224,
+      "step": 42796
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.192246472017769e-08,
+      "loss": 0.5727,
+      "step": 42797
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1857286676092455e-08,
+      "loss": 0.6396,
+      "step": 42798
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.179215923224789e-08,
+      "loss": 0.617,
+      "step": 42799
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1727082388972604e-08,
+      "loss": 0.6076,
+      "step": 42800
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.166205614659857e-08,
+      "loss": 0.615,
+      "step": 42801
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.159708050545663e-08,
+      "loss": 0.6018,
+      "step": 42802
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.153215546587652e-08,
+      "loss": 0.5638,
+      "step": 42803
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1467281028187976e-08,
+      "loss": 0.5428,
+      "step": 42804
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1402457192720734e-08,
+      "loss": 0.5963,
+      "step": 42805
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1337683959803424e-08,
+      "loss": 0.6189,
+      "step": 42806
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.127296132976577e-08,
+      "loss": 0.6322,
+      "step": 42807
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.120828930293752e-08,
+      "loss": 0.6,
+      "step": 42808
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.114366787964619e-08,
+      "loss": 0.5533,
+      "step": 42809
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.10790970602204e-08,
+      "loss": 0.6584,
+      "step": 42810
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.1014576844988776e-08,
+      "loss": 0.5983,
+      "step": 42811
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.095010723427772e-08,
+      "loss": 0.5739,
+      "step": 42812
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0885688228416985e-08,
+      "loss": 0.5795,
+      "step": 42813
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.082131982773185e-08,
+      "loss": 0.6531,
+      "step": 42814
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.075700203255095e-08,
+      "loss": 0.6205,
+      "step": 42815
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0692734843199574e-08,
+      "loss": 0.5663,
+      "step": 42816
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0628518260005245e-08,
+      "loss": 0.5824,
+      "step": 42817
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.056435228329436e-08,
+      "loss": 0.6656,
+      "step": 42818
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.0500236913393334e-08,
+      "loss": 0.5791,
+      "step": 42819
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.043617215062745e-08,
+      "loss": 0.6306,
+      "step": 42820
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.037215799532091e-08,
+      "loss": 0.6115,
+      "step": 42821
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.030819444780121e-08,
+      "loss": 0.5927,
+      "step": 42822
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.024428150839255e-08,
+      "loss": 0.6055,
+      "step": 42823
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.018041917741911e-08,
+      "loss": 0.5753,
+      "step": 42824
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.011660745520618e-08,
+      "loss": 0.6168,
+      "step": 42825
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.005284634207685e-08,
+      "loss": 0.6038,
+      "step": 42826
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.9989135838357505e-08,
+      "loss": 0.5842,
+      "step": 42827
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.992547594436791e-08,
+      "loss": 0.5853,
+      "step": 42828
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.9861866660435566e-08,
+      "loss": 0.6063,
+      "step": 42829
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.9798307986882446e-08,
+      "loss": 0.6141,
+      "step": 42830
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.973479992403051e-08,
+      "loss": 0.6067,
+      "step": 42831
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.967134247220284e-08,
+      "loss": 0.6146,
+      "step": 42832
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.960793563172138e-08,
+      "loss": 0.6638,
+      "step": 42833
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.954457940291034e-08,
+      "loss": 0.5102,
+      "step": 42834
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9481273786089455e-08,
+      "loss": 0.5784,
+      "step": 42835
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.94180187815818e-08,
+      "loss": 0.655,
+      "step": 42836
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.935481438970823e-08,
+      "loss": 0.632,
+      "step": 42837
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.92916606107896e-08,
+      "loss": 0.5945,
+      "step": 42838
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.922855744514786e-08,
+      "loss": 0.6246,
+      "step": 42839
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.916550489310167e-08,
+      "loss": 0.6594,
+      "step": 42840
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.9102502954975196e-08,
+      "loss": 0.5201,
+      "step": 42841
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.903955163108375e-08,
+      "loss": 0.5835,
+      "step": 42842
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8976650921750405e-08,
+      "loss": 0.6446,
+      "step": 42843
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8913800827293793e-08,
+      "loss": 0.5961,
+      "step": 42844
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8851001348033654e-08,
+      "loss": 0.5602,
+      "step": 42845
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8788252484288636e-08,
+      "loss": 0.6431,
+      "step": 42846
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.872555423637736e-08,
+      "loss": 0.6377,
+      "step": 42847
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.866290660461958e-08,
+      "loss": 0.5503,
+      "step": 42848
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8600309589331694e-08,
+      "loss": 0.5702,
+      "step": 42849
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8537763190834575e-08,
+      "loss": 0.6244,
+      "step": 42850
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.847526740944352e-08,
+      "loss": 0.5944,
+      "step": 42851
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8412822245477153e-08,
+      "loss": 0.5712,
+      "step": 42852
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.835042769925301e-08,
+      "loss": 0.5739,
+      "step": 42853
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8288083771087505e-08,
+      "loss": 0.5882,
+      "step": 42854
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.822579046129704e-08,
+      "loss": 0.685,
+      "step": 42855
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.816354777019915e-08,
+      "loss": 0.5982,
+      "step": 42856
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8101355698110244e-08,
+      "loss": 0.594,
+      "step": 42857
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.8039214245345624e-08,
+      "loss": 0.5643,
+      "step": 42858
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7977123412220594e-08,
+      "loss": 0.579,
+      "step": 42859
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7915083199051574e-08,
+      "loss": 0.579,
+      "step": 42860
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.785309360615386e-08,
+      "loss": 0.5968,
+      "step": 42861
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7791154633841643e-08,
+      "loss": 0.5929,
+      "step": 42862
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.772926628243023e-08,
+      "loss": 0.6176,
+      "step": 42863
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.766742855223493e-08,
+      "loss": 0.605,
+      "step": 42864
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.760564144356771e-08,
+      "loss": 0.6215,
+      "step": 42865
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.754390495674498e-08,
+      "loss": 0.6029,
+      "step": 42866
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.748221909207872e-08,
+      "loss": 0.5508,
+      "step": 42867
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.742058384988312e-08,
+      "loss": 0.5324,
+      "step": 42868
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.735899923047126e-08,
+      "loss": 0.6059,
+      "step": 42869
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.729746523415623e-08,
+      "loss": 0.5799,
+      "step": 42870
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.723598186124999e-08,
+      "loss": 0.6381,
+      "step": 42871
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.717454911206564e-08,
+      "loss": 0.6024,
+      "step": 42872
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.711316698691514e-08,
+      "loss": 0.5977,
+      "step": 42873
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7051835486110465e-08,
+      "loss": 0.5716,
+      "step": 42874
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.699055460996359e-08,
+      "loss": 0.6577,
+      "step": 42875
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.692932435878427e-08,
+      "loss": 0.6125,
+      "step": 42876
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.686814473288558e-08,
+      "loss": 0.5787,
+      "step": 42877
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.680701573257839e-08,
+      "loss": 0.6125,
+      "step": 42878
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.674593735817245e-08,
+      "loss": 0.58,
+      "step": 42879
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6684909609977504e-08,
+      "loss": 0.6243,
+      "step": 42880
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6623932488304425e-08,
+      "loss": 0.6147,
+      "step": 42881
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.656300599346296e-08,
+      "loss": 0.6735,
+      "step": 42882
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.650213012576398e-08,
+      "loss": 0.569,
+      "step": 42883
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6441304885515004e-08,
+      "loss": 0.5801,
+      "step": 42884
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.638053027302579e-08,
+      "loss": 0.6402,
+      "step": 42885
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.631980628860387e-08,
+      "loss": 0.5722,
+      "step": 42886
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.625913293256012e-08,
+      "loss": 0.5831,
+      "step": 42887
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6198510205202044e-08,
+      "loss": 0.6487,
+      "step": 42888
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.613793810683719e-08,
+      "loss": 0.566,
+      "step": 42889
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.607741663777309e-08,
+      "loss": 0.627,
+      "step": 42890
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.601694579831727e-08,
+      "loss": 0.5864,
+      "step": 42891
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.595652558877838e-08,
+      "loss": 0.5985,
+      "step": 42892
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5896156009461725e-08,
+      "loss": 0.5765,
+      "step": 42893
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.583583706067595e-08,
+      "loss": 0.6367,
+      "step": 42894
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.577556874272636e-08,
+      "loss": 0.5598,
+      "step": 42895
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.571535105591828e-08,
+      "loss": 0.6202,
+      "step": 42896
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.565518400055923e-08,
+      "loss": 0.5695,
+      "step": 42897
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5595067576953415e-08,
+      "loss": 0.6036,
+      "step": 42898
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5535001785408366e-08,
+      "loss": 0.5994,
+      "step": 42899
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.547498662622828e-08,
+      "loss": 0.5658,
+      "step": 42900
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.541502209971737e-08,
+      "loss": 0.6254,
+      "step": 42901
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.535510820618204e-08,
+      "loss": 0.6295,
+      "step": 42902
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5295244945924286e-08,
+      "loss": 0.5964,
+      "step": 42903
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.5235432319250526e-08,
+      "loss": 0.6309,
+      "step": 42904
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.517567032646385e-08,
+      "loss": 0.6201,
+      "step": 42905
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.511595896786735e-08,
+      "loss": 0.6459,
+      "step": 42906
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.505629824376522e-08,
+      "loss": 0.5703,
+      "step": 42907
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.499668815446056e-08,
+      "loss": 0.6564,
+      "step": 42908
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.493712870025645e-08,
+      "loss": 0.5881,
+      "step": 42909
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.487761988145488e-08,
+      "loss": 0.5868,
+      "step": 42910
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.481816169835894e-08,
+      "loss": 0.6287,
+      "step": 42911
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.4758754151269505e-08,
+      "loss": 0.6118,
+      "step": 42912
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.469939724048965e-08,
+      "loss": 0.601,
+      "step": 42913
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.464009096632137e-08,
+      "loss": 0.6131,
+      "step": 42914
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.458083532906442e-08,
+      "loss": 0.6046,
+      "step": 42915
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.452163032902078e-08,
+      "loss": 0.5817,
+      "step": 42916
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.4462475966491324e-08,
+      "loss": 0.5937,
+      "step": 42917
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.4403372241776926e-08,
+      "loss": 0.6919,
+      "step": 42918
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.434431915517733e-08,
+      "loss": 0.573,
+      "step": 42919
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.428531670699231e-08,
+      "loss": 0.6106,
+      "step": 42920
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.422636489752384e-08,
+      "loss": 0.5276,
+      "step": 42921
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.416746372706836e-08,
+      "loss": 0.5912,
+      "step": 42922
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.410861319592673e-08,
+      "loss": 0.5874,
+      "step": 42923
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.40498133043976e-08,
+      "loss": 0.5975,
+      "step": 42924
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3991064052780744e-08,
+      "loss": 0.5765,
+      "step": 42925
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.393236544137479e-08,
+      "loss": 0.5847,
+      "step": 42926
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.387371747047619e-08,
+      "loss": 0.5533,
+      "step": 42927
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3815120140383575e-08,
+      "loss": 0.586,
+      "step": 42928
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3756573451395606e-08,
+      "loss": 0.5562,
+      "step": 42929
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.369807740380871e-08,
+      "loss": 0.6305,
+      "step": 42930
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.363963199792153e-08,
+      "loss": 0.6025,
+      "step": 42931
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3581237234029394e-08,
+      "loss": 0.6094,
+      "step": 42932
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3522893112429844e-08,
+      "loss": 0.5717,
+      "step": 42933
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.34645996334193e-08,
+      "loss": 0.5666,
+      "step": 42934
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.34063567972942e-08,
+      "loss": 0.5821,
+      "step": 42935
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.334816460434986e-08,
+      "loss": 0.6425,
+      "step": 42936
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.329002305488271e-08,
+      "loss": 0.5862,
+      "step": 42937
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.323193214918807e-08,
+      "loss": 0.6018,
+      "step": 42938
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.317389188756126e-08,
+      "loss": 0.6883,
+      "step": 42939
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3115902270296484e-08,
+      "loss": 0.6083,
+      "step": 42940
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.305796329768796e-08,
+      "loss": 0.6615,
+      "step": 42941
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.3000074970030993e-08,
+      "loss": 0.6039,
+      "step": 42942
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.2942237287620914e-08,
+      "loss": 0.6088,
+      "step": 42943
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.28844502507486e-08,
+      "loss": 0.59,
+      "step": 42944
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.282671385971048e-08,
+      "loss": 0.63,
+      "step": 42945
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.276902811479854e-08,
+      "loss": 0.6367,
+      "step": 42946
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.271139301630699e-08,
+      "loss": 0.6332,
+      "step": 42947
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.2653808564526714e-08,
+      "loss": 0.6326,
+      "step": 42948
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.259627475975191e-08,
+      "loss": 0.6227,
+      "step": 42949
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.253879160227458e-08,
+      "loss": 0.575,
+      "step": 42950
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.24813590923867e-08,
+      "loss": 0.5645,
+      "step": 42951
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.2423977230380265e-08,
+      "loss": 0.5884,
+      "step": 42952
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.2366646016546154e-08,
+      "loss": 0.6139,
+      "step": 42953
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.230936545117747e-08,
+      "loss": 0.578,
+      "step": 42954
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.2252135534563965e-08,
+      "loss": 0.598,
+      "step": 42955
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.219495626699542e-08,
+      "loss": 0.6051,
+      "step": 42956
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.213782764876383e-08,
+      "loss": 0.5875,
+      "step": 42957
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.208074968015895e-08,
+      "loss": 0.5791,
+      "step": 42958
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.202372236147167e-08,
+      "loss": 0.622,
+      "step": 42959
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.196674569299174e-08,
+      "loss": 0.6069,
+      "step": 42960
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.190981967500673e-08,
+      "loss": 0.5793,
+      "step": 42961
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.18529443078075e-08,
+      "loss": 0.6199,
+      "step": 42962
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1796119591683825e-08,
+      "loss": 0.6161,
+      "step": 42963
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.173934552692215e-08,
+      "loss": 0.6348,
+      "step": 42964
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.168262211381334e-08,
+      "loss": 0.6604,
+      "step": 42965
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.162594935264274e-08,
+      "loss": 0.6208,
+      "step": 42966
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.156932724370121e-08,
+      "loss": 0.6155,
+      "step": 42967
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1512755787275193e-08,
+      "loss": 0.5272,
+      "step": 42968
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1456234983652244e-08,
+      "loss": 0.5904,
+      "step": 42969
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.13997648331199e-08,
+      "loss": 0.543,
+      "step": 42970
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.134334533596461e-08,
+      "loss": 0.5626,
+      "step": 42971
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.128697649247281e-08,
+      "loss": 0.6184,
+      "step": 42972
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1230658302932036e-08,
+      "loss": 0.6496,
+      "step": 42973
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.117439076762763e-08,
+      "loss": 0.6069,
+      "step": 42974
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1118173886844905e-08,
+      "loss": 0.6111,
+      "step": 42975
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.106200766087142e-08,
+      "loss": 0.605,
+      "step": 42976
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1005892089990274e-08,
+      "loss": 0.6207,
+      "step": 42977
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.094982717448902e-08,
+      "loss": 0.5784,
+      "step": 42978
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.089381291464966e-08,
+      "loss": 0.5172,
+      "step": 42979
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.083784931075973e-08,
+      "loss": 0.5449,
+      "step": 42980
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.078193636310123e-08,
+      "loss": 0.5789,
+      "step": 42981
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.07260740719606e-08,
+      "loss": 0.5515,
+      "step": 42982
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.067026243761984e-08,
+      "loss": 0.582,
+      "step": 42983
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.061450146036316e-08,
+      "loss": 0.5947,
+      "step": 42984
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.055879114047366e-08,
+      "loss": 0.6498,
+      "step": 42985
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.050313147823447e-08,
+      "loss": 0.5691,
+      "step": 42986
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.044752247392868e-08,
+      "loss": 0.5975,
+      "step": 42987
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.03919641278394e-08,
+      "loss": 0.5696,
+      "step": 42988
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.033645644024752e-08,
+      "loss": 0.6085,
+      "step": 42989
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.028099941143614e-08,
+      "loss": 0.6134,
+      "step": 42990
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0225593041687264e-08,
+      "loss": 0.5758,
+      "step": 42991
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.017023733128066e-08,
+      "loss": 0.6311,
+      "step": 42992
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.011493228050055e-08,
+      "loss": 0.6025,
+      "step": 42993
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.0059677889624496e-08,
+      "loss": 0.5436,
+      "step": 42994
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.00044741589367e-08,
+      "loss": 0.5673,
+      "step": 42995
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9949321088714736e-08,
+      "loss": 0.6403,
+      "step": 42996
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9894218679240584e-08,
+      "loss": 0.6922,
+      "step": 42997
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.983916693079403e-08,
+      "loss": 0.5976,
+      "step": 42998
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9784165843654844e-08,
+      "loss": 0.5831,
+      "step": 42999
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9729215418100588e-08,
+      "loss": 0.5886,
+      "step": 43000
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9674315654414366e-08,
+      "loss": 0.6039,
+      "step": 43001
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9619466552871512e-08,
+      "loss": 0.5972,
+      "step": 43002
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9564668113752915e-08,
+      "loss": 0.5661,
+      "step": 43003
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.950992033733502e-08,
+      "loss": 0.6187,
+      "step": 43004
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.94552232238976e-08,
+      "loss": 0.5832,
+      "step": 43005
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9400576773719324e-08,
+      "loss": 0.5515,
+      "step": 43006
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.934598098707553e-08,
+      "loss": 0.5517,
+      "step": 43007
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9291435864245987e-08,
+      "loss": 0.6649,
+      "step": 43008
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9236941405504927e-08,
+      "loss": 0.5916,
+      "step": 43009
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.918249761113323e-08,
+      "loss": 0.5783,
+      "step": 43010
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9128104481402908e-08,
+      "loss": 0.5486,
+      "step": 43011
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.907376201659484e-08,
+      "loss": 0.5762,
+      "step": 43012
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.9019470216982148e-08,
+      "loss": 0.59,
+      "step": 43013
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8965229082842382e-08,
+      "loss": 0.5834,
+      "step": 43014
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8911038614449772e-08,
+      "loss": 0.6077,
+      "step": 43015
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8856898812080757e-08,
+      "loss": 0.6346,
+      "step": 43016
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.880280967600957e-08,
+      "loss": 0.5632,
+      "step": 43017
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8748771206511537e-08,
+      "loss": 0.5903,
+      "step": 43018
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8694783403862004e-08,
+      "loss": 0.6607,
+      "step": 43019
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8640846268332966e-08,
+      "loss": 0.6185,
+      "step": 43020
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8586959800200876e-08,
+      "loss": 0.6089,
+      "step": 43021
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8533123999738844e-08,
+      "loss": 0.5759,
+      "step": 43022
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8479338867221096e-08,
+      "loss": 0.6731,
+      "step": 43023
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8425604402918527e-08,
+      "loss": 0.5918,
+      "step": 43024
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8371920607106473e-08,
+      "loss": 0.632,
+      "step": 43025
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.831828748005805e-08,
+      "loss": 0.6009,
+      "step": 43026
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8264705022043037e-08,
+      "loss": 0.6374,
+      "step": 43027
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8211173233336774e-08,
+      "loss": 0.6159,
+      "step": 43028
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8157692114209046e-08,
+      "loss": 0.5999,
+      "step": 43029
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8104261664932962e-08,
+      "loss": 0.6035,
+      "step": 43030
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8050881885779423e-08,
+      "loss": 0.5704,
+      "step": 43031
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.799755277701932e-08,
+      "loss": 0.547,
+      "step": 43032
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.794427433892466e-08,
+      "loss": 0.6409,
+      "step": 43033
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7891046571765222e-08,
+      "loss": 0.631,
+      "step": 43034
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7837869475811906e-08,
+      "loss": 0.5809,
+      "step": 43035
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7784743051334496e-08,
+      "loss": 0.5912,
+      "step": 43036
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7731667298603883e-08,
+      "loss": 0.5409,
+      "step": 43037
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.767864221788874e-08,
+      "loss": 0.6536,
+      "step": 43038
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7625667809459967e-08,
+      "loss": 0.5857,
+      "step": 43039
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7572744073585123e-08,
+      "loss": 0.6529,
+      "step": 43040
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7519871010532884e-08,
+      "loss": 0.6407,
+      "step": 43041
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7467048620573033e-08,
+      "loss": 0.6402,
+      "step": 43042
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7414276903974246e-08,
+      "loss": 0.6102,
+      "step": 43043
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7361555861004085e-08,
+      "loss": 0.6024,
+      "step": 43044
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7308885491930113e-08,
+      "loss": 0.6693,
+      "step": 43045
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.725626579702101e-08,
+      "loss": 0.5737,
+      "step": 43046
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7203696776543222e-08,
+      "loss": 0.6133,
+      "step": 43047
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7151178430764313e-08,
+      "loss": 0.5999,
+      "step": 43048
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.709871075995074e-08,
+      "loss": 0.6185,
+      "step": 43049
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.7046293764368957e-08,
+      "loss": 0.6279,
+      "step": 43050
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6993927444285417e-08,
+      "loss": 0.6338,
+      "step": 43051
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6941611799967683e-08,
+      "loss": 0.6257,
+      "step": 43052
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6889346831678875e-08,
+      "loss": 0.5417,
+      "step": 43053
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.683713253968767e-08,
+      "loss": 0.5594,
+      "step": 43054
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6784968924256082e-08,
+      "loss": 0.6112,
+      "step": 43055
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6732855985651674e-08,
+      "loss": 0.5583,
+      "step": 43056
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6680793724138677e-08,
+      "loss": 0.5933,
+      "step": 43057
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6628782139981323e-08,
+      "loss": 0.621,
+      "step": 43058
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.657682123344385e-08,
+      "loss": 0.6246,
+      "step": 43059
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6524911004790487e-08,
+      "loss": 0.6147,
+      "step": 43060
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.647305145428436e-08,
+      "loss": 0.5703,
+      "step": 43061
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6421242582190808e-08,
+      "loss": 0.5941,
+      "step": 43062
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6369484388770737e-08,
+      "loss": 0.5448,
+      "step": 43063
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6317776874288382e-08,
+      "loss": 0.6096,
+      "step": 43064
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.626612003900686e-08,
+      "loss": 0.5788,
+      "step": 43065
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6214513883187077e-08,
+      "loss": 0.5812,
+      "step": 43066
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6162958407093263e-08,
+      "loss": 0.5901,
+      "step": 43067
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6111453610986325e-08,
+      "loss": 0.5551,
+      "step": 43068
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6059999495127163e-08,
+      "loss": 0.6157,
+      "step": 43069
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.6008596059778902e-08,
+      "loss": 0.5844,
+      "step": 43070
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.595724330520244e-08,
+      "loss": 0.6164,
+      "step": 43071
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5905941231657573e-08,
+      "loss": 0.5148,
+      "step": 43072
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.585468983940631e-08,
+      "loss": 0.5998,
+      "step": 43073
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5803489128707338e-08,
+      "loss": 0.5713,
+      "step": 43074
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5752339099822664e-08,
+      "loss": 0.6315,
+      "step": 43075
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5701239753010976e-08,
+      "loss": 0.6725,
+      "step": 43076
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.565019108853317e-08,
+      "loss": 0.5667,
+      "step": 43077
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5599193106646823e-08,
+      "loss": 0.5698,
+      "step": 43078
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5548245807612838e-08,
+      "loss": 0.5723,
+      "step": 43079
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.549734919168989e-08,
+      "loss": 0.5714,
+      "step": 43080
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.544650325913556e-08,
+      "loss": 0.5979,
+      "step": 43081
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5395708010208518e-08,
+      "loss": 0.6119,
+      "step": 43082
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.534496344516635e-08,
+      "loss": 0.6102,
+      "step": 43083
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5294269564268836e-08,
+      "loss": 0.6306,
+      "step": 43084
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5243626367772445e-08,
+      "loss": 0.616,
+      "step": 43085
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5193033855933636e-08,
+      "loss": 0.6114,
+      "step": 43086
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.514249202901109e-08,
+      "loss": 0.6378,
+      "step": 43087
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.509200088726016e-08,
+      "loss": 0.583,
+      "step": 43088
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.5041560430938417e-08,
+      "loss": 0.5733,
+      "step": 43089
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4991170660301212e-08,
+      "loss": 0.602,
+      "step": 43090
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4940831575606117e-08,
+      "loss": 0.5606,
+      "step": 43091
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4890543177107374e-08,
+      "loss": 0.6122,
+      "step": 43092
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4840305465060333e-08,
+      "loss": 0.5258,
+      "step": 43093
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4790118439722567e-08,
+      "loss": 0.6029,
+      "step": 43094
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4739982101347204e-08,
+      "loss": 0.5795,
+      "step": 43095
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4689896450188487e-08,
+      "loss": 0.5829,
+      "step": 43096
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4639861486502882e-08,
+      "loss": 0.5705,
+      "step": 43097
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.458987721054351e-08,
+      "loss": 0.625,
+      "step": 43098
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.453994362256351e-08,
+      "loss": 0.5808,
+      "step": 43099
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.449006072281823e-08,
+      "loss": 0.5656,
+      "step": 43100
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4440228511559694e-08,
+      "loss": 0.656,
+      "step": 43101
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4390446989043247e-08,
+      "loss": 0.5663,
+      "step": 43102
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4340716155519807e-08,
+      "loss": 0.5908,
+      "step": 43103
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4291036011241388e-08,
+      "loss": 0.6455,
+      "step": 43104
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4241406556463343e-08,
+      "loss": 0.5965,
+      "step": 43105
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4191827791435475e-08,
+      "loss": 0.6507,
+      "step": 43106
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.414229971641091e-08,
+      "loss": 0.6227,
+      "step": 43107
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4092822331640566e-08,
+      "loss": 0.5847,
+      "step": 43108
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.4043395637376453e-08,
+      "loss": 0.6338,
+      "step": 43109
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3994019633869492e-08,
+      "loss": 0.5929,
+      "step": 43110
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.394469432137059e-08,
+      "loss": 0.6107,
+      "step": 43111
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.389541970012954e-08,
+      "loss": 0.6488,
+      "step": 43112
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3846195770398372e-08,
+      "loss": 0.6466,
+      "step": 43113
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3797022532425775e-08,
+      "loss": 0.5834,
+      "step": 43114
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3747899986462653e-08,
+      "loss": 0.5692,
+      "step": 43115
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.36988281327577e-08,
+      "loss": 0.609,
+      "step": 43116
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3649806971560718e-08,
+      "loss": 0.6507,
+      "step": 43117
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3600836503120393e-08,
+      "loss": 0.5991,
+      "step": 43118
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.355191672768653e-08,
+      "loss": 0.611,
+      "step": 43119
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3503047645507815e-08,
+      "loss": 0.6106,
+      "step": 43120
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.345422925683072e-08,
+      "loss": 0.6008,
+      "step": 43121
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3405461561905042e-08,
+      "loss": 0.6533,
+      "step": 43122
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.335674456097725e-08,
+      "loss": 0.6123,
+      "step": 43123
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3308078254297152e-08,
+      "loss": 0.6249,
+      "step": 43124
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3259462642108988e-08,
+      "loss": 0.615,
+      "step": 43125
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.321089772466145e-08,
+      "loss": 0.6053,
+      "step": 43126
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.3162383502202124e-08,
+      "loss": 0.6323,
+      "step": 43127
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.3113919974976362e-08,
+      "loss": 0.6649,
+      "step": 43128
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.3065507143229527e-08,
+      "loss": 0.6267,
+      "step": 43129
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.3017145007210305e-08,
+      "loss": 0.6502,
+      "step": 43130
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.296883356716184e-08,
+      "loss": 0.5867,
+      "step": 43131
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.29205728233306e-08,
+      "loss": 0.5735,
+      "step": 43132
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2872362775961944e-08,
+      "loss": 0.5661,
+      "step": 43133
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2824203425301227e-08,
+      "loss": 0.629,
+      "step": 43134
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2776094771592704e-08,
+      "loss": 0.6039,
+      "step": 43135
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2728036815080623e-08,
+      "loss": 0.6623,
+      "step": 43136
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.268002955600812e-08,
+      "loss": 0.5644,
+      "step": 43137
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2632072994622777e-08,
+      "loss": 0.5645,
+      "step": 43138
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2584167131164404e-08,
+      "loss": 0.6348,
+      "step": 43139
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2536311965877246e-08,
+      "loss": 0.6145,
+      "step": 43140
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2488507499006663e-08,
+      "loss": 0.6468,
+      "step": 43141
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2440753730793573e-08,
+      "loss": 0.5798,
+      "step": 43142
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.239305066148001e-08,
+      "loss": 0.5604,
+      "step": 43143
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2345398291310215e-08,
+      "loss": 0.5859,
+      "step": 43144
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2297796620526226e-08,
+      "loss": 0.5938,
+      "step": 43145
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.225024564936895e-08,
+      "loss": 0.5958,
+      "step": 43146
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2202745378079315e-08,
+      "loss": 0.5835,
+      "step": 43147
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.2155295806900457e-08,
+      "loss": 0.6145,
+      "step": 43148
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.210789693607329e-08,
+      "loss": 0.5478,
+      "step": 43149
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.206054876583652e-08,
+      "loss": 0.5342,
+      "step": 43150
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.201325129643328e-08,
+      "loss": 0.6402,
+      "step": 43151
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.196600452810227e-08,
+      "loss": 0.6182,
+      "step": 43152
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1918808461084407e-08,
+      "loss": 0.6688,
+      "step": 43153
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.18716630956195e-08,
+      "loss": 0.5606,
+      "step": 43154
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1824568431947356e-08,
+      "loss": 0.5793,
+      "step": 43155
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1777524470306678e-08,
+      "loss": 0.5181,
+      "step": 43156
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1730531210936155e-08,
+      "loss": 0.6267,
+      "step": 43157
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1683588654074496e-08,
+      "loss": 0.599,
+      "step": 43158
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1636696799962608e-08,
+      "loss": 0.6259,
+      "step": 43159
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1589855648835867e-08,
+      "loss": 0.5373,
+      "step": 43160
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1543065200932968e-08,
+      "loss": 0.5176,
+      "step": 43161
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1496325456492605e-08,
+      "loss": 0.5543,
+      "step": 43162
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1449636415752374e-08,
+      "loss": 0.5994,
+      "step": 43163
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1402998078947636e-08,
+      "loss": 0.6125,
+      "step": 43164
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.135641044631709e-08,
+      "loss": 0.6039,
+      "step": 43165
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1309873518097214e-08,
+      "loss": 0.5776,
+      "step": 43166
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1263387294524484e-08,
+      "loss": 0.6175,
+      "step": 43167
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.121695177583427e-08,
+      "loss": 0.5867,
+      "step": 43168
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.117056696226305e-08,
+      "loss": 0.5911,
+      "step": 43169
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.11242328540473e-08,
+      "loss": 0.6572,
+      "step": 43170
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1077949451422387e-08,
+      "loss": 0.7076,
+      "step": 43171
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1031716754621456e-08,
+      "loss": 0.606,
+      "step": 43172
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0985534763880987e-08,
+      "loss": 0.5882,
+      "step": 43173
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0939403479436347e-08,
+      "loss": 0.5931,
+      "step": 43174
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0893322901520685e-08,
+      "loss": 0.6367,
+      "step": 43175
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0847293030368254e-08,
+      "loss": 0.6093,
+      "step": 43176
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0801313866214425e-08,
+      "loss": 0.5443,
+      "step": 43177
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.075538540929012e-08,
+      "loss": 0.7059,
+      "step": 43178
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0709507659830708e-08,
+      "loss": 0.5795,
+      "step": 43179
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0663680618069337e-08,
+      "loss": 0.526,
+      "step": 43180
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0617904284238044e-08,
+      "loss": 0.6148,
+      "step": 43181
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0572178658569974e-08,
+      "loss": 0.5708,
+      "step": 43182
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0526503741297166e-08,
+      "loss": 0.569,
+      "step": 43183
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0480879532651654e-08,
+      "loss": 0.5761,
+      "step": 43184
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0435306032865476e-08,
+      "loss": 0.5385,
+      "step": 43185
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.038978324217067e-08,
+      "loss": 0.5956,
+      "step": 43186
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0344311160798157e-08,
+      "loss": 0.6048,
+      "step": 43187
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.029888978897776e-08,
+      "loss": 0.5716,
+      "step": 43188
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0253519126942624e-08,
+      "loss": 0.6048,
+      "step": 43189
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.020819917492145e-08,
+      "loss": 0.5722,
+      "step": 43190
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.016292993314517e-08,
+      "loss": 0.6339,
+      "step": 43191
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.01177114018436e-08,
+      "loss": 0.6561,
+      "step": 43192
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0072543581246552e-08,
+      "loss": 0.6488,
+      "step": 43193
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.0027426471583844e-08,
+      "loss": 0.6169,
+      "step": 43194
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9982360073084185e-08,
+      "loss": 0.5914,
+      "step": 43195
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9937344385977385e-08,
+      "loss": 0.5567,
+      "step": 43196
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9892379410492156e-08,
+      "loss": 0.6425,
+      "step": 43197
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9847465146856092e-08,
+      "loss": 0.571,
+      "step": 43198
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.98026015952979e-08,
+      "loss": 0.5558,
+      "step": 43199
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9757788756046282e-08,
+      "loss": 0.6178,
+      "step": 43200
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9713026629326614e-08,
+      "loss": 0.616,
+      "step": 43201
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9668315215368717e-08,
+      "loss": 0.6075,
+      "step": 43202
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9623654514399072e-08,
+      "loss": 0.5791,
+      "step": 43203
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9579044526645273e-08,
+      "loss": 0.6352,
+      "step": 43204
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9534485252331592e-08,
+      "loss": 0.5496,
+      "step": 43205
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.948997669168784e-08,
+      "loss": 0.6375,
+      "step": 43206
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9445518844937172e-08,
+      "loss": 0.5498,
+      "step": 43207
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9401111712307187e-08,
+      "loss": 0.595,
+      "step": 43208
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9356755294023255e-08,
+      "loss": 0.5825,
+      "step": 43209
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9312449590310756e-08,
+      "loss": 0.5589,
+      "step": 43210
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.926819460139395e-08,
+      "loss": 0.6305,
+      "step": 43211
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.922399032749933e-08,
+      "loss": 0.5908,
+      "step": 43212
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9179836768848935e-08,
+      "loss": 0.6059,
+      "step": 43213
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9135733925671474e-08,
+      "loss": 0.5991,
+      "step": 43214
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9091681798186768e-08,
+      "loss": 0.5965,
+      "step": 43215
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9047680386620192e-08,
+      "loss": 0.5962,
+      "step": 43216
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.9003729691194905e-08,
+      "loss": 0.6064,
+      "step": 43217
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.895982971213628e-08,
+      "loss": 0.6632,
+      "step": 43218
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.891598044966525e-08,
+      "loss": 0.5961,
+      "step": 43219
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8872181904004972e-08,
+      "loss": 0.5911,
+      "step": 43220
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8828434075378598e-08,
+      "loss": 0.5824,
+      "step": 43221
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8784736964007067e-08,
+      "loss": 0.6037,
+      "step": 43222
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.874109057011464e-08,
+      "loss": 0.5826,
+      "step": 43223
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.869749489392114e-08,
+      "loss": 0.6,
+      "step": 43224
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8653949935648617e-08,
+      "loss": 0.5623,
+      "step": 43225
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8610455695519114e-08,
+      "loss": 0.6056,
+      "step": 43226
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8567012173753563e-08,
+      "loss": 0.6181,
+      "step": 43227
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.852361937057068e-08,
+      "loss": 0.5878,
+      "step": 43228
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8480277286193616e-08,
+      "loss": 0.5919,
+      "step": 43229
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8436985920842197e-08,
+      "loss": 0.5787,
+      "step": 43230
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8393745274734032e-08,
+      "loss": 0.5688,
+      "step": 43231
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8350555348091158e-08,
+      "loss": 0.5475,
+      "step": 43232
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8307416141132296e-08,
+      "loss": 0.6504,
+      "step": 43233
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8264327654077263e-08,
+      "loss": 0.6023,
+      "step": 43234
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8221289887143667e-08,
+      "loss": 0.5376,
+      "step": 43235
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8178302840552443e-08,
+      "loss": 0.5622,
+      "step": 43236
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.813536651451897e-08,
+      "loss": 0.6029,
+      "step": 43237
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8092480909264188e-08,
+      "loss": 0.5689,
+      "step": 43238
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8049646025003475e-08,
+      "loss": 0.5492,
+      "step": 43239
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.8006861861956658e-08,
+      "loss": 0.5797,
+      "step": 43240
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7964128420341343e-08,
+      "loss": 0.6176,
+      "step": 43241
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7921445700371798e-08,
+      "loss": 0.6165,
+      "step": 43242
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.787881370226785e-08,
+      "loss": 0.6076,
+      "step": 43243
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7836232426244882e-08,
+      "loss": 0.6377,
+      "step": 43244
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7793701872519385e-08,
+      "loss": 0.5704,
+      "step": 43245
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7751222041307857e-08,
+      "loss": 0.6268,
+      "step": 43246
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7708792932825682e-08,
+      "loss": 0.5722,
+      "step": 43247
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7666414547288234e-08,
+      "loss": 0.5583,
+      "step": 43248
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7624086884910908e-08,
+      "loss": 0.6286,
+      "step": 43249
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7581809945910187e-08,
+      "loss": 0.5964,
+      "step": 43250
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7539583730498134e-08,
+      "loss": 0.51,
+      "step": 43251
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7497408238892344e-08,
+      "loss": 0.5947,
+      "step": 43252
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7455283471304872e-08,
+      "loss": 0.6105,
+      "step": 43253
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7413209427951107e-08,
+      "loss": 0.6613,
+      "step": 43254
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7371186109044203e-08,
+      "loss": 0.5758,
+      "step": 43255
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7329213514798436e-08,
+      "loss": 0.5649,
+      "step": 43256
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7287291645425862e-08,
+      "loss": 0.5469,
+      "step": 43257
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.724542050114075e-08,
+      "loss": 0.5415,
+      "step": 43258
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7203600082154048e-08,
+      "loss": 0.503,
+      "step": 43259
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7161830388681132e-08,
+      "loss": 0.5784,
+      "step": 43260
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.712011142093073e-08,
+      "loss": 0.6321,
+      "step": 43261
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.707844317911822e-08,
+      "loss": 0.5744,
+      "step": 43262
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.7036825663453438e-08,
+      "loss": 0.6117,
+      "step": 43263
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6995258874147326e-08,
+      "loss": 0.6293,
+      "step": 43264
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6953742811413043e-08,
+      "loss": 0.6175,
+      "step": 43265
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6912277475459315e-08,
+      "loss": 0.5401,
+      "step": 43266
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6870862866498193e-08,
+      "loss": 0.5673,
+      "step": 43267
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6829498984740624e-08,
+      "loss": 0.6447,
+      "step": 43268
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.678818583039643e-08,
+      "loss": 0.5451,
+      "step": 43269
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6746923403674343e-08,
+      "loss": 0.6323,
+      "step": 43270
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6705711704784188e-08,
+      "loss": 0.6338,
+      "step": 43271
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6664550733936912e-08,
+      "loss": 0.6417,
+      "step": 43272
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.6623440491341235e-08,
+      "loss": 0.5719,
+      "step": 43273
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.658238097720477e-08,
+      "loss": 0.6028,
+      "step": 43274
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6541372191737348e-08,
+      "loss": 0.6483,
+      "step": 43275
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6500414135147692e-08,
+      "loss": 0.594,
+      "step": 43276
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.64595068076423e-08,
+      "loss": 0.568,
+      "step": 43277
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6418650209431008e-08,
+      "loss": 0.5916,
+      "step": 43278
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6377844340719208e-08,
+      "loss": 0.6075,
+      "step": 43279
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6337089201715617e-08,
+      "loss": 0.5848,
+      "step": 43280
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.629638479262785e-08,
+      "loss": 0.6115,
+      "step": 43281
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.62557311136613e-08,
+      "loss": 0.6339,
+      "step": 43282
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6215128165023575e-08,
+      "loss": 0.6312,
+      "step": 43283
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6174575946921178e-08,
+      "loss": 0.6024,
+      "step": 43284
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6134074459558392e-08,
+      "loss": 0.588,
+      "step": 43285
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6093623703142826e-08,
+      "loss": 0.5865,
+      "step": 43286
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.605322367787987e-08,
+      "loss": 0.6236,
+      "step": 43287
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6012874383973808e-08,
+      "loss": 0.6485,
+      "step": 43288
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.597257582163114e-08,
+      "loss": 0.5694,
+      "step": 43289
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.593232799105504e-08,
+      "loss": 0.5615,
+      "step": 43290
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5892130892450896e-08,
+      "loss": 0.5723,
+      "step": 43291
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.585198452602299e-08,
+      "loss": 0.6073,
+      "step": 43292
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5811888891975602e-08,
+      "loss": 0.6575,
+      "step": 43293
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5771843990510793e-08,
+      "loss": 0.5842,
+      "step": 43294
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5731849821833955e-08,
+      "loss": 0.6005,
+      "step": 43295
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5691906386148258e-08,
+      "loss": 0.6063,
+      "step": 43296
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5652013683654656e-08,
+      "loss": 0.6378,
+      "step": 43297
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5612171714558532e-08,
+      "loss": 0.5282,
+      "step": 43298
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5572380479060845e-08,
+      "loss": 0.644,
+      "step": 43299
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.553263997736254e-08,
+      "loss": 0.5613,
+      "step": 43300
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5492950209669013e-08,
+      "loss": 0.6651,
+      "step": 43301
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5453311176178988e-08,
+      "loss": 0.566,
+      "step": 43302
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.541372287709453e-08,
+      "loss": 0.6274,
+      "step": 43303
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5374185312616586e-08,
+      "loss": 0.6061,
+      "step": 43304
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5334698482947218e-08,
+      "loss": 0.6161,
+      "step": 43305
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.529526238828738e-08,
+      "loss": 0.6296,
+      "step": 43306
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5255877028834687e-08,
+      "loss": 0.5824,
+      "step": 43307
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.521654240479231e-08,
+      "loss": 0.5905,
+      "step": 43308
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5177258516357873e-08,
+      "loss": 0.5415,
+      "step": 43309
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5138025363732323e-08,
+      "loss": 0.5982,
+      "step": 43310
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.50988429471155e-08,
+      "loss": 0.651,
+      "step": 43311
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5059711266705024e-08,
+      "loss": 0.586,
+      "step": 43312
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.5020630322699625e-08,
+      "loss": 0.5722,
+      "step": 43313
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4981600115300255e-08,
+      "loss": 0.5957,
+      "step": 43314
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.494262064470231e-08,
+      "loss": 0.5861,
+      "step": 43315
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4903691911105633e-08,
+      "loss": 0.6373,
+      "step": 43316
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4864813914707843e-08,
+      "loss": 0.6353,
+      "step": 43317
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4825986655705448e-08,
+      "loss": 0.6079,
+      "step": 43318
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.478721013429718e-08,
+      "loss": 0.6122,
+      "step": 43319
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4748484350679548e-08,
+      "loss": 0.5634,
+      "step": 43320
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.470980930504795e-08,
+      "loss": 0.5625,
+      "step": 43321
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4671184997601117e-08,
+      "loss": 0.6227,
+      "step": 43322
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4632611428534448e-08,
+      "loss": 0.5534,
+      "step": 43323
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4594088598043344e-08,
+      "loss": 0.5924,
+      "step": 43324
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.455561650632431e-08,
+      "loss": 0.6515,
+      "step": 43325
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4517195153572749e-08,
+      "loss": 0.5792,
+      "step": 43326
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.447882453998406e-08,
+      "loss": 0.6263,
+      "step": 43327
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.444050466575253e-08,
+      "loss": 0.6494,
+      "step": 43328
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.440223553107245e-08,
+      "loss": 0.6181,
+      "step": 43329
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4364017136139219e-08,
+      "loss": 0.6106,
+      "step": 43330
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4325849481147125e-08,
+      "loss": 0.5838,
+      "step": 43331
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4287732566290457e-08,
+      "loss": 0.578,
+      "step": 43332
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4249666391761286e-08,
+      "loss": 0.6388,
+      "step": 43333
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4211650957755007e-08,
+      "loss": 0.5899,
+      "step": 43334
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4173686264462583e-08,
+      "loss": 0.5579,
+      "step": 43335
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.413577231207941e-08,
+      "loss": 0.5864,
+      "step": 43336
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4097909100796448e-08,
+      "loss": 0.6255,
+      "step": 43337
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4060096630805764e-08,
+      "loss": 0.6007,
+      "step": 43338
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.4022334902300539e-08,
+      "loss": 0.5658,
+      "step": 43339
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3984623915472839e-08,
+      "loss": 0.6582,
+      "step": 43340
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3946963670513624e-08,
+      "loss": 0.6029,
+      "step": 43341
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3909354167614964e-08,
+      "loss": 0.5744,
+      "step": 43342
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3871795406966703e-08,
+      "loss": 0.6129,
+      "step": 43343
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3834287388760914e-08,
+      "loss": 0.5853,
+      "step": 43344
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3796830113187443e-08,
+      "loss": 0.6067,
+      "step": 43345
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3759423580436138e-08,
+      "loss": 0.612,
+      "step": 43346
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3722067790699068e-08,
+      "loss": 0.6067,
+      "step": 43347
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.368476274416386e-08,
+      "loss": 0.6331,
+      "step": 43348
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3647508441021473e-08,
+      "loss": 0.6374,
+      "step": 43349
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3610304881460646e-08,
+      "loss": 0.6334,
+      "step": 43350
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3573152065670114e-08,
+      "loss": 0.5971,
+      "step": 43351
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3536049993839729e-08,
+      "loss": 0.6341,
+      "step": 43352
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3498998666157115e-08,
+      "loss": 0.5681,
+      "step": 43353
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3461998082811013e-08,
+      "loss": 0.5698,
+      "step": 43354
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3425048243989048e-08,
+      "loss": 0.5867,
+      "step": 43355
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3388149149878849e-08,
+      "loss": 0.5821,
+      "step": 43356
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3351300800668044e-08,
+      "loss": 0.5895,
+      "step": 43357
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3314503196544259e-08,
+      "loss": 0.5982,
+      "step": 43358
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3277756337694015e-08,
+      "loss": 0.6609,
+      "step": 43359
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3241060224304936e-08,
+      "loss": 0.6468,
+      "step": 43360
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.320441485656132e-08,
+      "loss": 0.5836,
+      "step": 43361
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3167820234651908e-08,
+      "loss": 0.5897,
+      "step": 43362
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3131276358762102e-08,
+      "loss": 0.5321,
+      "step": 43363
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3094783229076202e-08,
+      "loss": 0.6414,
+      "step": 43364
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.3058340845780726e-08,
+      "loss": 0.589,
+      "step": 43365
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.302194920906108e-08,
+      "loss": 0.6061,
+      "step": 43366
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2985608319100451e-08,
+      "loss": 0.6233,
+      "step": 43367
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2949318176086468e-08,
+      "loss": 0.6263,
+      "step": 43368
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2913078780200095e-08,
+      "loss": 0.6558,
+      "step": 43369
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2876890131628961e-08,
+      "loss": 0.5835,
+      "step": 43370
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2840752230554033e-08,
+      "loss": 0.5959,
+      "step": 43371
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2804665077160716e-08,
+      "loss": 0.5797,
+      "step": 43372
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2768628671631089e-08,
+      "loss": 0.5632,
+      "step": 43373
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2732643014149448e-08,
+      "loss": 0.5913,
+      "step": 43374
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.269670810489787e-08,
+      "loss": 0.6023,
+      "step": 43375
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2660823944059542e-08,
+      "loss": 0.5791,
+      "step": 43376
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2624990531816539e-08,
+      "loss": 0.5694,
+      "step": 43377
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.258920786835094e-08,
+      "loss": 0.5796,
+      "step": 43378
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2553475953844819e-08,
+      "loss": 0.5959,
+      "step": 43379
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2517794788479143e-08,
+      "loss": 0.5499,
+      "step": 43380
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.248216437243488e-08,
+      "loss": 0.5956,
+      "step": 43381
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2446584705894104e-08,
+      "loss": 0.6263,
+      "step": 43382
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2411055789037785e-08,
+      "loss": 0.5746,
+      "step": 43383
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2375577622044665e-08,
+      "loss": 0.5847,
+      "step": 43384
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2340150205097934e-08,
+      "loss": 0.5825,
+      "step": 43385
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2304773538374115e-08,
+      "loss": 0.5676,
+      "step": 43386
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2269447622056396e-08,
+      "loss": 0.5112,
+      "step": 43387
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2234172456321303e-08,
+      "loss": 0.5888,
+      "step": 43388
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2198948041350912e-08,
+      "loss": 0.5938,
+      "step": 43389
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2163774377321746e-08,
+      "loss": 0.5997,
+      "step": 43390
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2128651464413665e-08,
+      "loss": 0.5972,
+      "step": 43391
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2093579302804303e-08,
+      "loss": 0.5996,
+      "step": 43392
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2058557892673518e-08,
+      "loss": 0.5765,
+      "step": 43393
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.2023587234197832e-08,
+      "loss": 0.5976,
+      "step": 43394
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1988667327555992e-08,
+      "loss": 0.6099,
+      "step": 43395
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1953798172924525e-08,
+      "loss": 0.6022,
+      "step": 43396
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1918979770481065e-08,
+      "loss": 0.6306,
+      "step": 43397
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1884212120402139e-08,
+      "loss": 0.6199,
+      "step": 43398
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1849495222864271e-08,
+      "loss": 0.6166,
+      "step": 43399
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1814829078045098e-08,
+      "loss": 0.6316,
+      "step": 43400
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1780213686120034e-08,
+      "loss": 0.5978,
+      "step": 43401
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1745649047264495e-08,
+      "loss": 0.5539,
+      "step": 43402
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1711135161653897e-08,
+      "loss": 0.6085,
+      "step": 43403
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1676672029464765e-08,
+      "loss": 0.5635,
+      "step": 43404
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1642259650872512e-08,
+      "loss": 0.6429,
+      "step": 43405
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1607898026050336e-08,
+      "loss": 0.6191,
+      "step": 43406
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1573587155173649e-08,
+      "loss": 0.6003,
+      "step": 43407
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1539327038416759e-08,
+      "loss": 0.5974,
+      "step": 43408
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.150511767595397e-08,
+      "loss": 0.6312,
+      "step": 43409
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1470959067958477e-08,
+      "loss": 0.5903,
+      "step": 43410
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1436851214604582e-08,
+      "loss": 0.5618,
+      "step": 43411
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1402794116066596e-08,
+      "loss": 0.619,
+      "step": 43412
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1368787772514377e-08,
+      "loss": 0.5825,
+      "step": 43413
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1334832184124457e-08,
+      "loss": 0.5834,
+      "step": 43414
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1300927351065583e-08,
+      "loss": 0.645,
+      "step": 43415
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1267073273513173e-08,
+      "loss": 0.5906,
+      "step": 43416
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1233269951638204e-08,
+      "loss": 0.6321,
+      "step": 43417
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1199517385612758e-08,
+      "loss": 0.6229,
+      "step": 43418
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1165815575606698e-08,
+      "loss": 0.5769,
+      "step": 43419
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1132164521793221e-08,
+      "loss": 0.5754,
+      "step": 43420
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.1098564224342189e-08,
+      "loss": 0.6564,
+      "step": 43421
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.106501468342569e-08,
+      "loss": 0.6509,
+      "step": 43422
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.1031515899212475e-08,
+      "loss": 0.5952,
+      "step": 43423
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0998067871873519e-08,
+      "loss": 0.6075,
+      "step": 43424
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0964670601578686e-08,
+      "loss": 0.5641,
+      "step": 43425
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0931324088497841e-08,
+      "loss": 0.5521,
+      "step": 43426
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0898028332800848e-08,
+      "loss": 0.5398,
+      "step": 43427
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.086478333465535e-08,
+      "loss": 0.5849,
+      "step": 43428
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.083158909423232e-08,
+      "loss": 0.5839,
+      "step": 43429
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0798445611699404e-08,
+      "loss": 0.5959,
+      "step": 43430
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0765352887225356e-08,
+      "loss": 0.5839,
+      "step": 43431
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0732310920976708e-08,
+      "loss": 0.5487,
+      "step": 43432
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0699319713124434e-08,
+      "loss": 0.6229,
+      "step": 43433
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0666379263832849e-08,
+      "loss": 0.549,
+      "step": 43434
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0633489573271816e-08,
+      "loss": 0.641,
+      "step": 43435
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0600650641606758e-08,
+      "loss": 0.6665,
+      "step": 43436
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0567862469006429e-08,
+      "loss": 0.6202,
+      "step": 43437
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0535125055636252e-08,
+      "loss": 0.5737,
+      "step": 43438
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0502438401661651e-08,
+      "loss": 0.6011,
+      "step": 43439
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0469802507250272e-08,
+      "loss": 0.5879,
+      "step": 43440
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0437217372567532e-08,
+      "loss": 0.6624,
+      "step": 43441
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0404682997778859e-08,
+      "loss": 0.5868,
+      "step": 43442
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0372199383048564e-08,
+      "loss": 0.5431,
+      "step": 43443
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0339766528544293e-08,
+      "loss": 0.5906,
+      "step": 43444
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0307384434428137e-08,
+      "loss": 0.6126,
+      "step": 43445
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0275053100866627e-08,
+      "loss": 0.5971,
+      "step": 43446
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0242772528022971e-08,
+      "loss": 0.596,
+      "step": 43447
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0210542716060368e-08,
+      "loss": 0.5553,
+      "step": 43448
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0178363665145352e-08,
+      "loss": 0.6293,
+      "step": 43449
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0146235375438906e-08,
+      "loss": 0.5954,
+      "step": 43450
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0114157847104233e-08,
+      "loss": 0.6351,
+      "step": 43451
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0082131080306756e-08,
+      "loss": 0.5729,
+      "step": 43452
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0050155075206347e-08,
+      "loss": 0.5951,
+      "step": 43453
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.0018229831967319e-08,
+      "loss": 0.6413,
+      "step": 43454
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.986355350751764e-09,
+      "loss": 0.6219,
+      "step": 43455
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.954531631720666e-09,
+      "loss": 0.6417,
+      "step": 43456
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.922758675037226e-09,
+      "loss": 0.608,
+      "step": 43457
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.891036480861316e-09,
+      "loss": 0.626,
+      "step": 43458
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.85936504935392e-09,
+      "loss": 0.5919,
+      "step": 43459
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.827744380678239e-09,
+      "loss": 0.6207,
+      "step": 43460
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.796174474993036e-09,
+      "loss": 0.654,
+      "step": 43461
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.764655332459295e-09,
+      "loss": 0.5896,
+      "step": 43462
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.733186953236885e-09,
+      "loss": 0.5938,
+      "step": 43463
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.701769337485679e-09,
+      "loss": 0.6156,
+      "step": 43464
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.67040248536666e-09,
+      "loss": 0.6665,
+      "step": 43465
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.639086397037478e-09,
+      "loss": 0.5949,
+      "step": 43466
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.607821072658008e-09,
+      "loss": 0.5978,
+      "step": 43467
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.5766065123859e-09,
+      "loss": 0.625,
+      "step": 43468
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.545442716382136e-09,
+      "loss": 0.604,
+      "step": 43469
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.514329684803259e-09,
+      "loss": 0.6227,
+      "step": 43470
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.48326741780914e-09,
+      "loss": 0.5979,
+      "step": 43471
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.452255915555209e-09,
+      "loss": 0.6006,
+      "step": 43472
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.42129517820023e-09,
+      "loss": 0.578,
+      "step": 43473
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.390385205901853e-09,
+      "loss": 0.5491,
+      "step": 43474
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.35952599881662e-09,
+      "loss": 0.6008,
+      "step": 43475
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.328717557102185e-09,
+      "loss": 0.5606,
+      "step": 43476
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.297959880913976e-09,
+      "loss": 0.5663,
+      "step": 43477
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.267252970408536e-09,
+      "loss": 0.5172,
+      "step": 43478
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.236596825742406e-09,
+      "loss": 0.6403,
+      "step": 43479
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.205991447071016e-09,
+      "loss": 0.6066,
+      "step": 43480
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.1754368345498e-09,
+      "loss": 0.6498,
+      "step": 43481
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.144932988334188e-09,
+      "loss": 0.5828,
+      "step": 43482
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.1144799085785e-09,
+      "loss": 0.668,
+      "step": 43483
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.084077595439277e-09,
+      "loss": 0.6438,
+      "step": 43484
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.053726049068623e-09,
+      "loss": 0.558,
+      "step": 43485
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.023425269623076e-09,
+      "loss": 0.5954,
+      "step": 43486
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.993175257254738e-09,
+      "loss": 0.5661,
+      "step": 43487
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.962976012119041e-09,
+      "loss": 0.6424,
+      "step": 43488
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.932827534366973e-09,
+      "loss": 0.6054,
+      "step": 43489
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.902729824155077e-09,
+      "loss": 0.6198,
+      "step": 43490
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.872682881633233e-09,
+      "loss": 0.6162,
+      "step": 43491
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.842686706955761e-09,
+      "loss": 0.6054,
+      "step": 43492
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.812741300273652e-09,
+      "loss": 0.5681,
+      "step": 43493
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.782846661741228e-09,
+      "loss": 0.6086,
+      "step": 43494
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.753002791508369e-09,
+      "loss": 0.6407,
+      "step": 43495
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.723209689728285e-09,
+      "loss": 0.5809,
+      "step": 43496
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.693467356551965e-09,
+      "loss": 0.6168,
+      "step": 43497
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.663775792129292e-09,
+      "loss": 0.5931,
+      "step": 43498
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.634134996612364e-09,
+      "loss": 0.5531,
+      "step": 43499
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.604544970151062e-09,
+      "loss": 0.5708,
+      "step": 43500
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.575005712897489e-09,
+      "loss": 0.5847,
+      "step": 43501
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.5455172249993e-09,
+      "loss": 0.6528,
+      "step": 43502
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.516079506607488e-09,
+      "loss": 0.5412,
+      "step": 43503
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.486692557871934e-09,
+      "loss": 0.611,
+      "step": 43504
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.457356378941406e-09,
+      "loss": 0.6161,
+      "step": 43505
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.428070969965786e-09,
+      "loss": 0.6205,
+      "step": 43506
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.39883633109273e-09,
+      "loss": 0.5744,
+      "step": 43507
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.369652462472122e-09,
+      "loss": 0.5882,
+      "step": 43508
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.34051936425051e-09,
+      "loss": 0.6178,
+      "step": 43509
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.311437036577774e-09,
+      "loss": 0.6779,
+      "step": 43510
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.282405479600463e-09,
+      "loss": 0.6398,
+      "step": 43511
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.253424693466239e-09,
+      "loss": 0.6029,
+      "step": 43512
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.224494678322758e-09,
+      "loss": 0.6201,
+      "step": 43513
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.195615434317683e-09,
+      "loss": 0.5465,
+      "step": 43514
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.16678696159534e-09,
+      "loss": 0.5819,
+      "step": 43515
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.138009260304502e-09,
+      "loss": 0.548,
+      "step": 43516
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.109282330590607e-09,
+      "loss": 0.5415,
+      "step": 43517
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.080606172600203e-09,
+      "loss": 0.5899,
+      "step": 43518
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.05198078647762e-09,
+      "loss": 0.6565,
+      "step": 43519
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 8.023406172370519e-09,
+      "loss": 0.6527,
+      "step": 43520
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.994882330422115e-09,
+      "loss": 0.5792,
+      "step": 43521
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.966409260777851e-09,
+      "loss": 0.5786,
+      "step": 43522
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.937986963582056e-09,
+      "loss": 0.6236,
+      "step": 43523
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.909615438980168e-09,
+      "loss": 0.6436,
+      "step": 43524
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.881294687116514e-09,
+      "loss": 0.5786,
+      "step": 43525
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.853024708133206e-09,
+      "loss": 0.5347,
+      "step": 43526
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.82480550217568e-09,
+      "loss": 0.6124,
+      "step": 43527
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.796637069386048e-09,
+      "loss": 0.6233,
+      "step": 43528
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.768519409908637e-09,
+      "loss": 0.6091,
+      "step": 43529
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.740452523885555e-09,
+      "loss": 0.5941,
+      "step": 43530
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.712436411460023e-09,
+      "loss": 0.5481,
+      "step": 43531
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.684471072773037e-09,
+      "loss": 0.5796,
+      "step": 43532
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.656556507968927e-09,
+      "loss": 0.5719,
+      "step": 43533
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.62869271718758e-09,
+      "loss": 0.6808,
+      "step": 43534
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.600879700571106e-09,
+      "loss": 0.6488,
+      "step": 43535
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.573117458260503e-09,
+      "loss": 0.6019,
+      "step": 43536
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.54540599039788e-09,
+      "loss": 0.608,
+      "step": 43537
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.517745297124235e-09,
+      "loss": 0.6435,
+      "step": 43538
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.490135378578345e-09,
+      "loss": 0.6133,
+      "step": 43539
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.46257623490121e-09,
+      "loss": 0.5776,
+      "step": 43540
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.435067866233825e-09,
+      "loss": 0.584,
+      "step": 43541
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.407610272714971e-09,
+      "loss": 0.6572,
+      "step": 43542
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.380203454485646e-09,
+      "loss": 0.6343,
+      "step": 43543
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.3528474116824065e-09,
+      "loss": 0.6136,
+      "step": 43544
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.325542144447362e-09,
+      "loss": 0.6465,
+      "step": 43545
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.298287652915958e-09,
+      "loss": 0.5963,
+      "step": 43546
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.2710839372303055e-09,
+      "loss": 0.5175,
+      "step": 43547
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.2439309975247395e-09,
+      "loss": 0.5965,
+      "step": 43548
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.21682883394137e-09,
+      "loss": 0.5689,
+      "step": 43549
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.189777446614532e-09,
+      "loss": 0.6315,
+      "step": 43550
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.162776835683005e-09,
+      "loss": 0.6436,
+      "step": 43551
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.1358270012844564e-09,
+      "loss": 0.5125,
+      "step": 43552
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.108927943555444e-09,
+      "loss": 0.5816,
+      "step": 43553
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.0820796626314135e-09,
+      "loss": 0.6589,
+      "step": 43554
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.0552821586511445e-09,
+      "loss": 0.5731,
+      "step": 43555
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.028535431748973e-09,
+      "loss": 0.5534,
+      "step": 43556
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.0018394820614566e-09,
+      "loss": 0.585,
+      "step": 43557
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.975194309724043e-09,
+      "loss": 0.5707,
+      "step": 43558
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.9485999148721786e-09,
+      "loss": 0.6238,
+      "step": 43559
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.922056297641311e-09,
+      "loss": 0.5668,
+      "step": 43560
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.895563458166887e-09,
+      "loss": 0.6028,
+      "step": 43561
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.869121396581024e-09,
+      "loss": 0.6079,
+      "step": 43562
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.84273011302139e-09,
+      "loss": 0.5821,
+      "step": 43563
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.816389607618989e-09,
+      "loss": 0.6263,
+      "step": 43564
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.790099880510381e-09,
+      "loss": 0.6227,
+      "step": 43565
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.763860931827681e-09,
+      "loss": 0.6421,
+      "step": 43566
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.737672761704117e-09,
+      "loss": 0.5624,
+      "step": 43567
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 6.711535370272915e-09,
+      "loss": 0.6029,
+      "step": 43568
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.685448757667301e-09,
+      "loss": 0.5828,
+      "step": 43569
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.659412924020503e-09,
+      "loss": 0.6335,
+      "step": 43570
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.633427869462416e-09,
+      "loss": 0.6103,
+      "step": 43571
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.6074935941273785e-09,
+      "loss": 0.6174,
+      "step": 43572
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.581610098146396e-09,
+      "loss": 0.6569,
+      "step": 43573
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.555777381650474e-09,
+      "loss": 0.5917,
+      "step": 43574
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.52999544477173e-09,
+      "loss": 0.5964,
+      "step": 43575
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.504264287641171e-09,
+      "loss": 0.6259,
+      "step": 43576
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.478583910388691e-09,
+      "loss": 0.6198,
+      "step": 43577
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.452954313145299e-09,
+      "loss": 0.5991,
+      "step": 43578
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.427375496040889e-09,
+      "loss": 0.5493,
+      "step": 43579
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.4018474592064675e-09,
+      "loss": 0.6212,
+      "step": 43580
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.376370202770821e-09,
+      "loss": 0.5948,
+      "step": 43581
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.3509437268627346e-09,
+      "loss": 0.5914,
+      "step": 43582
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.325568031613216e-09,
+      "loss": 0.5653,
+      "step": 43583
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.30024311714994e-09,
+      "loss": 0.6207,
+      "step": 43584
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.274968983602803e-09,
+      "loss": 0.6279,
+      "step": 43585
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.249745631098369e-09,
+      "loss": 0.601,
+      "step": 43586
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.224573059765426e-09,
+      "loss": 0.5644,
+      "step": 43587
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.199451269732759e-09,
+      "loss": 0.6003,
+      "step": 43588
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.1743802611280434e-09,
+      "loss": 0.5994,
+      "step": 43589
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.149360034076734e-09,
+      "loss": 0.6148,
+      "step": 43590
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.124390588708729e-09,
+      "loss": 0.6738,
+      "step": 43591
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.099471925149481e-09,
+      "loss": 0.6302,
+      "step": 43592
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.074604043524446e-09,
+      "loss": 0.6317,
+      "step": 43593
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.04978694396241e-09,
+      "loss": 0.6256,
+      "step": 43594
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.025020626587718e-09,
+      "loss": 0.615,
+      "step": 43595
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 6.000305091526937e-09,
+      "loss": 0.5905,
+      "step": 43596
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.97564033890552e-09,
+      "loss": 0.6248,
+      "step": 43597
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.951026368848922e-09,
+      "loss": 0.6201,
+      "step": 43598
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.92646318148149e-09,
+      "loss": 0.6323,
+      "step": 43599
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.901950776928678e-09,
+      "loss": 0.5792,
+      "step": 43600
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.8774891553148305e-09,
+      "loss": 0.621,
+      "step": 43601
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.853078316765404e-09,
+      "loss": 0.5932,
+      "step": 43602
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.828718261402522e-09,
+      "loss": 0.6116,
+      "step": 43603
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.804408989351639e-09,
+      "loss": 0.596,
+      "step": 43604
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.7801505007348825e-09,
+      "loss": 0.5347,
+      "step": 43605
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.755942795675484e-09,
+      "loss": 0.5324,
+      "step": 43606
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.73178587429779e-09,
+      "loss": 0.6293,
+      "step": 43607
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.707679736723926e-09,
+      "loss": 0.5536,
+      "step": 43608
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.683624383076014e-09,
+      "loss": 0.5908,
+      "step": 43609
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.65961981347729e-09,
+      "loss": 0.6601,
+      "step": 43610
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.63566602804877e-09,
+      "loss": 0.5966,
+      "step": 43611
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.611763026911466e-09,
+      "loss": 0.5716,
+      "step": 43612
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.5879108101886126e-09,
+      "loss": 0.636,
+      "step": 43613
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.5641093780012254e-09,
+      "loss": 0.5566,
+      "step": 43614
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.540358730468098e-09,
+      "loss": 0.6029,
+      "step": 43615
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.516658867712466e-09,
+      "loss": 0.6295,
+      "step": 43616
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.493009789853121e-09,
+      "loss": 0.5999,
+      "step": 43617
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.469411497011079e-09,
+      "loss": 0.6433,
+      "step": 43618
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.445863989306244e-09,
+      "loss": 0.6103,
+      "step": 43619
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.422367266857409e-09,
+      "loss": 0.6564,
+      "step": 43620
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.398921329785589e-09,
+      "loss": 0.583,
+      "step": 43621
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.375526178207358e-09,
+      "loss": 0.5592,
+      "step": 43622
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.352181812244839e-09,
+      "loss": 0.5856,
+      "step": 43623
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.328888232013496e-09,
+      "loss": 0.6087,
+      "step": 43624
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.305645437634344e-09,
+      "loss": 0.6139,
+      "step": 43625
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.282453429223955e-09,
+      "loss": 0.6285,
+      "step": 43626
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.2593122069000135e-09,
+      "loss": 0.5798,
+      "step": 43627
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.236221770780203e-09,
+      "loss": 0.6443,
+      "step": 43628
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.2131821209833175e-09,
+      "loss": 0.5773,
+      "step": 43629
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.19019325762482e-09,
+      "loss": 0.6235,
+      "step": 43630
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.167255180822394e-09,
+      "loss": 0.6312,
+      "step": 43631
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.1443678906926145e-09,
+      "loss": 0.6305,
+      "step": 43632
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.121531387350942e-09,
+      "loss": 0.5934,
+      "step": 43633
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.098745670913951e-09,
+      "loss": 0.5925,
+      "step": 43634
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.0760107414971065e-09,
+      "loss": 0.5312,
+      "step": 43635
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.053326599215869e-09,
+      "loss": 0.664,
+      "step": 43636
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.030693244186813e-09,
+      "loss": 0.6202,
+      "step": 43637
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 5.008110676523181e-09,
+      "loss": 0.6066,
+      "step": 43638
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.985578896341548e-09,
+      "loss": 0.6485,
+      "step": 43639
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.9630979037540435e-09,
+      "loss": 0.6231,
+      "step": 43640
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.940667698877244e-09,
+      "loss": 0.5497,
+      "step": 43641
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.91828828182328e-09,
+      "loss": 0.5927,
+      "step": 43642
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.895959652707616e-09,
+      "loss": 0.5735,
+      "step": 43643
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.873681811642383e-09,
+      "loss": 0.6739,
+      "step": 43644
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.851454758740826e-09,
+      "loss": 0.6146,
+      "step": 43645
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.829278494116185e-09,
+      "loss": 0.627,
+      "step": 43646
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.807153017881705e-09,
+      "loss": 0.5623,
+      "step": 43647
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.785078330149517e-09,
+      "loss": 0.5946,
+      "step": 43648
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.763054431030644e-09,
+      "loss": 0.6323,
+      "step": 43649
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.74108132063833e-09,
+      "loss": 0.586,
+      "step": 43650
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.719158999083595e-09,
+      "loss": 0.5464,
+      "step": 43651
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.697287466478573e-09,
+      "loss": 0.6054,
+      "step": 43652
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.675466722934285e-09,
+      "loss": 0.6123,
+      "step": 43653
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.653696768560645e-09,
+      "loss": 0.633,
+      "step": 43654
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.6319776034686736e-09,
+      "loss": 0.5539,
+      "step": 43655
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.610309227769394e-09,
+      "loss": 0.6034,
+      "step": 43656
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.588691641571608e-09,
+      "loss": 0.5892,
+      "step": 43657
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.567124844986337e-09,
+      "loss": 0.5963,
+      "step": 43658
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.545608838123494e-09,
+      "loss": 0.615,
+      "step": 43659
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.5241436210907705e-09,
+      "loss": 0.6422,
+      "step": 43660
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.502729193999189e-09,
+      "loss": 0.6502,
+      "step": 43661
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.4813655569553304e-09,
+      "loss": 0.586,
+      "step": 43662
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.460052710070218e-09,
+      "loss": 0.5252,
+      "step": 43663
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.438790653450431e-09,
+      "loss": 0.5996,
+      "step": 43664
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.417579387203663e-09,
+      "loss": 0.6083,
+      "step": 43665
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.396418911439826e-09,
+      "loss": 0.5529,
+      "step": 43666
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.37530922626439e-09,
+      "loss": 0.6731,
+      "step": 43667
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.354250331785048e-09,
+      "loss": 0.5453,
+      "step": 43668
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.333242228110602e-09,
+      "loss": 0.5869,
+      "step": 43669
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.3122849153443e-09,
+      "loss": 0.6119,
+      "step": 43670
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.291378393596057e-09,
+      "loss": 0.5987,
+      "step": 43671
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.270522662970233e-09,
+      "loss": 0.5775,
+      "step": 43672
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.24971772357341e-09,
+      "loss": 0.5967,
+      "step": 43673
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.228963575511058e-09,
+      "loss": 0.6529,
+      "step": 43674
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.208260218888649e-09,
+      "loss": 0.6228,
+      "step": 43675
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.187607653811654e-09,
+      "loss": 0.5936,
+      "step": 43676
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.167005880384434e-09,
+      "loss": 0.6034,
+      "step": 43677
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.14645489871246e-09,
+      "loss": 0.5758,
+      "step": 43678
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.1259547089000925e-09,
+      "loss": 0.6648,
+      "step": 43679
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.105505311050584e-09,
+      "loss": 0.5973,
+      "step": 43680
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.0851067052682934e-09,
+      "loss": 0.5546,
+      "step": 43681
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.064758891657583e-09,
+      "loss": 0.5899,
+      "step": 43682
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.044461870320593e-09,
+      "loss": 0.6101,
+      "step": 43683
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.024215641361684e-09,
+      "loss": 0.5632,
+      "step": 43684
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.0040202048841074e-09,
+      "loss": 0.6241,
+      "step": 43685
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.983875560987782e-09,
+      "loss": 0.5864,
+      "step": 43686
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.9637817097781805e-09,
+      "loss": 0.5715,
+      "step": 43687
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.943738651355222e-09,
+      "loss": 0.5824,
+      "step": 43688
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.923746385822158e-09,
+      "loss": 0.5335,
+      "step": 43689
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.903804913280018e-09,
+      "loss": 0.5966,
+      "step": 43690
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.883914233829833e-09,
+      "loss": 0.581,
+      "step": 43691
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.864074347572633e-09,
+      "loss": 0.5533,
+      "step": 43692
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.844285254610558e-09,
+      "loss": 0.6305,
+      "step": 43693
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.8245469550424184e-09,
+      "loss": 0.6623,
+      "step": 43694
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.804859448969245e-09,
+      "loss": 0.6116,
+      "step": 43695
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.7852227364909564e-09,
+      "loss": 0.5605,
+      "step": 43696
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.765636817708584e-09,
+      "loss": 0.5294,
+      "step": 43697
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.746101692719828e-09,
+      "loss": 0.6003,
+      "step": 43698
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.726617361624607e-09,
+      "loss": 0.5731,
+      "step": 43699
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.7071838245217316e-09,
+      "loss": 0.6286,
+      "step": 43700
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.687801081511122e-09,
+      "loss": 0.5849,
+      "step": 43701
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.668469132689367e-09,
+      "loss": 0.5545,
+      "step": 43702
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.6491879781563877e-09,
+      "loss": 0.6229,
+      "step": 43703
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.6299576180087727e-09,
+      "loss": 0.6033,
+      "step": 43704
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.6107780523453317e-09,
+      "loss": 0.5626,
+      "step": 43705
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5916492812626546e-09,
+      "loss": 0.6124,
+      "step": 43706
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.572571304859551e-09,
+      "loss": 0.5863,
+      "step": 43707
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5535441232303904e-09,
+      "loss": 0.5802,
+      "step": 43708
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5345677364739817e-09,
+      "loss": 0.6235,
+      "step": 43709
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5156421446858046e-09,
+      "loss": 0.6172,
+      "step": 43710
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.4967673479624487e-09,
+      "loss": 0.5847,
+      "step": 43711
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.4779433463993927e-09,
+      "loss": 0.6268,
+      "step": 43712
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.4591701400932265e-09,
+      "loss": 0.5746,
+      "step": 43713
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.4404477291383186e-09,
+      "loss": 0.5997,
+      "step": 43714
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.421776113629038e-09,
+      "loss": 0.6305,
+      "step": 43715
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.4031552936630854e-09,
+      "loss": 0.5729,
+      "step": 43716
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.3845852693326075e-09,
+      "loss": 0.5927,
+      "step": 43717
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.3660660407319745e-09,
+      "loss": 0.62,
+      "step": 43718
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.347597607956665e-09,
+      "loss": 0.5756,
+      "step": 43719
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.3291799710999384e-09,
+      "loss": 0.569,
+      "step": 43720
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.3108131302550526e-09,
+      "loss": 0.5462,
+      "step": 43721
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.2924970855163775e-09,
+      "loss": 0.5284,
+      "step": 43722
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.2742318369749506e-09,
+      "loss": 0.6108,
+      "step": 43723
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.256017384725141e-09,
+      "loss": 0.6275,
+      "step": 43724
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.2378537288602074e-09,
+      "loss": 0.6168,
+      "step": 43725
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.219740869470078e-09,
+      "loss": 0.5637,
+      "step": 43726
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.2016788066491223e-09,
+      "loss": 0.5745,
+      "step": 43727
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.1836675404883776e-09,
+      "loss": 0.6308,
+      "step": 43728
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.1657070710777727e-09,
+      "loss": 0.5637,
+      "step": 43729
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.147797398510566e-09,
+      "loss": 0.5912,
+      "step": 43730
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.1299385228777958e-09,
+      "loss": 0.6251,
+      "step": 43731
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.1121304442682797e-09,
+      "loss": 0.5945,
+      "step": 43732
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.094373162775277e-09,
+      "loss": 0.6019,
+      "step": 43733
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.0766666784864952e-09,
+      "loss": 0.6544,
+      "step": 43734
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.0590109914929723e-09,
+      "loss": 0.5745,
+      "step": 43735
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.0414061018846365e-09,
+      "loss": 0.592,
+      "step": 43736
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.0238520097503053e-09,
+      "loss": 0.5847,
+      "step": 43737
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.0063487151787974e-09,
+      "loss": 0.6247,
+      "step": 43738
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.98889621826115e-09,
+      "loss": 0.5986,
+      "step": 43739
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.9714945190839615e-09,
+      "loss": 0.5785,
+      "step": 43740
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.9541436177371596e-09,
+      "loss": 0.6543,
+      "step": 43741
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.9368435143073413e-09,
+      "loss": 0.5558,
+      "step": 43742
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.919594208883325e-09,
+      "loss": 0.6011,
+      "step": 43743
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.9023957015528183e-09,
+      "loss": 0.597,
+      "step": 43744
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.885247992403528e-09,
+      "loss": 0.6611,
+      "step": 43745
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8681510815209425e-09,
+      "loss": 0.5679,
+      "step": 43746
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8511049689938785e-09,
+      "loss": 0.5537,
+      "step": 43747
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.834109654907824e-09,
+      "loss": 0.6034,
+      "step": 43748
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.817165139349376e-09,
+      "loss": 0.617,
+      "step": 43749
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.800271422405132e-09,
+      "loss": 0.5729,
+      "step": 43750
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.7834285041605792e-09,
+      "loss": 0.5824,
+      "step": 43751
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.7666363847000943e-09,
+      "loss": 0.6894,
+      "step": 43752
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.749895064111385e-09,
+      "loss": 0.6165,
+      "step": 43753
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.7332045424777186e-09,
+      "loss": 0.5912,
+      "step": 43754
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.7165648198845818e-09,
+      "loss": 0.6181,
+      "step": 43755
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6999758964174617e-09,
+      "loss": 0.5519,
+      "step": 43756
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6834377721585148e-09,
+      "loss": 0.6233,
+      "step": 43757
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6669504471932284e-09,
+      "loss": 0.6233,
+      "step": 43758
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.650513921603759e-09,
+      "loss": 0.6401,
+      "step": 43759
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.634128195476704e-09,
+      "loss": 0.5856,
+      "step": 43760
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6177932688919992e-09,
+      "loss": 0.5904,
+      "step": 43761
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.6015091419351323e-09,
+      "loss": 0.5661,
+      "step": 43762
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.5852758146882597e-09,
+      "loss": 0.5864,
+      "step": 43763
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.5690932872324272e-09,
+      "loss": 0.596,
+      "step": 43764
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.552961559652012e-09,
+      "loss": 0.5789,
+      "step": 43765
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.53688063202695e-09,
+      "loss": 0.6194,
+      "step": 43766
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.5208505044393983e-09,
+      "loss": 0.607,
+      "step": 43767
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.504871176972623e-09,
+      "loss": 0.6426,
+      "step": 43768
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.4889426497054504e-09,
+      "loss": 0.6045,
+      "step": 43769
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.473064922721147e-09,
+      "loss": 0.5937,
+      "step": 43770
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.4572379960985384e-09,
+      "loss": 0.6014,
+      "step": 43771
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.441461869917561e-09,
+      "loss": 0.5703,
+      "step": 43772
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.425736544260371e-09,
+      "loss": 0.6643,
+      "step": 43773
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.4100620192057943e-09,
+      "loss": 0.5836,
+      "step": 43774
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.3944382948337675e-09,
+      "loss": 0.63,
+      "step": 43775
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.378865371223116e-09,
+      "loss": 0.5552,
+      "step": 43776
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.3633432484548858e-09,
+      "loss": 0.5368,
+      "step": 43777
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.3478719266045723e-09,
+      "loss": 0.5566,
+      "step": 43778
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.332451405753222e-09,
+      "loss": 0.6084,
+      "step": 43779
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.3170816859785505e-09,
+      "loss": 0.5503,
+      "step": 43780
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.3017627673593833e-09,
+      "loss": 0.6747,
+      "step": 43781
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.286494649972326e-09,
+      "loss": 0.6905,
+      "step": 43782
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.2712773338962047e-09,
+      "loss": 0.5921,
+      "step": 43783
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.256110819207624e-09,
+      "loss": 0.5887,
+      "step": 43784
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.2409951059831902e-09,
+      "loss": 0.5809,
+      "step": 43785
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.2259301943006185e-09,
+      "loss": 0.5666,
+      "step": 43786
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.2109160842354038e-09,
+      "loss": 0.6108,
+      "step": 43787
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.195952775865262e-09,
+      "loss": 0.5675,
+      "step": 43788
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1810402692645784e-09,
+      "loss": 0.5923,
+      "step": 43789
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1661785645110678e-09,
+      "loss": 0.556,
+      "step": 43790
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1513676616791155e-09,
+      "loss": 0.6075,
+      "step": 43791
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1366075608431067e-09,
+      "loss": 0.613,
+      "step": 43792
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.121898262079647e-09,
+      "loss": 0.6327,
+      "step": 43793
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1072397654631206e-09,
+      "loss": 0.6155,
+      "step": 43794
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.0926320710668024e-09,
+      "loss": 0.5734,
+      "step": 43795
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.078075178967298e-09,
+      "loss": 0.5763,
+      "step": 43796
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.063569089235662e-09,
+      "loss": 0.6029,
+      "step": 43797
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.049113801948499e-09,
+      "loss": 0.654,
+      "step": 43798
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.0347093171768638e-09,
+      "loss": 0.606,
+      "step": 43799
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.0203556349951415e-09,
+      "loss": 0.6239,
+      "step": 43800
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.006052755475496e-09,
+      "loss": 0.6205,
+      "step": 43801
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9918006786912024e-09,
+      "loss": 0.5993,
+      "step": 43802
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9775994047155355e-09,
+      "loss": 0.6202,
+      "step": 43803
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9634489336195495e-09,
+      "loss": 0.6038,
+      "step": 43804
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9493492654742984e-09,
+      "loss": 0.5574,
+      "step": 43805
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9353004003530575e-09,
+      "loss": 0.5968,
+      "step": 43806
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9213023383268804e-09,
+      "loss": 0.5957,
+      "step": 43807
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.9073550794668216e-09,
+      "loss": 0.6141,
+      "step": 43808
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.8934586238428254e-09,
+      "loss": 0.6133,
+      "step": 43809
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.879612971527056e-09,
+      "loss": 0.6107,
+      "step": 43810
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.8658181225883478e-09,
+      "loss": 0.6317,
+      "step": 43811
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.852074077097754e-09,
+      "loss": 0.6269,
+      "step": 43812
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.8383808351252198e-09,
+      "loss": 0.6007,
+      "step": 43813
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.8247383967395782e-09,
+      "loss": 0.699,
+      "step": 43814
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.8111467620107736e-09,
+      "loss": 0.5509,
+      "step": 43815
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.79760593100764e-09,
+      "loss": 0.6376,
+      "step": 43816
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7841159037990108e-09,
+      "loss": 0.6435,
+      "step": 43817
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7706766804537202e-09,
+      "loss": 0.5957,
+      "step": 43818
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7572882610383813e-09,
+      "loss": 0.5825,
+      "step": 43819
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7439506456240485e-09,
+      "loss": 0.5497,
+      "step": 43820
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.730663834276225e-09,
+      "loss": 0.6282,
+      "step": 43821
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7174278270626343e-09,
+      "loss": 0.6218,
+      "step": 43822
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.7042426240510002e-09,
+      "loss": 0.6196,
+      "step": 43823
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6911082253079358e-09,
+      "loss": 0.6131,
+      "step": 43824
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.678024630901165e-09,
+      "loss": 0.6186,
+      "step": 43825
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6649918408961906e-09,
+      "loss": 0.5701,
+      "step": 43826
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6520098553596265e-09,
+      "loss": 0.6059,
+      "step": 43827
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6390786743580856e-09,
+      "loss": 0.6689,
+      "step": 43828
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.626198297955961e-09,
+      "loss": 0.6005,
+      "step": 43829
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.613368726218756e-09,
+      "loss": 0.5475,
+      "step": 43830
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6005899592119733e-09,
+      "loss": 0.6539,
+      "step": 43831
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5878619970022269e-09,
+      "loss": 0.6729,
+      "step": 43832
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.575184839651689e-09,
+      "loss": 0.6119,
+      "step": 43833
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5625584872258626e-09,
+      "loss": 0.5357,
+      "step": 43834
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5499829397891408e-09,
+      "loss": 0.5774,
+      "step": 43835
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5374581974048065e-09,
+      "loss": 0.6349,
+      "step": 43836
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5249842601372522e-09,
+      "loss": 0.5947,
+      "step": 43837
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.5125611280486508e-09,
+      "loss": 0.589,
+      "step": 43838
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.500188801203395e-09,
+      "loss": 0.6168,
+      "step": 43839
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4878672796647675e-09,
+      "loss": 0.5594,
+      "step": 43840
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4755965634927204e-09,
+      "loss": 0.6121,
+      "step": 43841
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4633766527527571e-09,
+      "loss": 0.5555,
+      "step": 43842
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4512075475048294e-09,
+      "loss": 0.583,
+      "step": 43843
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4390892478122198e-09,
+      "loss": 0.5952,
+      "step": 43844
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4270217537348808e-09,
+      "loss": 0.627,
+      "step": 43845
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4150050653360947e-09,
+      "loss": 0.6019,
+      "step": 43846
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.403039182675814e-09,
+      "loss": 0.6035,
+      "step": 43847
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3911241058139902e-09,
+      "loss": 0.5859,
+      "step": 43848
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.379259834812796e-09,
+      "loss": 0.6373,
+      "step": 43849
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3674463697310737e-09,
+      "loss": 0.5654,
+      "step": 43850
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.355683710630995e-09,
+      "loss": 0.5993,
+      "step": 43851
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3439718575691818e-09,
+      "loss": 0.5936,
+      "step": 43852
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3323108106078063e-09,
+      "loss": 0.5897,
+      "step": 43853
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3207005698057106e-09,
+      "loss": 0.625,
+      "step": 43854
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.3091411352206263e-09,
+      "loss": 0.6015,
+      "step": 43855
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2976325069125052e-09,
+      "loss": 0.6143,
+      "step": 43856
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2861746849390788e-09,
+      "loss": 0.5792,
+      "step": 43857
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2747676693591893e-09,
+      "loss": 0.6016,
+      "step": 43858
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2634114602294578e-09,
+      "loss": 0.6202,
+      "step": 43859
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2521060576098366e-09,
+      "loss": 0.6096,
+      "step": 43860
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2408514615558366e-09,
+      "loss": 0.621,
+      "step": 43861
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.2296476721251893e-09,
+      "loss": 0.5609,
+      "step": 43862
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.2184946893756267e-09,
+      "loss": 0.5849,
+      "step": 43863
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.2073925133626597e-09,
+      "loss": 0.5888,
+      "step": 43864
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1963411441440197e-09,
+      "loss": 0.6152,
+      "step": 43865
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.185340581774108e-09,
+      "loss": 0.6301,
+      "step": 43866
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1743908263106563e-09,
+      "loss": 0.6319,
+      "step": 43867
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.163491877808065e-09,
+      "loss": 0.6353,
+      "step": 43868
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1526437363218456e-09,
+      "loss": 0.6183,
+      "step": 43869
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1418464019075093e-09,
+      "loss": 0.5627,
+      "step": 43870
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1310998746194567e-09,
+      "loss": 0.6322,
+      "step": 43871
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1204041545120891e-09,
+      "loss": 0.5778,
+      "step": 43872
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.1097592416409175e-09,
+      "loss": 0.6748,
+      "step": 43873
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0991651360603427e-09,
+      "loss": 0.5789,
+      "step": 43874
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0886218378214352e-09,
+      "loss": 0.6569,
+      "step": 43875
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0781293469808163e-09,
+      "loss": 0.5803,
+      "step": 43876
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.067687663589556e-09,
+      "loss": 0.5941,
+      "step": 43877
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.057296787703166e-09,
+      "loss": 0.663,
+      "step": 43878
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0469567193716057e-09,
+      "loss": 0.61,
+      "step": 43879
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0366674586492765e-09,
+      "loss": 0.5825,
+      "step": 43880
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0264290055883586e-09,
+      "loss": 0.6554,
+      "step": 43881
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0162413602399223e-09,
+      "loss": 0.6191,
+      "step": 43882
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.0061045226561482e-09,
+      "loss": 0.6223,
+      "step": 43883
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.96018492889217e-10,
+      "loss": 0.5847,
+      "step": 43884
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.859832709890882e-10,
+      "loss": 0.6194,
+      "step": 43885
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.759988570079426e-10,
+      "loss": 0.551,
+      "step": 43886
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.660652509968505e-10,
+      "loss": 0.5273,
+      "step": 43887
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.561824530046615e-10,
+      "loss": 0.5853,
+      "step": 43888
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.46350463082446e-10,
+      "loss": 0.5725,
+      "step": 43889
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.365692812812743e-10,
+      "loss": 0.6232,
+      "step": 43890
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.268389076488859e-10,
+      "loss": 0.6499,
+      "step": 43891
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.17159342235241e-10,
+      "loss": 0.6055,
+      "step": 43892
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.075305850902994e-10,
+      "loss": 0.5666,
+      "step": 43893
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.97952636262911e-10,
+      "loss": 0.5747,
+      "step": 43894
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.884254958008154e-10,
+      "loss": 0.6438,
+      "step": 43895
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.789491637528624e-10,
+      "loss": 0.5884,
+      "step": 43896
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.695236401679019e-10,
+      "loss": 0.6587,
+      "step": 43897
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.601489250936734e-10,
+      "loss": 0.6146,
+      "step": 43898
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.508250185768063e-10,
+      "loss": 0.593,
+      "step": 43899
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.415519206650402e-10,
+      "loss": 0.631,
+      "step": 43900
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.323296314061147e-10,
+      "loss": 0.5288,
+      "step": 43901
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.231581508455489e-10,
+      "loss": 0.6839,
+      "step": 43902
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.140374790321925e-10,
+      "loss": 0.6392,
+      "step": 43903
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 8.049676160104547e-10,
+      "loss": 0.5873,
+      "step": 43904
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.959485618280749e-10,
+      "loss": 0.529,
+      "step": 43905
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.869803165283518e-10,
+      "loss": 0.5855,
+      "step": 43906
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.780628801590251e-10,
+      "loss": 0.6013,
+      "step": 43907
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.691962527656138e-10,
+      "loss": 0.6097,
+      "step": 43908
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.603804343914167e-10,
+      "loss": 0.5983,
+      "step": 43909
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.516154250819529e-10,
+      "loss": 0.6141,
+      "step": 43910
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.429012248827416e-10,
+      "loss": 0.6295,
+      "step": 43911
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.342378338370815e-10,
+      "loss": 0.6077,
+      "step": 43912
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.256252519882711e-10,
+      "loss": 0.5337,
+      "step": 43913
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.170634793818298e-10,
+      "loss": 0.615,
+      "step": 43914
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.085525160599461e-10,
+      "loss": 0.6103,
+      "step": 43915
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.000923620670286e-10,
+      "loss": 0.6185,
+      "step": 43916
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.91683017445266e-10,
+      "loss": 0.6383,
+      "step": 43917
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.833244822368468e-10,
+      "loss": 0.7027,
+      "step": 43918
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.750167564850695e-10,
+      "loss": 0.595,
+      "step": 43919
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.667598402321229e-10,
+      "loss": 0.5835,
+      "step": 43920
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.585537335201952e-10,
+      "loss": 0.6298,
+      "step": 43921
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.503984363903648e-10,
+      "loss": 0.6086,
+      "step": 43922
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.422939488837099e-10,
+      "loss": 0.6466,
+      "step": 43923
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.342402710424189e-10,
+      "loss": 0.5972,
+      "step": 43924
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.262374029075701e-10,
+      "loss": 0.603,
+      "step": 43925
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.182853445180215e-10,
+      "loss": 0.6017,
+      "step": 43926
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.103840959170715e-10,
+      "loss": 0.6502,
+      "step": 43927
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 6.025336571424678e-10,
+      "loss": 0.6149,
+      "step": 43928
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.947340282341785e-10,
+      "loss": 0.5345,
+      "step": 43929
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.869852092332818e-10,
+      "loss": 0.5864,
+      "step": 43930
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.792872001775252e-10,
+      "loss": 0.6514,
+      "step": 43931
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.71640001107987e-10,
+      "loss": 0.5679,
+      "step": 43932
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.640436120613046e-10,
+      "loss": 0.6447,
+      "step": 43933
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.564980330785563e-10,
+      "loss": 0.6548,
+      "step": 43934
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.490032641952692e-10,
+      "loss": 0.5284,
+      "step": 43935
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.415593054514112e-10,
+      "loss": 0.5839,
+      "step": 43936
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.341661568847301e-10,
+      "loss": 0.6254,
+      "step": 43937
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.268238185329733e-10,
+      "loss": 0.5869,
+      "step": 43938
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.19532290431668e-10,
+      "loss": 0.6172,
+      "step": 43939
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.12291572619672e-10,
+      "loss": 0.6409,
+      "step": 43940
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 5.051016651336227e-10,
+      "loss": 0.6561,
+      "step": 43941
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.979625680090472e-10,
+      "loss": 0.5999,
+      "step": 43942
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.908742812825829e-10,
+      "loss": 0.586,
+      "step": 43943
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.838368049908671e-10,
+      "loss": 0.6042,
+      "step": 43944
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.768501391694269e-10,
+      "loss": 0.6056,
+      "step": 43945
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.699142838537895e-10,
+      "loss": 0.583,
+      "step": 43946
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.6302923907837193e-10,
+      "loss": 0.5916,
+      "step": 43947
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.5619500487870117e-10,
+      "loss": 0.58,
+      "step": 43948
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.4941158129030436e-10,
+      "loss": 0.5911,
+      "step": 43949
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.4267896834648827e-10,
+      "loss": 0.6226,
+      "step": 43950
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.3599716608277997e-10,
+      "loss": 0.578,
+      "step": 43951
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.29366174531376e-10,
+      "loss": 0.5878,
+      "step": 43952
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2278599372669316e-10,
+      "loss": 0.6095,
+      "step": 43953
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.1625662370314844e-10,
+      "loss": 0.5988,
+      "step": 43954
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.0977806449293836e-10,
+      "loss": 0.6116,
+      "step": 43955
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.033503161282593e-10,
+      "loss": 0.5896,
+      "step": 43956
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.9697337864352816e-10,
+      "loss": 0.6386,
+      "step": 43957
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.9064725206983124e-10,
+      "loss": 0.5719,
+      "step": 43958
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.843719364404752e-10,
+      "loss": 0.6027,
+      "step": 43959
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.7814743178654635e-10,
+      "loss": 0.6606,
+      "step": 43960
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.7197373813913083e-10,
+      "loss": 0.6072,
+      "step": 43961
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.658508555304252e-10,
+      "loss": 0.562,
+      "step": 43962
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.5977878399151565e-10,
+      "loss": 0.6622,
+      "step": 43963
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.537575235534885e-10,
+      "loss": 0.6358,
+      "step": 43964
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.4778707424631965e-10,
+      "loss": 0.5748,
+      "step": 43965
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.4186743609998517e-10,
+      "loss": 0.6129,
+      "step": 43966
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.359986091455714e-10,
+      "loss": 0.5742,
+      "step": 43967
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.301805934130542e-10,
+      "loss": 0.6453,
+      "step": 43968
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.2441338893018924e-10,
+      "loss": 0.5845,
+      "step": 43969
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.1869699572806276e-10,
+      "loss": 0.5723,
+      "step": 43970
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.130314138355406e-10,
+      "loss": 0.635,
+      "step": 43971
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.0741664328148844e-10,
+      "loss": 0.6258,
+      "step": 43972
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.0185268409255176e-10,
+      "loss": 0.6212,
+      "step": 43973
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.963395362998167e-10,
+      "loss": 0.6185,
+      "step": 43974
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.9087719992881844e-10,
+      "loss": 0.6399,
+      "step": 43975
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8546567500953306e-10,
+      "loss": 0.5936,
+      "step": 43976
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8010496156749554e-10,
+      "loss": 0.6374,
+      "step": 43977
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.7479505963157183e-10,
+      "loss": 0.5662,
+      "step": 43978
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.6953596922729695e-10,
+      "loss": 0.5496,
+      "step": 43979
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.6432769038242654e-10,
+      "loss": 0.5672,
+      "step": 43980
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.5917022312249576e-10,
+      "loss": 0.6269,
+      "step": 43981
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.540635674752601e-10,
+      "loss": 0.6341,
+      "step": 43982
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.490077234640342e-10,
+      "loss": 0.687,
+      "step": 43983
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.4400269111768404e-10,
+      "loss": 0.616,
+      "step": 43984
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.390484704595242e-10,
+      "loss": 0.6207,
+      "step": 43985
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.3414506151619997e-10,
+      "loss": 0.5337,
+      "step": 43986
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.2929246431102613e-10,
+      "loss": 0.6001,
+      "step": 43987
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.2449067886842757e-10,
+      "loss": 0.5666,
+      "step": 43988
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1973970521504963e-10,
+      "loss": 0.5649,
+      "step": 43989
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1503954337309673e-10,
+      "loss": 0.589,
+      "step": 43990
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1039019336810407e-10,
+      "loss": 0.5978,
+      "step": 43991
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.0579165522227607e-10,
+      "loss": 0.6254,
+      "step": 43992
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.0124392895892741e-10,
+      "loss": 0.5973,
+      "step": 43993
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.967470146013728e-10,
+      "loss": 0.6034,
+      "step": 43994
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.923009121729269e-10,
+      "loss": 0.5903,
+      "step": 43995
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.879056216969044e-10,
+      "loss": 0.621,
+      "step": 43996
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.8356114319328931e-10,
+      "loss": 0.6489,
+      "step": 43997
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.7926747668650656e-10,
+      "loss": 0.5989,
+      "step": 43998
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.7502462219765036e-10,
+      "loss": 0.5304,
+      "step": 43999
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.7083257974781497e-10,
+      "loss": 0.5393,
+      "step": 44000
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 44082,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 2000,
+  "total_flos": 1.4439702322248417e+19,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}