diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,73026 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 12166,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.479452054794521e-08,
+      "loss": 1.9902,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0958904109589042e-07,
+      "loss": 1.9961,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6438356164383561e-07,
+      "loss": 1.96,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1917808219178084e-07,
+      "loss": 2.0156,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.73972602739726e-07,
+      "loss": 1.9951,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2876712328767123e-07,
+      "loss": 2.0127,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.835616438356165e-07,
+      "loss": 1.9746,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.383561643835617e-07,
+      "loss": 1.9561,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.931506849315068e-07,
+      "loss": 1.9727,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.47945205479452e-07,
+      "loss": 1.9912,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.027397260273974e-07,
+      "loss": 2.0293,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.575342465753425e-07,
+      "loss": 1.9609,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.123287671232878e-07,
+      "loss": 1.9805,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.67123287671233e-07,
+      "loss": 1.9883,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.219178082191781e-07,
+      "loss": 1.9785,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.767123287671234e-07,
+      "loss": 1.9824,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.315068493150686e-07,
+      "loss": 1.9199,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.863013698630137e-07,
+      "loss": 1.9453,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.041095890410959e-06,
+      "loss": 1.9004,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.095890410958904e-06,
+      "loss": 1.8984,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1506849315068494e-06,
+      "loss": 1.8926,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2054794520547947e-06,
+      "loss": 1.877,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.26027397260274e-06,
+      "loss": 1.8477,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.315068493150685e-06,
+      "loss": 1.8691,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3698630136986302e-06,
+      "loss": 1.8193,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4246575342465755e-06,
+      "loss": 1.793,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4794520547945206e-06,
+      "loss": 1.8398,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.534246575342466e-06,
+      "loss": 1.7891,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5890410958904112e-06,
+      "loss": 1.7754,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6438356164383561e-06,
+      "loss": 1.7773,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6986301369863014e-06,
+      "loss": 1.7783,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7534246575342468e-06,
+      "loss": 1.7988,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.808219178082192e-06,
+      "loss": 1.833,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8630136986301372e-06,
+      "loss": 1.751,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9178082191780823e-06,
+      "loss": 1.748,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9726027397260274e-06,
+      "loss": 1.7734,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.027397260273973e-06,
+      "loss": 1.7373,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.082191780821918e-06,
+      "loss": 1.7305,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1369863013698635e-06,
+      "loss": 1.7061,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.191780821917808e-06,
+      "loss": 1.7354,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2465753424657537e-06,
+      "loss": 1.7002,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.301369863013699e-06,
+      "loss": 1.71,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.356164383561644e-06,
+      "loss": 1.7129,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4109589041095894e-06,
+      "loss": 1.7012,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4657534246575345e-06,
+      "loss": 1.709,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.52054794520548e-06,
+      "loss": 1.7207,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5753424657534247e-06,
+      "loss": 1.6846,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.63013698630137e-06,
+      "loss": 1.6943,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6849315068493153e-06,
+      "loss": 1.6934,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7397260273972604e-06,
+      "loss": 1.708,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.794520547945206e-06,
+      "loss": 1.6924,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.849315068493151e-06,
+      "loss": 1.7178,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9041095890410957e-06,
+      "loss": 1.7275,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9589041095890413e-06,
+      "loss": 1.6963,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0136986301369864e-06,
+      "loss": 1.6699,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.068493150684932e-06,
+      "loss": 1.6289,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.123287671232877e-06,
+      "loss": 1.6748,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1780821917808225e-06,
+      "loss": 1.6904,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2328767123287676e-06,
+      "loss": 1.6484,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2876712328767123e-06,
+      "loss": 1.6895,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.342465753424658e-06,
+      "loss": 1.6904,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.397260273972603e-06,
+      "loss": 1.6611,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4520547945205484e-06,
+      "loss": 1.6611,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5068493150684935e-06,
+      "loss": 1.6475,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5616438356164386e-06,
+      "loss": 1.6357,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.616438356164384e-06,
+      "loss": 1.6094,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.671232876712329e-06,
+      "loss": 1.7188,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7260273972602743e-06,
+      "loss": 1.6377,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7808219178082194e-06,
+      "loss": 1.6201,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8356164383561645e-06,
+      "loss": 1.5957,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.89041095890411e-06,
+      "loss": 1.6641,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.945205479452055e-06,
+      "loss": 1.5645,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.5889,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.054794520547946e-06,
+      "loss": 1.6289,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.109589041095891e-06,
+      "loss": 1.6309,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.164383561643836e-06,
+      "loss": 1.6533,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.219178082191781e-06,
+      "loss": 1.625,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.273972602739727e-06,
+      "loss": 1.6318,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.328767123287671e-06,
+      "loss": 1.6582,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.383561643835616e-06,
+      "loss": 1.6182,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.438356164383562e-06,
+      "loss": 1.6416,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.493150684931507e-06,
+      "loss": 1.5723,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5479452054794525e-06,
+      "loss": 1.7109,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.602739726027398e-06,
+      "loss": 1.6152,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.657534246575343e-06,
+      "loss": 1.6348,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.712328767123288e-06,
+      "loss": 1.5801,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.767123287671233e-06,
+      "loss": 1.5762,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.821917808219179e-06,
+      "loss": 1.5791,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.876712328767124e-06,
+      "loss": 1.6562,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.931506849315069e-06,
+      "loss": 1.5889,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.986301369863014e-06,
+      "loss": 1.6162,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.04109589041096e-06,
+      "loss": 1.5713,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.095890410958904e-06,
+      "loss": 1.6123,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1506849315068494e-06,
+      "loss": 1.6201,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2054794520547945e-06,
+      "loss": 1.5957,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.26027397260274e-06,
+      "loss": 1.5801,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3150684931506856e-06,
+      "loss": 1.6191,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.369863013698631e-06,
+      "loss": 1.6328,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.424657534246576e-06,
+      "loss": 1.6289,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.479452054794521e-06,
+      "loss": 1.5791,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.534246575342466e-06,
+      "loss": 1.5537,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.589041095890412e-06,
+      "loss": 1.5967,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.643835616438357e-06,
+      "loss": 1.5889,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.698630136986302e-06,
+      "loss": 1.5684,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.753424657534246e-06,
+      "loss": 1.5547,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8082191780821915e-06,
+      "loss": 1.5664,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.863013698630137e-06,
+      "loss": 1.6104,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9178082191780825e-06,
+      "loss": 1.6055,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.972602739726028e-06,
+      "loss": 1.5273,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.027397260273973e-06,
+      "loss": 1.5811,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.082191780821919e-06,
+      "loss": 1.5684,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.136986301369864e-06,
+      "loss": 1.5596,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.191780821917809e-06,
+      "loss": 1.5781,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.246575342465754e-06,
+      "loss": 1.624,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.301369863013699e-06,
+      "loss": 1.5381,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.356164383561645e-06,
+      "loss": 1.5557,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.41095890410959e-06,
+      "loss": 1.5801,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.465753424657535e-06,
+      "loss": 1.5566,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5205479452054794e-06,
+      "loss": 1.6064,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5753424657534245e-06,
+      "loss": 1.5625,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.630136986301371e-06,
+      "loss": 1.5498,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.684931506849316e-06,
+      "loss": 1.5752,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.739726027397261e-06,
+      "loss": 1.5557,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.794520547945206e-06,
+      "loss": 1.5342,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.849315068493151e-06,
+      "loss": 1.5566,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.904109589041097e-06,
+      "loss": 1.5371,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.958904109589042e-06,
+      "loss": 1.5908,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.013698630136987e-06,
+      "loss": 1.5732,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.068493150684932e-06,
+      "loss": 1.5928,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.123287671232877e-06,
+      "loss": 1.5703,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.178082191780823e-06,
+      "loss": 1.5439,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.232876712328768e-06,
+      "loss": 1.4863,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.287671232876713e-06,
+      "loss": 1.5381,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.342465753424658e-06,
+      "loss": 1.5664,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.397260273972603e-06,
+      "loss": 1.5156,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.452054794520549e-06,
+      "loss": 1.5195,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.506849315068494e-06,
+      "loss": 1.5693,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.561643835616439e-06,
+      "loss": 1.5312,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.616438356164384e-06,
+      "loss": 1.5771,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.671232876712329e-06,
+      "loss": 1.5352,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.726027397260276e-06,
+      "loss": 1.5283,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.78082191780822e-06,
+      "loss": 1.5664,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.835616438356164e-06,
+      "loss": 1.5508,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.89041095890411e-06,
+      "loss": 1.5566,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.945205479452055e-06,
+      "loss": 1.5293,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.5488,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.054794520547946e-06,
+      "loss": 1.5176,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.109589041095892e-06,
+      "loss": 1.5234,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.164383561643837e-06,
+      "loss": 1.5049,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.219178082191782e-06,
+      "loss": 1.5449,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.273972602739727e-06,
+      "loss": 1.5469,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.328767123287672e-06,
+      "loss": 1.5625,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.383561643835617e-06,
+      "loss": 1.5469,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.438356164383562e-06,
+      "loss": 1.5479,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.493150684931507e-06,
+      "loss": 1.5322,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.547945205479454e-06,
+      "loss": 1.5225,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.602739726027397e-06,
+      "loss": 1.5537,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.657534246575343e-06,
+      "loss": 1.5703,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.712328767123288e-06,
+      "loss": 1.5166,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.767123287671233e-06,
+      "loss": 1.5176,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.82191780821918e-06,
+      "loss": 1.5625,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.876712328767125e-06,
+      "loss": 1.5303,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.93150684931507e-06,
+      "loss": 1.541,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.986301369863015e-06,
+      "loss": 1.499,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.04109589041096e-06,
+      "loss": 1.5029,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.095890410958905e-06,
+      "loss": 1.5205,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.15068493150685e-06,
+      "loss": 1.5391,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.205479452054795e-06,
+      "loss": 1.4932,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.26027397260274e-06,
+      "loss": 1.5303,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.315068493150685e-06,
+      "loss": 1.5391,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.36986301369863e-06,
+      "loss": 1.5195,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.424657534246576e-06,
+      "loss": 1.4854,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.47945205479452e-06,
+      "loss": 1.5586,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.534246575342466e-06,
+      "loss": 1.5088,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.589041095890411e-06,
+      "loss": 1.5449,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.643835616438358e-06,
+      "loss": 1.5176,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.698630136986303e-06,
+      "loss": 1.5391,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.753424657534248e-06,
+      "loss": 1.4658,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.808219178082193e-06,
+      "loss": 1.4824,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.863013698630138e-06,
+      "loss": 1.5117,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.917808219178083e-06,
+      "loss": 1.5537,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972602739726028e-06,
+      "loss": 1.4658,
+      "step": 182
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0027397260273975e-05,
+      "loss": 1.4736,
+      "step": 183
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.008219178082192e-05,
+      "loss": 1.4971,
+      "step": 184
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0136986301369864e-05,
+      "loss": 1.5049,
+      "step": 185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0191780821917809e-05,
+      "loss": 1.5137,
+      "step": 186
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0246575342465754e-05,
+      "loss": 1.4893,
+      "step": 187
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0301369863013699e-05,
+      "loss": 1.4941,
+      "step": 188
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0356164383561644e-05,
+      "loss": 1.5049,
+      "step": 189
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0410958904109589e-05,
+      "loss": 1.5127,
+      "step": 190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0465753424657534e-05,
+      "loss": 1.4727,
+      "step": 191
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.052054794520548e-05,
+      "loss": 1.4893,
+      "step": 192
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0575342465753426e-05,
+      "loss": 1.5547,
+      "step": 193
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0630136986301371e-05,
+      "loss": 1.4902,
+      "step": 194
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0684931506849316e-05,
+      "loss": 1.4648,
+      "step": 195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0739726027397261e-05,
+      "loss": 1.4775,
+      "step": 196
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0794520547945206e-05,
+      "loss": 1.5107,
+      "step": 197
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0849315068493152e-05,
+      "loss": 1.5039,
+      "step": 198
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0904109589041097e-05,
+      "loss": 1.5254,
+      "step": 199
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0958904109589042e-05,
+      "loss": 1.5264,
+      "step": 200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1013698630136987e-05,
+      "loss": 1.5176,
+      "step": 201
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1068493150684932e-05,
+      "loss": 1.5322,
+      "step": 202
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1123287671232879e-05,
+      "loss": 1.4883,
+      "step": 203
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1178082191780824e-05,
+      "loss": 1.5459,
+      "step": 204
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1232876712328769e-05,
+      "loss": 1.5205,
+      "step": 205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1287671232876714e-05,
+      "loss": 1.5078,
+      "step": 206
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1342465753424659e-05,
+      "loss": 1.4961,
+      "step": 207
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1397260273972604e-05,
+      "loss": 1.4912,
+      "step": 208
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1452054794520548e-05,
+      "loss": 1.501,
+      "step": 209
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1506849315068493e-05,
+      "loss": 1.4297,
+      "step": 210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1561643835616438e-05,
+      "loss": 1.5098,
+      "step": 211
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1616438356164383e-05,
+      "loss": 1.5049,
+      "step": 212
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1671232876712331e-05,
+      "loss": 1.5078,
+      "step": 213
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1726027397260275e-05,
+      "loss": 1.5449,
+      "step": 214
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.178082191780822e-05,
+      "loss": 1.4707,
+      "step": 215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1835616438356165e-05,
+      "loss": 1.4824,
+      "step": 216
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.189041095890411e-05,
+      "loss": 1.5117,
+      "step": 217
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1945205479452055e-05,
+      "loss": 1.4961,
+      "step": 218
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2e-05,
+      "loss": 1.5127,
+      "step": 219
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2054794520547945e-05,
+      "loss": 1.4893,
+      "step": 220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.210958904109589e-05,
+      "loss": 1.501,
+      "step": 221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2164383561643837e-05,
+      "loss": 1.5078,
+      "step": 222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2219178082191782e-05,
+      "loss": 1.4385,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2273972602739727e-05,
+      "loss": 1.4697,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2328767123287673e-05,
+      "loss": 1.5029,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2383561643835618e-05,
+      "loss": 1.4688,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2438356164383563e-05,
+      "loss": 1.5049,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2493150684931508e-05,
+      "loss": 1.5049,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2547945205479453e-05,
+      "loss": 1.4785,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2602739726027398e-05,
+      "loss": 1.4766,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2657534246575343e-05,
+      "loss": 1.4844,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.271232876712329e-05,
+      "loss": 1.4951,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2767123287671235e-05,
+      "loss": 1.4541,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.282191780821918e-05,
+      "loss": 1.5215,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2876712328767125e-05,
+      "loss": 1.4814,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.293150684931507e-05,
+      "loss": 1.4922,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2986301369863015e-05,
+      "loss": 1.4756,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3041095890410959e-05,
+      "loss": 1.4766,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3095890410958904e-05,
+      "loss": 1.4922,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3150684931506849e-05,
+      "loss": 1.5039,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3205479452054794e-05,
+      "loss": 1.457,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3260273972602743e-05,
+      "loss": 1.4756,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3315068493150686e-05,
+      "loss": 1.5273,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3369863013698631e-05,
+      "loss": 1.4502,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3424657534246576e-05,
+      "loss": 1.4639,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3479452054794521e-05,
+      "loss": 1.4727,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3534246575342466e-05,
+      "loss": 1.4697,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3589041095890412e-05,
+      "loss": 1.5088,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3643835616438357e-05,
+      "loss": 1.5,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3698630136986302e-05,
+      "loss": 1.5303,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3753424657534247e-05,
+      "loss": 1.4365,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3808219178082194e-05,
+      "loss": 1.5254,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3863013698630139e-05,
+      "loss": 1.4492,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3917808219178084e-05,
+      "loss": 1.4551,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3972602739726029e-05,
+      "loss": 1.4746,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4027397260273974e-05,
+      "loss": 1.5205,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4082191780821919e-05,
+      "loss": 1.5234,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4136986301369864e-05,
+      "loss": 1.5195,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.419178082191781e-05,
+      "loss": 1.4717,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4246575342465754e-05,
+      "loss": 1.4785,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.43013698630137e-05,
+      "loss": 1.4648,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4356164383561646e-05,
+      "loss": 1.4951,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4410958904109591e-05,
+      "loss": 1.4609,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4465753424657537e-05,
+      "loss": 1.4971,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4520547945205482e-05,
+      "loss": 1.4541,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4575342465753427e-05,
+      "loss": 1.5244,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.463013698630137e-05,
+      "loss": 1.46,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4684931506849315e-05,
+      "loss": 1.5117,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.473972602739726e-05,
+      "loss": 1.4248,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4794520547945205e-05,
+      "loss": 1.4336,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.484931506849315e-05,
+      "loss": 1.4863,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4904109589041097e-05,
+      "loss": 1.4697,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4958904109589042e-05,
+      "loss": 1.4932,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5013698630136988e-05,
+      "loss": 1.4531,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5068493150684933e-05,
+      "loss": 1.5234,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5123287671232878e-05,
+      "loss": 1.4717,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5178082191780823e-05,
+      "loss": 1.498,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5232876712328768e-05,
+      "loss": 1.4746,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5287671232876713e-05,
+      "loss": 1.4834,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5342465753424658e-05,
+      "loss": 1.4824,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5397260273972603e-05,
+      "loss": 1.4775,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.545205479452055e-05,
+      "loss": 1.4756,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5506849315068497e-05,
+      "loss": 1.4717,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.556164383561644e-05,
+      "loss": 1.4121,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5616438356164384e-05,
+      "loss": 1.5273,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.567123287671233e-05,
+      "loss": 1.4365,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5726027397260274e-05,
+      "loss": 1.4609,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.578082191780822e-05,
+      "loss": 1.459,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5835616438356164e-05,
+      "loss": 1.4287,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.589041095890411e-05,
+      "loss": 1.5039,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5945205479452054e-05,
+      "loss": 1.5078,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.4551,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6054794520547948e-05,
+      "loss": 1.498,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6109589041095893e-05,
+      "loss": 1.4375,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6164383561643838e-05,
+      "loss": 1.4932,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6219178082191783e-05,
+      "loss": 1.5117,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6273972602739728e-05,
+      "loss": 1.4902,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6328767123287673e-05,
+      "loss": 1.4902,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.638356164383562e-05,
+      "loss": 1.5098,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6438356164383563e-05,
+      "loss": 1.46,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.649315068493151e-05,
+      "loss": 1.4443,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6547945205479454e-05,
+      "loss": 1.4316,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.66027397260274e-05,
+      "loss": 1.502,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6657534246575344e-05,
+      "loss": 1.5088,
+      "step": 304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.671232876712329e-05,
+      "loss": 1.4658,
+      "step": 305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6767123287671234e-05,
+      "loss": 1.5068,
+      "step": 306
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.682191780821918e-05,
+      "loss": 1.4814,
+      "step": 307
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6876712328767124e-05,
+      "loss": 1.4346,
+      "step": 308
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.693150684931507e-05,
+      "loss": 1.4473,
+      "step": 309
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6986301369863014e-05,
+      "loss": 1.4482,
+      "step": 310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7041095890410963e-05,
+      "loss": 1.4697,
+      "step": 311
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7095890410958908e-05,
+      "loss": 1.4941,
+      "step": 312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.715068493150685e-05,
+      "loss": 1.4873,
+      "step": 313
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7205479452054795e-05,
+      "loss": 1.5215,
+      "step": 314
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.726027397260274e-05,
+      "loss": 1.5137,
+      "step": 315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7315068493150685e-05,
+      "loss": 1.4795,
+      "step": 316
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.736986301369863e-05,
+      "loss": 1.4375,
+      "step": 317
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7424657534246575e-05,
+      "loss": 1.4697,
+      "step": 318
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.747945205479452e-05,
+      "loss": 1.4473,
+      "step": 319
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7534246575342465e-05,
+      "loss": 1.4639,
+      "step": 320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7589041095890414e-05,
+      "loss": 1.459,
+      "step": 321
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.764383561643836e-05,
+      "loss": 1.4863,
+      "step": 322
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7698630136986304e-05,
+      "loss": 1.4814,
+      "step": 323
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.775342465753425e-05,
+      "loss": 1.4521,
+      "step": 324
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7808219178082194e-05,
+      "loss": 1.4512,
+      "step": 325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.786301369863014e-05,
+      "loss": 1.5312,
+      "step": 326
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7917808219178085e-05,
+      "loss": 1.501,
+      "step": 327
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.797260273972603e-05,
+      "loss": 1.4854,
+      "step": 328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8027397260273975e-05,
+      "loss": 1.4531,
+      "step": 329
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.808219178082192e-05,
+      "loss": 1.4795,
+      "step": 330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8136986301369865e-05,
+      "loss": 1.4697,
+      "step": 331
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.819178082191781e-05,
+      "loss": 1.4922,
+      "step": 332
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8246575342465755e-05,
+      "loss": 1.5176,
+      "step": 333
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.83013698630137e-05,
+      "loss": 1.4912,
+      "step": 334
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8356164383561645e-05,
+      "loss": 1.4814,
+      "step": 335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.841095890410959e-05,
+      "loss": 1.4512,
+      "step": 336
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8465753424657535e-05,
+      "loss": 1.5127,
+      "step": 337
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.852054794520548e-05,
+      "loss": 1.4775,
+      "step": 338
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8575342465753426e-05,
+      "loss": 1.4121,
+      "step": 339
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.863013698630137e-05,
+      "loss": 1.4824,
+      "step": 340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8684931506849316e-05,
+      "loss": 1.4805,
+      "step": 341
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.873972602739726e-05,
+      "loss": 1.4668,
+      "step": 342
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8794520547945206e-05,
+      "loss": 1.4814,
+      "step": 343
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.884931506849315e-05,
+      "loss": 1.5088,
+      "step": 344
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8904109589041096e-05,
+      "loss": 1.4824,
+      "step": 345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.895890410958904e-05,
+      "loss": 1.46,
+      "step": 346
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9013698630136986e-05,
+      "loss": 1.4043,
+      "step": 347
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.906849315068493e-05,
+      "loss": 1.4502,
+      "step": 348
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9123287671232877e-05,
+      "loss": 1.4434,
+      "step": 349
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9178082191780822e-05,
+      "loss": 1.4453,
+      "step": 350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.923287671232877e-05,
+      "loss": 1.4863,
+      "step": 351
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9287671232876715e-05,
+      "loss": 1.4326,
+      "step": 352
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.934246575342466e-05,
+      "loss": 1.4688,
+      "step": 353
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9397260273972606e-05,
+      "loss": 1.4346,
+      "step": 354
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.945205479452055e-05,
+      "loss": 1.4912,
+      "step": 355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9506849315068496e-05,
+      "loss": 1.4766,
+      "step": 356
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.956164383561644e-05,
+      "loss": 1.4541,
+      "step": 357
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9616438356164386e-05,
+      "loss": 1.4912,
+      "step": 358
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.967123287671233e-05,
+      "loss": 1.4678,
+      "step": 359
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9726027397260276e-05,
+      "loss": 1.4658,
+      "step": 360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978082191780822e-05,
+      "loss": 1.4678,
+      "step": 361
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9835616438356166e-05,
+      "loss": 1.4365,
+      "step": 362
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.989041095890411e-05,
+      "loss": 1.4453,
+      "step": 363
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9945205479452057e-05,
+      "loss": 1.4736,
+      "step": 364
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.4746,
+      "step": 365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999645650256e-05,
+      "loss": 1.5137,
+      "step": 366
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999858260104e-05,
+      "loss": 1.4336,
+      "step": 367
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996810852434e-05,
+      "loss": 1.5156,
+      "step": 368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999433040456e-05,
+      "loss": 1.4453,
+      "step": 369
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999991141257594e-05,
+      "loss": 1.4688,
+      "step": 370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999987243411763e-05,
+      "loss": 1.4639,
+      "step": 371
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999982636867345e-05,
+      "loss": 1.4219,
+      "step": 372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999977321624663e-05,
+      "loss": 1.4189,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999971297684094e-05,
+      "loss": 1.4121,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999964565046067e-05,
+      "loss": 1.4346,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999995712371106e-05,
+      "loss": 1.457,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99999489736796e-05,
+      "loss": 1.4443,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999940114952255e-05,
+      "loss": 1.4658,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999930547529666e-05,
+      "loss": 1.4551,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999920271412507e-05,
+      "loss": 1.4619,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999909286601505e-05,
+      "loss": 1.4854,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999989759309744e-05,
+      "loss": 1.4238,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988519090114e-05,
+      "loss": 1.4395,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999872080013478e-05,
+      "loss": 1.502,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999985826043539e-05,
+      "loss": 1.4658,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999984373216786e-05,
+      "loss": 1.4854,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999828495211905e-05,
+      "loss": 1.418,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999812549568614e-05,
+      "loss": 1.4395,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999795895239114e-05,
+      "loss": 1.4561,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999778532224588e-05,
+      "loss": 1.4951,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999760460526265e-05,
+      "loss": 1.5137,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999741680145423e-05,
+      "loss": 1.458,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999722191083394e-05,
+      "loss": 1.4971,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999701993341562e-05,
+      "loss": 1.498,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999681086921355e-05,
+      "loss": 1.4424,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999965947182426e-05,
+      "loss": 1.4346,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999637148051803e-05,
+      "loss": 1.4023,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999614115605566e-05,
+      "loss": 1.4287,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999590374487184e-05,
+      "loss": 1.4844,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999565924698342e-05,
+      "loss": 1.4531,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999954076624077e-05,
+      "loss": 1.4521,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999514899116246e-05,
+      "loss": 1.4473,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999488323326612e-05,
+      "loss": 1.4834,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999461038873748e-05,
+      "loss": 1.4482,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999433045759583e-05,
+      "loss": 1.4844,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999404343986112e-05,
+      "loss": 1.4766,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999374933555356e-05,
+      "loss": 1.4316,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999344814469406e-05,
+      "loss": 1.5117,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999313986730397e-05,
+      "loss": 1.4355,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999282450340514e-05,
+      "loss": 1.4795,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999925020530199e-05,
+      "loss": 1.4072,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999921725161711e-05,
+      "loss": 1.4443,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999918358928821e-05,
+      "loss": 1.4365,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999149218317677e-05,
+      "loss": 1.46,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999911413870795e-05,
+      "loss": 1.457,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99990783504615e-05,
+      "loss": 1.4854,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999041853580883e-05,
+      "loss": 1.4707,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999004648068673e-05,
+      "loss": 1.4658,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998966733927513e-05,
+      "loss": 1.4287,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998928111160087e-05,
+      "loss": 1.498,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999888877976913e-05,
+      "loss": 1.4316,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998848739757434e-05,
+      "loss": 1.4424,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998807991127836e-05,
+      "loss": 1.459,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998766533883223e-05,
+      "loss": 1.4219,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998724368026526e-05,
+      "loss": 1.4883,
+      "step": 425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999868149356075e-05,
+      "loss": 1.4951,
+      "step": 426
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998637910488915e-05,
+      "loss": 1.4092,
+      "step": 427
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998593618814125e-05,
+      "loss": 1.4473,
+      "step": 428
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998548618539507e-05,
+      "loss": 1.4365,
+      "step": 429
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998502909668256e-05,
+      "loss": 1.4658,
+      "step": 430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998456492203613e-05,
+      "loss": 1.4932,
+      "step": 431
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999840936614886e-05,
+      "loss": 1.4629,
+      "step": 432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999836153150735e-05,
+      "loss": 1.4531,
+      "step": 433
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998312988282458e-05,
+      "loss": 1.4023,
+      "step": 434
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998263736477638e-05,
+      "loss": 1.4375,
+      "step": 435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999821377609637e-05,
+      "loss": 1.4375,
+      "step": 436
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998163107142197e-05,
+      "loss": 1.4248,
+      "step": 437
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998111729618715e-05,
+      "loss": 1.4346,
+      "step": 438
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998059643529557e-05,
+      "loss": 1.4277,
+      "step": 439
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998006848878422e-05,
+      "loss": 1.4258,
+      "step": 440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999795334566905e-05,
+      "loss": 1.46,
+      "step": 441
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997899133905227e-05,
+      "loss": 1.4326,
+      "step": 442
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997844213590803e-05,
+      "loss": 1.4395,
+      "step": 443
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997788584729667e-05,
+      "loss": 1.4551,
+      "step": 444
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997732247325758e-05,
+      "loss": 1.4512,
+      "step": 445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999767520138307e-05,
+      "loss": 1.4297,
+      "step": 446
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999761744690565e-05,
+      "loss": 1.4277,
+      "step": 447
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997558983897593e-05,
+      "loss": 1.418,
+      "step": 448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999749981236303e-05,
+      "loss": 1.4238,
+      "step": 449
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997439932306166e-05,
+      "loss": 1.4199,
+      "step": 450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999737934373124e-05,
+      "loss": 1.4424,
+      "step": 451
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997318046642545e-05,
+      "loss": 1.4414,
+      "step": 452
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997256041044427e-05,
+      "loss": 1.457,
+      "step": 453
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997193326941283e-05,
+      "loss": 1.4658,
+      "step": 454
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997129904337553e-05,
+      "loss": 1.4434,
+      "step": 455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997065773237736e-05,
+      "loss": 1.4443,
+      "step": 456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999700093364637e-05,
+      "loss": 1.4082,
+      "step": 457
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996935385568057e-05,
+      "loss": 1.4473,
+      "step": 458
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999686912900744e-05,
+      "loss": 1.4375,
+      "step": 459
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996802163969214e-05,
+      "loss": 1.3906,
+      "step": 460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999673449045813e-05,
+      "loss": 1.458,
+      "step": 461
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996666108478972e-05,
+      "loss": 1.417,
+      "step": 462
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99965970180366e-05,
+      "loss": 1.4404,
+      "step": 463
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996527219135902e-05,
+      "loss": 1.4473,
+      "step": 464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996456711781823e-05,
+      "loss": 1.4736,
+      "step": 465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996385495979366e-05,
+      "loss": 1.4658,
+      "step": 466
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996313571733578e-05,
+      "loss": 1.4385,
+      "step": 467
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996240939049554e-05,
+      "loss": 1.4287,
+      "step": 468
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996167597932436e-05,
+      "loss": 1.4277,
+      "step": 469
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996093548387433e-05,
+      "loss": 1.4062,
+      "step": 470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996018790419785e-05,
+      "loss": 1.4668,
+      "step": 471
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995943324034792e-05,
+      "loss": 1.4287,
+      "step": 472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995867149237802e-05,
+      "loss": 1.418,
+      "step": 473
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995790266034213e-05,
+      "loss": 1.4785,
+      "step": 474
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995712674429474e-05,
+      "loss": 1.4678,
+      "step": 475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999563437442909e-05,
+      "loss": 1.4697,
+      "step": 476
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99955553660386e-05,
+      "loss": 1.4854,
+      "step": 477
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995475649263605e-05,
+      "loss": 1.4229,
+      "step": 478
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995395224109764e-05,
+      "loss": 1.4453,
+      "step": 479
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999531409058277e-05,
+      "loss": 1.4229,
+      "step": 480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999523224868837e-05,
+      "loss": 1.4092,
+      "step": 481
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995149698432365e-05,
+      "loss": 1.4053,
+      "step": 482
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995066439820612e-05,
+      "loss": 1.4404,
+      "step": 483
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994982472859004e-05,
+      "loss": 1.3789,
+      "step": 484
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99948977975535e-05,
+      "loss": 1.4336,
+      "step": 485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999481241391009e-05,
+      "loss": 1.4395,
+      "step": 486
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994726321934832e-05,
+      "loss": 1.4482,
+      "step": 487
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994639521633826e-05,
+      "loss": 1.4492,
+      "step": 488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994552013013227e-05,
+      "loss": 1.4521,
+      "step": 489
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999446379607923e-05,
+      "loss": 1.4326,
+      "step": 490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994374870838095e-05,
+      "loss": 1.4863,
+      "step": 491
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994285237296112e-05,
+      "loss": 1.4443,
+      "step": 492
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999419489545965e-05,
+      "loss": 1.4189,
+      "step": 493
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994103845335097e-05,
+      "loss": 1.4404,
+      "step": 494
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994012086928913e-05,
+      "loss": 1.4453,
+      "step": 495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993919620247598e-05,
+      "loss": 1.4043,
+      "step": 496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993826445297705e-05,
+      "loss": 1.416,
+      "step": 497
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999373256208584e-05,
+      "loss": 1.4463,
+      "step": 498
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993637970618655e-05,
+      "loss": 1.4395,
+      "step": 499
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993542670902853e-05,
+      "loss": 1.418,
+      "step": 500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999344666294519e-05,
+      "loss": 1.4062,
+      "step": 501
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993349946752464e-05,
+      "loss": 1.417,
+      "step": 502
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999325252233154e-05,
+      "loss": 1.4043,
+      "step": 503
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993154389689313e-05,
+      "loss": 1.4189,
+      "step": 504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993055548832742e-05,
+      "loss": 1.4316,
+      "step": 505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999295599976883e-05,
+      "loss": 1.4434,
+      "step": 506
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992855742504633e-05,
+      "loss": 1.4629,
+      "step": 507
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999275477704726e-05,
+      "loss": 1.4658,
+      "step": 508
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999265310340386e-05,
+      "loss": 1.4043,
+      "step": 509
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992550721581637e-05,
+      "loss": 1.4102,
+      "step": 510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992447631587858e-05,
+      "loss": 1.416,
+      "step": 511
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992343833429814e-05,
+      "loss": 1.4336,
+      "step": 512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992239327114876e-05,
+      "loss": 1.4346,
+      "step": 513
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999213411265044e-05,
+      "loss": 1.4424,
+      "step": 514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992028190043965e-05,
+      "loss": 1.46,
+      "step": 515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999192155930296e-05,
+      "loss": 1.4102,
+      "step": 516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999181422043498e-05,
+      "loss": 1.458,
+      "step": 517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991706173447638e-05,
+      "loss": 1.4473,
+      "step": 518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991597418348582e-05,
+      "loss": 1.4756,
+      "step": 519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999148795514552e-05,
+      "loss": 1.4102,
+      "step": 520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999137778384622e-05,
+      "loss": 1.4248,
+      "step": 521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991266904458477e-05,
+      "loss": 1.4102,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991155316990157e-05,
+      "loss": 1.4707,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991043021449168e-05,
+      "loss": 1.4131,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990930017843465e-05,
+      "loss": 1.4131,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990816306181057e-05,
+      "loss": 1.4004,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990701886470006e-05,
+      "loss": 1.4287,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999058675871842e-05,
+      "loss": 1.4072,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990470922934454e-05,
+      "loss": 1.3779,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999035437912632e-05,
+      "loss": 1.4277,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999023712730228e-05,
+      "loss": 1.4023,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999011916747064e-05,
+      "loss": 1.4229,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990000499639762e-05,
+      "loss": 1.4727,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989881123818054e-05,
+      "loss": 1.4482,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998976104001398e-05,
+      "loss": 1.4834,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998964024823604e-05,
+      "loss": 1.4453,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998951874849281e-05,
+      "loss": 1.4287,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989396540792886e-05,
+      "loss": 1.4004,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989273625144943e-05,
+      "loss": 1.4023,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989150001557678e-05,
+      "loss": 1.4551,
+      "step": 540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989025670039863e-05,
+      "loss": 1.4482,
+      "step": 541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988900630600306e-05,
+      "loss": 1.3906,
+      "step": 542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988774883247864e-05,
+      "loss": 1.4258,
+      "step": 543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988648427991452e-05,
+      "loss": 1.4668,
+      "step": 544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988521264840034e-05,
+      "loss": 1.3877,
+      "step": 545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998839339380262e-05,
+      "loss": 1.3711,
+      "step": 546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988264814888275e-05,
+      "loss": 1.4219,
+      "step": 547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988135528106105e-05,
+      "loss": 1.4658,
+      "step": 548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998800553346528e-05,
+      "loss": 1.4189,
+      "step": 549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998787483097501e-05,
+      "loss": 1.3643,
+      "step": 550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987743420644554e-05,
+      "loss": 1.4307,
+      "step": 551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987611302483226e-05,
+      "loss": 1.4111,
+      "step": 552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987478476500395e-05,
+      "loss": 1.4297,
+      "step": 553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987344942705472e-05,
+      "loss": 1.4121,
+      "step": 554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987210701107915e-05,
+      "loss": 1.4785,
+      "step": 555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987075751717243e-05,
+      "loss": 1.3926,
+      "step": 556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986940094543022e-05,
+      "loss": 1.377,
+      "step": 557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998680372959486e-05,
+      "loss": 1.4082,
+      "step": 558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986666656882426e-05,
+      "loss": 1.4004,
+      "step": 559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986528876415428e-05,
+      "loss": 1.4346,
+      "step": 560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986390388203638e-05,
+      "loss": 1.415,
+      "step": 561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998625119225687e-05,
+      "loss": 1.3779,
+      "step": 562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998611128858498e-05,
+      "loss": 1.4053,
+      "step": 563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985970677197894e-05,
+      "loss": 1.4346,
+      "step": 564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998582935810557e-05,
+      "loss": 1.4219,
+      "step": 565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985687331318027e-05,
+      "loss": 1.4111,
+      "step": 566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985544596845328e-05,
+      "loss": 1.4473,
+      "step": 567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998540115469759e-05,
+      "loss": 1.4102,
+      "step": 568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998525700488498e-05,
+      "loss": 1.4258,
+      "step": 569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998511214741771e-05,
+      "loss": 1.4736,
+      "step": 570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998496658230605e-05,
+      "loss": 1.4238,
+      "step": 571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984820309560314e-05,
+      "loss": 1.4463,
+      "step": 572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998467332919087e-05,
+      "loss": 1.4414,
+      "step": 573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998452564120813e-05,
+      "loss": 1.4082,
+      "step": 574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984377245622565e-05,
+      "loss": 1.4229,
+      "step": 575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998422814244469e-05,
+      "loss": 1.457,
+      "step": 576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998407833168508e-05,
+      "loss": 1.4121,
+      "step": 577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998392781335434e-05,
+      "loss": 1.3984,
+      "step": 578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983776587463137e-05,
+      "loss": 1.417,
+      "step": 579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983624654022198e-05,
+      "loss": 1.4092,
+      "step": 580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983472013042284e-05,
+      "loss": 1.4102,
+      "step": 581
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983318664534218e-05,
+      "loss": 1.4072,
+      "step": 582
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983164608508863e-05,
+      "loss": 1.458,
+      "step": 583
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983009844977137e-05,
+      "loss": 1.4375,
+      "step": 584
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998285437395001e-05,
+      "loss": 1.4541,
+      "step": 585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99826981954385e-05,
+      "loss": 1.459,
+      "step": 586
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982541309453675e-05,
+      "loss": 1.4209,
+      "step": 587
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998238371600665e-05,
+      "loss": 1.4082,
+      "step": 588
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982225415108603e-05,
+      "loss": 1.416,
+      "step": 589
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982066406770744e-05,
+      "loss": 1.3984,
+      "step": 590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981906691004344e-05,
+      "loss": 1.4385,
+      "step": 591
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981746267820723e-05,
+      "loss": 1.418,
+      "step": 592
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998158513723125e-05,
+      "loss": 1.4336,
+      "step": 593
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998142329924735e-05,
+      "loss": 1.4053,
+      "step": 594
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998126075388048e-05,
+      "loss": 1.4756,
+      "step": 595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981097501142167e-05,
+      "loss": 1.4355,
+      "step": 596
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980933541043983e-05,
+      "loss": 1.4893,
+      "step": 597
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980768873597543e-05,
+      "loss": 1.4287,
+      "step": 598
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980603498814516e-05,
+      "loss": 1.4277,
+      "step": 599
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998043741670663e-05,
+      "loss": 1.3955,
+      "step": 600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998027062728565e-05,
+      "loss": 1.3848,
+      "step": 601
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980103130563397e-05,
+      "loss": 1.4131,
+      "step": 602
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979934926551738e-05,
+      "loss": 1.4492,
+      "step": 603
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979766015262597e-05,
+      "loss": 1.418,
+      "step": 604
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979596396707945e-05,
+      "loss": 1.4316,
+      "step": 605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979426070899804e-05,
+      "loss": 1.3838,
+      "step": 606
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997925503785024e-05,
+      "loss": 1.4561,
+      "step": 607
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997908329757138e-05,
+      "loss": 1.4346,
+      "step": 608
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997891085007539e-05,
+      "loss": 1.4434,
+      "step": 609
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978737695374497e-05,
+      "loss": 1.3936,
+      "step": 610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997856383348097e-05,
+      "loss": 1.3975,
+      "step": 611
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978389264407126e-05,
+      "loss": 1.4307,
+      "step": 612
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978213988165343e-05,
+      "loss": 1.4453,
+      "step": 613
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997803800476804e-05,
+      "loss": 1.4502,
+      "step": 614
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977861314227692e-05,
+      "loss": 1.417,
+      "step": 615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977683916556816e-05,
+      "loss": 1.4395,
+      "step": 616
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977505811767988e-05,
+      "loss": 1.3896,
+      "step": 617
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977326999873826e-05,
+      "loss": 1.4141,
+      "step": 618
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977147480887005e-05,
+      "loss": 1.4014,
+      "step": 619
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976967254820252e-05,
+      "loss": 1.3916,
+      "step": 620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976786321686335e-05,
+      "loss": 1.3975,
+      "step": 621
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976604681498076e-05,
+      "loss": 1.4219,
+      "step": 622
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997642233426835e-05,
+      "loss": 1.3965,
+      "step": 623
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976239280010075e-05,
+      "loss": 1.3467,
+      "step": 624
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997605551873623e-05,
+      "loss": 1.3984,
+      "step": 625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997587105045984e-05,
+      "loss": 1.418,
+      "step": 626
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997568587519397e-05,
+      "loss": 1.4297,
+      "step": 627
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975499992951746e-05,
+      "loss": 1.46,
+      "step": 628
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997531340374635e-05,
+      "loss": 1.4248,
+      "step": 629
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997512610759099e-05,
+      "loss": 1.415,
+      "step": 630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974938104498954e-05,
+      "loss": 1.4229,
+      "step": 631
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997474939448356e-05,
+      "loss": 1.417,
+      "step": 632
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997455997755818e-05,
+      "loss": 1.4092,
+      "step": 633
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974369853736246e-05,
+      "loss": 1.4395,
+      "step": 634
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997417902303122e-05,
+      "loss": 1.4297,
+      "step": 635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973987485456635e-05,
+      "loss": 1.4023,
+      "step": 636
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997379524102606e-05,
+      "loss": 1.377,
+      "step": 637
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973602289753126e-05,
+      "loss": 1.4414,
+      "step": 638
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99734086316515e-05,
+      "loss": 1.4287,
+      "step": 639
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997321426673491e-05,
+      "loss": 1.458,
+      "step": 640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997301919501713e-05,
+      "loss": 1.4648,
+      "step": 641
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972823416511988e-05,
+      "loss": 1.4756,
+      "step": 642
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972626931233355e-05,
+      "loss": 1.416,
+      "step": 643
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972429739195155e-05,
+      "loss": 1.416,
+      "step": 644
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972231840411365e-05,
+      "loss": 1.417,
+      "step": 645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972033234896014e-05,
+      "loss": 1.4189,
+      "step": 646
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971833922663172e-05,
+      "loss": 1.4316,
+      "step": 647
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997163390372696e-05,
+      "loss": 1.3623,
+      "step": 648
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971433178101566e-05,
+      "loss": 1.3877,
+      "step": 649
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971231745801205e-05,
+      "loss": 1.4307,
+      "step": 650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971029606840155e-05,
+      "loss": 1.3975,
+      "step": 651
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997082676123274e-05,
+      "loss": 1.3896,
+      "step": 652
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970623208993346e-05,
+      "loss": 1.4424,
+      "step": 653
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970418950136383e-05,
+      "loss": 1.457,
+      "step": 654
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997021398467634e-05,
+      "loss": 1.4668,
+      "step": 655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970008312627733e-05,
+      "loss": 1.3965,
+      "step": 656
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996980193400514e-05,
+      "loss": 1.3809,
+      "step": 657
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99695948488232e-05,
+      "loss": 1.4111,
+      "step": 658
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969387057096567e-05,
+      "loss": 1.3721,
+      "step": 659
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969178558839984e-05,
+      "loss": 1.3691,
+      "step": 660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996896935406822e-05,
+      "loss": 1.4336,
+      "step": 661
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968759442796104e-05,
+      "loss": 1.4453,
+      "step": 662
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968548825038515e-05,
+      "loss": 1.4092,
+      "step": 663
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968337500810372e-05,
+      "loss": 1.4141,
+      "step": 664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968125470126655e-05,
+      "loss": 1.4121,
+      "step": 665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967912733002396e-05,
+      "loss": 1.3594,
+      "step": 666
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967699289452662e-05,
+      "loss": 1.373,
+      "step": 667
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967485139492585e-05,
+      "loss": 1.4229,
+      "step": 668
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967270283137347e-05,
+      "loss": 1.4463,
+      "step": 669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967054720402163e-05,
+      "loss": 1.3877,
+      "step": 670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996683845130232e-05,
+      "loss": 1.3857,
+      "step": 671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966621475853142e-05,
+      "loss": 1.4141,
+      "step": 672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996640379407e-05,
+      "loss": 1.4395,
+      "step": 673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996618540596833e-05,
+      "loss": 1.417,
+      "step": 674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965966311563606e-05,
+      "loss": 1.3887,
+      "step": 675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965746510871354e-05,
+      "loss": 1.4287,
+      "step": 676
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996552600390715e-05,
+      "loss": 1.4199,
+      "step": 677
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965304790686626e-05,
+      "loss": 1.4092,
+      "step": 678
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965082871225455e-05,
+      "loss": 1.4092,
+      "step": 679
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964860245539367e-05,
+      "loss": 1.3994,
+      "step": 680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996463691364414e-05,
+      "loss": 1.3701,
+      "step": 681
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99644128755556e-05,
+      "loss": 1.4082,
+      "step": 682
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996418813128962e-05,
+      "loss": 1.4404,
+      "step": 683
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963962680862135e-05,
+      "loss": 1.4102,
+      "step": 684
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996373652428912e-05,
+      "loss": 1.4219,
+      "step": 685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963509661586607e-05,
+      "loss": 1.376,
+      "step": 686
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963282092770667e-05,
+      "loss": 1.3965,
+      "step": 687
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963053817857426e-05,
+      "loss": 1.3926,
+      "step": 688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996282483686307e-05,
+      "loss": 1.4219,
+      "step": 689
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962595149803822e-05,
+      "loss": 1.4014,
+      "step": 690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996236475669596e-05,
+      "loss": 1.4082,
+      "step": 691
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996213365755581e-05,
+      "loss": 1.3848,
+      "step": 692
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996190185239976e-05,
+      "loss": 1.4189,
+      "step": 693
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961669341244225e-05,
+      "loss": 1.4062,
+      "step": 694
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961436124105692e-05,
+      "loss": 1.458,
+      "step": 695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996120220100068e-05,
+      "loss": 1.4678,
+      "step": 696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996096757194578e-05,
+      "loss": 1.3877,
+      "step": 697
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996073223695761e-05,
+      "loss": 1.3887,
+      "step": 698
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960496196052852e-05,
+      "loss": 1.4297,
+      "step": 699
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960259449248234e-05,
+      "loss": 1.3701,
+      "step": 700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960021996560533e-05,
+      "loss": 1.3994,
+      "step": 701
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959783838006576e-05,
+      "loss": 1.4277,
+      "step": 702
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995954497360325e-05,
+      "loss": 1.3965,
+      "step": 703
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995930540336747e-05,
+      "loss": 1.4082,
+      "step": 704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959065127316222e-05,
+      "loss": 1.3662,
+      "step": 705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958824145466536e-05,
+      "loss": 1.3887,
+      "step": 706
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958582457835484e-05,
+      "loss": 1.3818,
+      "step": 707
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99583400644402e-05,
+      "loss": 1.4375,
+      "step": 708
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958096965297863e-05,
+      "loss": 1.4082,
+      "step": 709
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957853160425694e-05,
+      "loss": 1.3555,
+      "step": 710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995760864984098e-05,
+      "loss": 1.3955,
+      "step": 711
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957363433561046e-05,
+      "loss": 1.3848,
+      "step": 712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957117511603267e-05,
+      "loss": 1.3564,
+      "step": 713
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956870883985078e-05,
+      "loss": 1.4141,
+      "step": 714
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956623550723954e-05,
+      "loss": 1.4316,
+      "step": 715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995637551183742e-05,
+      "loss": 1.3936,
+      "step": 716
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956126767343062e-05,
+      "loss": 1.4053,
+      "step": 717
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99558773172585e-05,
+      "loss": 1.4297,
+      "step": 718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955627161601424e-05,
+      "loss": 1.418,
+      "step": 719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955376300389552e-05,
+      "loss": 1.4062,
+      "step": 720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955124733640667e-05,
+      "loss": 1.4316,
+      "step": 721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99548724613726e-05,
+      "loss": 1.3809,
+      "step": 722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995461948360322e-05,
+      "loss": 1.4268,
+      "step": 723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954365800350465e-05,
+      "loss": 1.3877,
+      "step": 724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995411141163231e-05,
+      "loss": 1.3965,
+      "step": 725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953856317466785e-05,
+      "loss": 1.4082,
+      "step": 726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953600517871968e-05,
+      "loss": 1.4014,
+      "step": 727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953344012865984e-05,
+      "loss": 1.3506,
+      "step": 728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953086802467016e-05,
+      "loss": 1.3809,
+      "step": 729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995282888669329e-05,
+      "loss": 1.3955,
+      "step": 730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952570265563085e-05,
+      "loss": 1.4424,
+      "step": 731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995231093909473e-05,
+      "loss": 1.3965,
+      "step": 732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952050907306607e-05,
+      "loss": 1.4189,
+      "step": 733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995179017021714e-05,
+      "loss": 1.4531,
+      "step": 734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951528727844804e-05,
+      "loss": 1.3906,
+      "step": 735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951266580208133e-05,
+      "loss": 1.3701,
+      "step": 736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951003727325705e-05,
+      "loss": 1.416,
+      "step": 737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950740169216147e-05,
+      "loss": 1.3926,
+      "step": 738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950475905898135e-05,
+      "loss": 1.3828,
+      "step": 739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950210937390405e-05,
+      "loss": 1.3975,
+      "step": 740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949945263711728e-05,
+      "loss": 1.374,
+      "step": 741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949678884880934e-05,
+      "loss": 1.3779,
+      "step": 742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949411800916902e-05,
+      "loss": 1.4023,
+      "step": 743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994914401183856e-05,
+      "loss": 1.4385,
+      "step": 744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948875517664887e-05,
+      "loss": 1.3789,
+      "step": 745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948606318414908e-05,
+      "loss": 1.3936,
+      "step": 746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948336414107708e-05,
+      "loss": 1.3818,
+      "step": 747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948065804762406e-05,
+      "loss": 1.4062,
+      "step": 748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947794490398188e-05,
+      "loss": 1.4053,
+      "step": 749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994752247103428e-05,
+      "loss": 1.3506,
+      "step": 750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947249746689957e-05,
+      "loss": 1.3838,
+      "step": 751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946976317384545e-05,
+      "loss": 1.3994,
+      "step": 752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994670218313743e-05,
+      "loss": 1.4434,
+      "step": 753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946427343968035e-05,
+      "loss": 1.4307,
+      "step": 754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994615179989584e-05,
+      "loss": 1.4023,
+      "step": 755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994587555094037e-05,
+      "loss": 1.4111,
+      "step": 756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99455985971212e-05,
+      "loss": 1.3564,
+      "step": 757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945320938457965e-05,
+      "loss": 1.3965,
+      "step": 758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945042574970336e-05,
+      "loss": 1.4189,
+      "step": 759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944763506678048e-05,
+      "loss": 1.3838,
+      "step": 760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944483733600872e-05,
+      "loss": 1.4131,
+      "step": 761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994420325575864e-05,
+      "loss": 1.4473,
+      "step": 762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943922073171227e-05,
+      "loss": 1.4385,
+      "step": 763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994364018585856e-05,
+      "loss": 1.3887,
+      "step": 764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994335759384062e-05,
+      "loss": 1.3848,
+      "step": 765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943074297137424e-05,
+      "loss": 1.3828,
+      "step": 766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942790295769066e-05,
+      "loss": 1.3789,
+      "step": 767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942505589755656e-05,
+      "loss": 1.4072,
+      "step": 768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942220179117383e-05,
+      "loss": 1.3555,
+      "step": 769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994193406387447e-05,
+      "loss": 1.4258,
+      "step": 770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994164724404719e-05,
+      "loss": 1.3398,
+      "step": 771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941359719655878e-05,
+      "loss": 1.4297,
+      "step": 772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941071490720906e-05,
+      "loss": 1.376,
+      "step": 773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99407825572627e-05,
+      "loss": 1.3535,
+      "step": 774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994049291930174e-05,
+      "loss": 1.375,
+      "step": 775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994020257685855e-05,
+      "loss": 1.4209,
+      "step": 776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939911529953708e-05,
+      "loss": 1.4355,
+      "step": 777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939619778607837e-05,
+      "loss": 1.4092,
+      "step": 778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993932732284162e-05,
+      "loss": 1.4014,
+      "step": 779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939034162675777e-05,
+      "loss": 1.4502,
+      "step": 780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938740298131086e-05,
+      "loss": 1.4072,
+      "step": 781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938445729228376e-05,
+      "loss": 1.4111,
+      "step": 782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993815045598852e-05,
+      "loss": 1.4014,
+      "step": 783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937854478432447e-05,
+      "loss": 1.3916,
+      "step": 784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937557796581127e-05,
+      "loss": 1.4082,
+      "step": 785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937260410455592e-05,
+      "loss": 1.3721,
+      "step": 786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936962320076917e-05,
+      "loss": 1.4121,
+      "step": 787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936663525466224e-05,
+      "loss": 1.4102,
+      "step": 788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993636402664469e-05,
+      "loss": 1.3789,
+      "step": 789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936063823633543e-05,
+      "loss": 1.3447,
+      "step": 790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935762916454056e-05,
+      "loss": 1.4111,
+      "step": 791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935461305127555e-05,
+      "loss": 1.4121,
+      "step": 792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935158989675413e-05,
+      "loss": 1.3984,
+      "step": 793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993485597011906e-05,
+      "loss": 1.3438,
+      "step": 794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934552246479966e-05,
+      "loss": 1.4443,
+      "step": 795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993424781877966e-05,
+      "loss": 1.4434,
+      "step": 796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933942687039713e-05,
+      "loss": 1.4131,
+      "step": 797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933636851281753e-05,
+      "loss": 1.3945,
+      "step": 798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933330311527452e-05,
+      "loss": 1.4102,
+      "step": 799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933023067798533e-05,
+      "loss": 1.4072,
+      "step": 800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932715120116777e-05,
+      "loss": 1.417,
+      "step": 801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932406468504e-05,
+      "loss": 1.4043,
+      "step": 802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932097112982083e-05,
+      "loss": 1.3652,
+      "step": 803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931787053572946e-05,
+      "loss": 1.3906,
+      "step": 804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931476290298562e-05,
+      "loss": 1.4121,
+      "step": 805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931164823180958e-05,
+      "loss": 1.4072,
+      "step": 806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993085265224221e-05,
+      "loss": 1.3965,
+      "step": 807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930539777504436e-05,
+      "loss": 1.415,
+      "step": 808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993022619898981e-05,
+      "loss": 1.3721,
+      "step": 809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929911916720557e-05,
+      "loss": 1.4219,
+      "step": 810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992959693071895e-05,
+      "loss": 1.4072,
+      "step": 811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992928124100731e-05,
+      "loss": 1.4209,
+      "step": 812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928964847608018e-05,
+      "loss": 1.4023,
+      "step": 813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928647750543487e-05,
+      "loss": 1.3672,
+      "step": 814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992832994983619e-05,
+      "loss": 1.3867,
+      "step": 815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928011445508657e-05,
+      "loss": 1.3574,
+      "step": 816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927692237583455e-05,
+      "loss": 1.3916,
+      "step": 817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927372326083207e-05,
+      "loss": 1.3916,
+      "step": 818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927051711030586e-05,
+      "loss": 1.3525,
+      "step": 819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926730392448314e-05,
+      "loss": 1.4082,
+      "step": 820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926408370359164e-05,
+      "loss": 1.4004,
+      "step": 821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926085644785953e-05,
+      "loss": 1.415,
+      "step": 822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925762215751554e-05,
+      "loss": 1.3916,
+      "step": 823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992543808327889e-05,
+      "loss": 1.3965,
+      "step": 824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992511324739094e-05,
+      "loss": 1.3945,
+      "step": 825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924787708110708e-05,
+      "loss": 1.4336,
+      "step": 826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992446146546128e-05,
+      "loss": 1.4277,
+      "step": 827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924134519465768e-05,
+      "loss": 1.3906,
+      "step": 828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992380687014735e-05,
+      "loss": 1.4092,
+      "step": 829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923478517529236e-05,
+      "loss": 1.3936,
+      "step": 830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923149461634704e-05,
+      "loss": 1.3896,
+      "step": 831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922819702487077e-05,
+      "loss": 1.3613,
+      "step": 832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922489240109718e-05,
+      "loss": 1.4238,
+      "step": 833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922158074526047e-05,
+      "loss": 1.3877,
+      "step": 834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992182620575954e-05,
+      "loss": 1.4023,
+      "step": 835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992149363383371e-05,
+      "loss": 1.3691,
+      "step": 836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992116035877213e-05,
+      "loss": 1.3535,
+      "step": 837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992082638059842e-05,
+      "loss": 1.3623,
+      "step": 838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920491699336248e-05,
+      "loss": 1.4082,
+      "step": 839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920156315009333e-05,
+      "loss": 1.4072,
+      "step": 840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919820227641436e-05,
+      "loss": 1.3828,
+      "step": 841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991948343725639e-05,
+      "loss": 1.4287,
+      "step": 842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919145943878053e-05,
+      "loss": 1.4014,
+      "step": 843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918807747530346e-05,
+      "loss": 1.3643,
+      "step": 844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918468848237237e-05,
+      "loss": 1.3799,
+      "step": 845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991812924602274e-05,
+      "loss": 1.3867,
+      "step": 846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917788940910928e-05,
+      "loss": 1.4121,
+      "step": 847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991744793292592e-05,
+      "loss": 1.4922,
+      "step": 848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917106222091875e-05,
+      "loss": 1.3877,
+      "step": 849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916763808433016e-05,
+      "loss": 1.375,
+      "step": 850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991642069197361e-05,
+      "loss": 1.3662,
+      "step": 851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916076872737973e-05,
+      "loss": 1.377,
+      "step": 852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991573235075047e-05,
+      "loss": 1.3994,
+      "step": 853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915387126035518e-05,
+      "loss": 1.4277,
+      "step": 854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991504119861758e-05,
+      "loss": 1.4053,
+      "step": 855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991469456852118e-05,
+      "loss": 1.3721,
+      "step": 856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914347235770877e-05,
+      "loss": 1.417,
+      "step": 857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913999200391285e-05,
+      "loss": 1.3252,
+      "step": 858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913650462407075e-05,
+      "loss": 1.3877,
+      "step": 859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913301021842958e-05,
+      "loss": 1.3574,
+      "step": 860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99129508787237e-05,
+      "loss": 1.3906,
+      "step": 861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991260003307412e-05,
+      "loss": 1.4502,
+      "step": 862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991224848491907e-05,
+      "loss": 1.3975,
+      "step": 863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991189623428348e-05,
+      "loss": 1.3682,
+      "step": 864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911543281192304e-05,
+      "loss": 1.4062,
+      "step": 865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911189625670556e-05,
+      "loss": 1.3916,
+      "step": 866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910835267743305e-05,
+      "loss": 1.4189,
+      "step": 867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991048020743566e-05,
+      "loss": 1.4102,
+      "step": 868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910124444772782e-05,
+      "loss": 1.3496,
+      "step": 869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990976797977989e-05,
+      "loss": 1.4131,
+      "step": 870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909410812482243e-05,
+      "loss": 1.4619,
+      "step": 871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909052942905155e-05,
+      "loss": 1.4248,
+      "step": 872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908694371073985e-05,
+      "loss": 1.377,
+      "step": 873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990833509701415e-05,
+      "loss": 1.4023,
+      "step": 874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907975120751107e-05,
+      "loss": 1.3789,
+      "step": 875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990761444231037e-05,
+      "loss": 1.376,
+      "step": 876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99072530617175e-05,
+      "loss": 1.3887,
+      "step": 877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990689097899811e-05,
+      "loss": 1.4082,
+      "step": 878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906528194177858e-05,
+      "loss": 1.417,
+      "step": 879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906164707282453e-05,
+      "loss": 1.3936,
+      "step": 880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905800518337653e-05,
+      "loss": 1.4453,
+      "step": 881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990543562736928e-05,
+      "loss": 1.376,
+      "step": 882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905070034403183e-05,
+      "loss": 1.4082,
+      "step": 883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904703739465273e-05,
+      "loss": 1.3789,
+      "step": 884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904336742581517e-05,
+      "loss": 1.3545,
+      "step": 885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990396904377791e-05,
+      "loss": 1.4199,
+      "step": 886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903600643080526e-05,
+      "loss": 1.3965,
+      "step": 887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903231540515465e-05,
+      "loss": 1.4111,
+      "step": 888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902861736108886e-05,
+      "loss": 1.373,
+      "step": 889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902491229887e-05,
+      "loss": 1.3867,
+      "step": 890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902120021876057e-05,
+      "loss": 1.3965,
+      "step": 891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901748112102375e-05,
+      "loss": 1.3447,
+      "step": 892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99013755005923e-05,
+      "loss": 1.3936,
+      "step": 893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901002187372253e-05,
+      "loss": 1.4258,
+      "step": 894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990062817246868e-05,
+      "loss": 1.3477,
+      "step": 895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990025345590809e-05,
+      "loss": 1.4033,
+      "step": 896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899878037717038e-05,
+      "loss": 1.3857,
+      "step": 897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989950191792213e-05,
+      "loss": 1.3711,
+      "step": 898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989912509655003e-05,
+      "loss": 1.3691,
+      "step": 899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989874757362743e-05,
+      "loss": 1.3447,
+      "step": 900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898369349181094e-05,
+      "loss": 1.3525,
+      "step": 901
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897990423237824e-05,
+      "loss": 1.4531,
+      "step": 902
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897610795824474e-05,
+      "loss": 1.3369,
+      "step": 903
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897230466967948e-05,
+      "loss": 1.3721,
+      "step": 904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98968494366952e-05,
+      "loss": 1.3252,
+      "step": 905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896467705033235e-05,
+      "loss": 1.3584,
+      "step": 906
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896085272009106e-05,
+      "loss": 1.3857,
+      "step": 907
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895702137649914e-05,
+      "loss": 1.4121,
+      "step": 908
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895318301982816e-05,
+      "loss": 1.3857,
+      "step": 909
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989493376503501e-05,
+      "loss": 1.3789,
+      "step": 910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989454852683375e-05,
+      "loss": 1.3877,
+      "step": 911
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894162587406337e-05,
+      "loss": 1.3701,
+      "step": 912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989377594678012e-05,
+      "loss": 1.3955,
+      "step": 913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893388604982507e-05,
+      "loss": 1.4326,
+      "step": 914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893000562040944e-05,
+      "loss": 1.3906,
+      "step": 915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892611817982933e-05,
+      "loss": 1.4561,
+      "step": 916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892222372836025e-05,
+      "loss": 1.3799,
+      "step": 917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891832226627818e-05,
+      "loss": 1.4004,
+      "step": 918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989144137938596e-05,
+      "loss": 1.4365,
+      "step": 919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891049831138158e-05,
+      "loss": 1.3848,
+      "step": 920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890657581912155e-05,
+      "loss": 1.3984,
+      "step": 921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890264631735746e-05,
+      "loss": 1.4404,
+      "step": 922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988987098063679e-05,
+      "loss": 1.3721,
+      "step": 923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889476628643175e-05,
+      "loss": 1.3672,
+      "step": 924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988908157578285e-05,
+      "loss": 1.3477,
+      "step": 925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888685822083823e-05,
+      "loss": 1.4092,
+      "step": 926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988828936757413e-05,
+      "loss": 1.3984,
+      "step": 927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887892212281867e-05,
+      "loss": 1.3984,
+      "step": 928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887494356235186e-05,
+      "loss": 1.4111,
+      "step": 929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988709579946228e-05,
+      "loss": 1.4092,
+      "step": 930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886696541991402e-05,
+      "loss": 1.3525,
+      "step": 931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886296583850834e-05,
+      "loss": 1.3545,
+      "step": 932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885895925068932e-05,
+      "loss": 1.4023,
+      "step": 933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988549456567409e-05,
+      "loss": 1.3906,
+      "step": 934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885092505694745e-05,
+      "loss": 1.415,
+      "step": 935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884689745159393e-05,
+      "loss": 1.3926,
+      "step": 936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884286284096582e-05,
+      "loss": 1.4268,
+      "step": 937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883882122534905e-05,
+      "loss": 1.4043,
+      "step": 938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883477260503e-05,
+      "loss": 1.4141,
+      "step": 939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883071698029566e-05,
+      "loss": 1.3721,
+      "step": 940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988266543514334e-05,
+      "loss": 1.416,
+      "step": 941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882258471873116e-05,
+      "loss": 1.373,
+      "step": 942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988185080824774e-05,
+      "loss": 1.4072,
+      "step": 943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988144244429609e-05,
+      "loss": 1.4219,
+      "step": 944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881033380047117e-05,
+      "loss": 1.3857,
+      "step": 945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880623615529814e-05,
+      "loss": 1.3711,
+      "step": 946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880213150773212e-05,
+      "loss": 1.4092,
+      "step": 947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879801985806406e-05,
+      "loss": 1.3604,
+      "step": 948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879390120658533e-05,
+      "loss": 1.4365,
+      "step": 949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878977555358783e-05,
+      "loss": 1.3975,
+      "step": 950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878564289936394e-05,
+      "loss": 1.376,
+      "step": 951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878150324420655e-05,
+      "loss": 1.3867,
+      "step": 952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877735658840905e-05,
+      "loss": 1.3926,
+      "step": 953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877320293226528e-05,
+      "loss": 1.3545,
+      "step": 954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876904227606966e-05,
+      "loss": 1.3555,
+      "step": 955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876487462011698e-05,
+      "loss": 1.3926,
+      "step": 956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876069996470267e-05,
+      "loss": 1.3779,
+      "step": 957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875651831012254e-05,
+      "loss": 1.3877,
+      "step": 958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875232965667294e-05,
+      "loss": 1.377,
+      "step": 959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987481340046508e-05,
+      "loss": 1.3418,
+      "step": 960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987439313543534e-05,
+      "loss": 1.3682,
+      "step": 961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873972170607855e-05,
+      "loss": 1.4141,
+      "step": 962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873550506012464e-05,
+      "loss": 1.3643,
+      "step": 963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873128141679052e-05,
+      "loss": 1.3125,
+      "step": 964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987270507763755e-05,
+      "loss": 1.3604,
+      "step": 965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872281313917936e-05,
+      "loss": 1.4014,
+      "step": 966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871856850550246e-05,
+      "loss": 1.3877,
+      "step": 967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987143168756456e-05,
+      "loss": 1.3965,
+      "step": 968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871005824991016e-05,
+      "loss": 1.3896,
+      "step": 969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870579262859787e-05,
+      "loss": 1.3193,
+      "step": 970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870152001201103e-05,
+      "loss": 1.3555,
+      "step": 971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869724040045254e-05,
+      "loss": 1.3525,
+      "step": 972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869295379422557e-05,
+      "loss": 1.4062,
+      "step": 973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868866019363396e-05,
+      "loss": 1.3799,
+      "step": 974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868435959898204e-05,
+      "loss": 1.374,
+      "step": 975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868005201057453e-05,
+      "loss": 1.3467,
+      "step": 976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867573742871678e-05,
+      "loss": 1.3779,
+      "step": 977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986714158537145e-05,
+      "loss": 1.3877,
+      "step": 978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866708728587396e-05,
+      "loss": 1.3701,
+      "step": 979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866275172550194e-05,
+      "loss": 1.3936,
+      "step": 980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865840917290574e-05,
+      "loss": 1.4482,
+      "step": 981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865405962839307e-05,
+      "loss": 1.4092,
+      "step": 982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986497030922722e-05,
+      "loss": 1.3467,
+      "step": 983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986453395648519e-05,
+      "loss": 1.2998,
+      "step": 984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864096904644135e-05,
+      "loss": 1.3662,
+      "step": 985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863659153735032e-05,
+      "loss": 1.3721,
+      "step": 986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863220703788903e-05,
+      "loss": 1.3789,
+      "step": 987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862781554836827e-05,
+      "loss": 1.3418,
+      "step": 988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862341706909923e-05,
+      "loss": 1.3965,
+      "step": 989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861901160039357e-05,
+      "loss": 1.3516,
+      "step": 990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986145991425636e-05,
+      "loss": 1.3916,
+      "step": 991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98610179695922e-05,
+      "loss": 1.3926,
+      "step": 992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986057532607819e-05,
+      "loss": 1.3721,
+      "step": 993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860131983745713e-05,
+      "loss": 1.415,
+      "step": 994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985968794262618e-05,
+      "loss": 1.3721,
+      "step": 995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985924320275106e-05,
+      "loss": 1.4209,
+      "step": 996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858797764151877e-05,
+      "loss": 1.3848,
+      "step": 997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858351626860192e-05,
+      "loss": 1.377,
+      "step": 998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857904790907635e-05,
+      "loss": 1.4062,
+      "step": 999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857457256325858e-05,
+      "loss": 1.3535,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857009023146586e-05,
+      "loss": 1.3799,
+      "step": 1001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856560091401587e-05,
+      "loss": 1.4141,
+      "step": 1002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985611046112267e-05,
+      "loss": 1.3447,
+      "step": 1003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985566013234171e-05,
+      "loss": 1.3857,
+      "step": 1004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855209105090613e-05,
+      "loss": 1.3848,
+      "step": 1005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854757379401345e-05,
+      "loss": 1.4434,
+      "step": 1006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854304955305925e-05,
+      "loss": 1.3818,
+      "step": 1007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985385183283641e-05,
+      "loss": 1.3779,
+      "step": 1008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853398012024913e-05,
+      "loss": 1.3574,
+      "step": 1009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98529434929036e-05,
+      "loss": 1.3994,
+      "step": 1010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852488275504683e-05,
+      "loss": 1.4326,
+      "step": 1011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985203235986042e-05,
+      "loss": 1.374,
+      "step": 1012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851575746003123e-05,
+      "loss": 1.374,
+      "step": 1013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851118433965153e-05,
+      "loss": 1.3867,
+      "step": 1014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985066042377892e-05,
+      "loss": 1.3965,
+      "step": 1015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850201715476877e-05,
+      "loss": 1.3848,
+      "step": 1016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984974230909154e-05,
+      "loss": 1.3496,
+      "step": 1017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984928220465547e-05,
+      "loss": 1.417,
+      "step": 1018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848821402201265e-05,
+      "loss": 1.3936,
+      "step": 1019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848359901761588e-05,
+      "loss": 1.3994,
+      "step": 1020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984789770336914e-05,
+      "loss": 1.3447,
+      "step": 1021
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847434807056687e-05,
+      "loss": 1.3916,
+      "step": 1022
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846971212857024e-05,
+      "loss": 1.3926,
+      "step": 1023
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846506920803012e-05,
+      "loss": 1.3516,
+      "step": 1024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846041930927554e-05,
+      "loss": 1.373,
+      "step": 1025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845576243263603e-05,
+      "loss": 1.3809,
+      "step": 1026
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845109857844163e-05,
+      "loss": 1.4248,
+      "step": 1027
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844642774702288e-05,
+      "loss": 1.3584,
+      "step": 1028
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844174993871077e-05,
+      "loss": 1.377,
+      "step": 1029
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843706515383682e-05,
+      "loss": 1.3896,
+      "step": 1030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984323733927331e-05,
+      "loss": 1.3984,
+      "step": 1031
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98427674655732e-05,
+      "loss": 1.3838,
+      "step": 1032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842296894316665e-05,
+      "loss": 1.4102,
+      "step": 1033
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841825625537042e-05,
+      "loss": 1.3809,
+      "step": 1034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841353659267743e-05,
+      "loss": 1.3711,
+      "step": 1035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840880995542204e-05,
+      "loss": 1.373,
+      "step": 1036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840407634393928e-05,
+      "loss": 1.3691,
+      "step": 1037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839933575856464e-05,
+      "loss": 1.3955,
+      "step": 1038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839458819963404e-05,
+      "loss": 1.3281,
+      "step": 1039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838983366748398e-05,
+      "loss": 1.3564,
+      "step": 1040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983850721624514e-05,
+      "loss": 1.3232,
+      "step": 1041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983803036848737e-05,
+      "loss": 1.377,
+      "step": 1042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983755282350889e-05,
+      "loss": 1.3477,
+      "step": 1043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983707458134354e-05,
+      "loss": 1.4141,
+      "step": 1044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836595642025213e-05,
+      "loss": 1.3984,
+      "step": 1045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836116005587854e-05,
+      "loss": 1.4043,
+      "step": 1046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835635672065447e-05,
+      "loss": 1.3555,
+      "step": 1047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983515464149204e-05,
+      "loss": 1.3535,
+      "step": 1048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834672913901728e-05,
+      "loss": 1.3848,
+      "step": 1049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834190489328636e-05,
+      "loss": 1.4072,
+      "step": 1050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983370736780697e-05,
+      "loss": 1.3818,
+      "step": 1051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833223549370958e-05,
+      "loss": 1.3818,
+      "step": 1052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832739034054893e-05,
+      "loss": 1.4004,
+      "step": 1053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983225382189311e-05,
+      "loss": 1.3926,
+      "step": 1054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983176791292e-05,
+      "loss": 1.3799,
+      "step": 1055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831281307169992e-05,
+      "loss": 1.3486,
+      "step": 1056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830794004677577e-05,
+      "loss": 1.3477,
+      "step": 1057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983030600547729e-05,
+      "loss": 1.3682,
+      "step": 1058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829817309603716e-05,
+      "loss": 1.3857,
+      "step": 1059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829327917091487e-05,
+      "loss": 1.3398,
+      "step": 1060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828837827975285e-05,
+      "loss": 1.3945,
+      "step": 1061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828347042289848e-05,
+      "loss": 1.3242,
+      "step": 1062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827855560069952e-05,
+      "loss": 1.3584,
+      "step": 1063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982736338135043e-05,
+      "loss": 1.3301,
+      "step": 1064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826870506166164e-05,
+      "loss": 1.3467,
+      "step": 1065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826376934552086e-05,
+      "loss": 1.4258,
+      "step": 1066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982588266654317e-05,
+      "loss": 1.3896,
+      "step": 1067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982538770217445e-05,
+      "loss": 1.3721,
+      "step": 1068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824892041480998e-05,
+      "loss": 1.3828,
+      "step": 1069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824395684497946e-05,
+      "loss": 1.3193,
+      "step": 1070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823898631260472e-05,
+      "loss": 1.3818,
+      "step": 1071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98234008818038e-05,
+      "loss": 1.3955,
+      "step": 1072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98229024361632e-05,
+      "loss": 1.3994,
+      "step": 1073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982240329437401e-05,
+      "loss": 1.3848,
+      "step": 1074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982190345647159e-05,
+      "loss": 1.4014,
+      "step": 1075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821402922491374e-05,
+      "loss": 1.3799,
+      "step": 1076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982090169246883e-05,
+      "loss": 1.3701,
+      "step": 1077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820399766439478e-05,
+      "loss": 1.3662,
+      "step": 1078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98198971444389e-05,
+      "loss": 1.4229,
+      "step": 1079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819393826502705e-05,
+      "loss": 1.3818,
+      "step": 1080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818889812666565e-05,
+      "loss": 1.3594,
+      "step": 1081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818385102966205e-05,
+      "loss": 1.3633,
+      "step": 1082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817879697437386e-05,
+      "loss": 1.3701,
+      "step": 1083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817373596115934e-05,
+      "loss": 1.3359,
+      "step": 1084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816866799037713e-05,
+      "loss": 1.3975,
+      "step": 1085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981635930623864e-05,
+      "loss": 1.3486,
+      "step": 1086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981585111775468e-05,
+      "loss": 1.4023,
+      "step": 1087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981534223362185e-05,
+      "loss": 1.374,
+      "step": 1088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814832653876207e-05,
+      "loss": 1.3486,
+      "step": 1089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981432237855388e-05,
+      "loss": 1.3545,
+      "step": 1090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813811407691018e-05,
+      "loss": 1.374,
+      "step": 1091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813299741323845e-05,
+      "loss": 1.3818,
+      "step": 1092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812787379488608e-05,
+      "loss": 1.3857,
+      "step": 1093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812274322221635e-05,
+      "loss": 1.418,
+      "step": 1094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981176056955927e-05,
+      "loss": 1.334,
+      "step": 1095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981124612153794e-05,
+      "loss": 1.4189,
+      "step": 1096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981073097819409e-05,
+      "loss": 1.4541,
+      "step": 1097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810215139564236e-05,
+      "loss": 1.3438,
+      "step": 1098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809698605684928e-05,
+      "loss": 1.3701,
+      "step": 1099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809181376592778e-05,
+      "loss": 1.3711,
+      "step": 1100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808663452324445e-05,
+      "loss": 1.3564,
+      "step": 1101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808144832916626e-05,
+      "loss": 1.3936,
+      "step": 1102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807625518406083e-05,
+      "loss": 1.3271,
+      "step": 1103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807105508829614e-05,
+      "loss": 1.3398,
+      "step": 1104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806584804224076e-05,
+      "loss": 1.3643,
+      "step": 1105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980606340462637e-05,
+      "loss": 1.3916,
+      "step": 1106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805541310073447e-05,
+      "loss": 1.3682,
+      "step": 1107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980501852060231e-05,
+      "loss": 1.3516,
+      "step": 1108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804495036250008e-05,
+      "loss": 1.3721,
+      "step": 1109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980397085705364e-05,
+      "loss": 1.375,
+      "step": 1110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803445983050352e-05,
+      "loss": 1.373,
+      "step": 1111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980292041427734e-05,
+      "loss": 1.3926,
+      "step": 1112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802394150771864e-05,
+      "loss": 1.4355,
+      "step": 1113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980186719257121e-05,
+      "loss": 1.4023,
+      "step": 1114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980133953971272e-05,
+      "loss": 1.3955,
+      "step": 1115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800811192233797e-05,
+      "loss": 1.3994,
+      "step": 1116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800282150171882e-05,
+      "loss": 1.3838,
+      "step": 1117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799752413564465e-05,
+      "loss": 1.3564,
+      "step": 1118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799221982449096e-05,
+      "loss": 1.3691,
+      "step": 1119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798690856863358e-05,
+      "loss": 1.3438,
+      "step": 1120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798159036844897e-05,
+      "loss": 1.3486,
+      "step": 1121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797626522431402e-05,
+      "loss": 1.3398,
+      "step": 1122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979709331366061e-05,
+      "loss": 1.3516,
+      "step": 1123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796559410570314e-05,
+      "loss": 1.3623,
+      "step": 1124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979602481319835e-05,
+      "loss": 1.3467,
+      "step": 1125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97954895215826e-05,
+      "loss": 1.373,
+      "step": 1126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794953535761008e-05,
+      "loss": 1.3516,
+      "step": 1127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794416855771554e-05,
+      "loss": 1.3643,
+      "step": 1128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793879481652275e-05,
+      "loss": 1.3721,
+      "step": 1129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793341413441252e-05,
+      "loss": 1.3271,
+      "step": 1130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979280265117662e-05,
+      "loss": 1.3652,
+      "step": 1131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979226319489656e-05,
+      "loss": 1.3545,
+      "step": 1132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791723044639306e-05,
+      "loss": 1.3496,
+      "step": 1133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791182200443136e-05,
+      "loss": 1.3516,
+      "step": 1134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790640662346377e-05,
+      "loss": 1.3359,
+      "step": 1135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790098430387412e-05,
+      "loss": 1.3262,
+      "step": 1136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978955550460467e-05,
+      "loss": 1.3506,
+      "step": 1137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789011885036622e-05,
+      "loss": 1.3877,
+      "step": 1138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97884675717218e-05,
+      "loss": 1.3613,
+      "step": 1139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787922564698778e-05,
+      "loss": 1.4102,
+      "step": 1140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978737686400618e-05,
+      "loss": 1.4102,
+      "step": 1141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978683046968268e-05,
+      "loss": 1.3545,
+      "step": 1142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786283381766996e-05,
+      "loss": 1.3613,
+      "step": 1143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978573560029791e-05,
+      "loss": 1.3506,
+      "step": 1144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785187125314234e-05,
+      "loss": 1.376,
+      "step": 1145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978463795685485e-05,
+      "loss": 1.3701,
+      "step": 1146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978408809495866e-05,
+      "loss": 1.3877,
+      "step": 1147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978353753966465e-05,
+      "loss": 1.4023,
+      "step": 1148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978298629101183e-05,
+      "loss": 1.3301,
+      "step": 1149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782434349039264e-05,
+      "loss": 1.4072,
+      "step": 1150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978188171378607e-05,
+      "loss": 1.4043,
+      "step": 1151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978132838529142e-05,
+      "loss": 1.4043,
+      "step": 1152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780774363594525e-05,
+      "loss": 1.4004,
+      "step": 1153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780219648734643e-05,
+      "loss": 1.377,
+      "step": 1154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779664240751088e-05,
+      "loss": 1.3398,
+      "step": 1155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779108139683228e-05,
+      "loss": 1.3896,
+      "step": 1156
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977855134557047e-05,
+      "loss": 1.3984,
+      "step": 1157
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977799385845227e-05,
+      "loss": 1.3564,
+      "step": 1158
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777435678368146e-05,
+      "loss": 1.3506,
+      "step": 1159
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776876805357647e-05,
+      "loss": 1.3936,
+      "step": 1160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776317239460384e-05,
+      "loss": 1.3955,
+      "step": 1161
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775756980716018e-05,
+      "loss": 1.3584,
+      "step": 1162
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977519602916425e-05,
+      "loss": 1.3428,
+      "step": 1163
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774634384844834e-05,
+      "loss": 1.3418,
+      "step": 1164
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774072047797572e-05,
+      "loss": 1.3799,
+      "step": 1165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977350901806232e-05,
+      "loss": 1.376,
+      "step": 1166
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772945295678983e-05,
+      "loss": 1.3496,
+      "step": 1167
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772380880687503e-05,
+      "loss": 1.3418,
+      "step": 1168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977181577312789e-05,
+      "loss": 1.3105,
+      "step": 1169
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771249973040186e-05,
+      "loss": 1.3799,
+      "step": 1170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977068348046449e-05,
+      "loss": 1.3877,
+      "step": 1171
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770116295440958e-05,
+      "loss": 1.4043,
+      "step": 1172
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769548418009773e-05,
+      "loss": 1.3877,
+      "step": 1173
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768979848211193e-05,
+      "loss": 1.3604,
+      "step": 1174
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97684105860855e-05,
+      "loss": 1.4209,
+      "step": 1175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976784063167305e-05,
+      "loss": 1.4023,
+      "step": 1176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767269985014223e-05,
+      "loss": 1.3789,
+      "step": 1177
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976669864614947e-05,
+      "loss": 1.3965,
+      "step": 1178
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976612661511928e-05,
+      "loss": 1.3408,
+      "step": 1179
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976555389196419e-05,
+      "loss": 1.3086,
+      "step": 1180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764980476724793e-05,
+      "loss": 1.3896,
+      "step": 1181
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764406369441724e-05,
+      "loss": 1.3838,
+      "step": 1182
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976383157015567e-05,
+      "loss": 1.3965,
+      "step": 1183
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763256078907367e-05,
+      "loss": 1.3447,
+      "step": 1184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762679895737605e-05,
+      "loss": 1.3525,
+      "step": 1185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762103020687208e-05,
+      "loss": 1.3809,
+      "step": 1186
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761525453797067e-05,
+      "loss": 1.3838,
+      "step": 1187
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760947195108115e-05,
+      "loss": 1.4199,
+      "step": 1188
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760368244661327e-05,
+      "loss": 1.3672,
+      "step": 1189
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759788602497734e-05,
+      "loss": 1.3613,
+      "step": 1190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975920826865842e-05,
+      "loss": 1.4043,
+      "step": 1191
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758627243184512e-05,
+      "loss": 1.3633,
+      "step": 1192
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758045526117184e-05,
+      "loss": 1.3457,
+      "step": 1193
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757463117497667e-05,
+      "loss": 1.376,
+      "step": 1194
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975688001736723e-05,
+      "loss": 1.377,
+      "step": 1195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97562962257672e-05,
+      "loss": 1.3818,
+      "step": 1196
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755711742738958e-05,
+      "loss": 1.3438,
+      "step": 1197
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755126568323916e-05,
+      "loss": 1.3184,
+      "step": 1198
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754540702563545e-05,
+      "loss": 1.4102,
+      "step": 1199
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975395414549937e-05,
+      "loss": 1.3574,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753366897172964e-05,
+      "loss": 1.3721,
+      "step": 1201
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752778957625934e-05,
+      "loss": 1.3662,
+      "step": 1202
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975219032689996e-05,
+      "loss": 1.3369,
+      "step": 1203
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751601005036748e-05,
+      "loss": 1.3525,
+      "step": 1204
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751010992078065e-05,
+      "loss": 1.3887,
+      "step": 1205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975042028806573e-05,
+      "loss": 1.3594,
+      "step": 1206
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749828893041604e-05,
+      "loss": 1.3516,
+      "step": 1207
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97492368070476e-05,
+      "loss": 1.3926,
+      "step": 1208
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748644030125675e-05,
+      "loss": 1.3398,
+      "step": 1209
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748050562317843e-05,
+      "loss": 1.3818,
+      "step": 1210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974745640366616e-05,
+      "loss": 1.3779,
+      "step": 1211
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746861554212738e-05,
+      "loss": 1.3809,
+      "step": 1212
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746266013999735e-05,
+      "loss": 1.335,
+      "step": 1213
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745669783069347e-05,
+      "loss": 1.416,
+      "step": 1214
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745072861463843e-05,
+      "loss": 1.3496,
+      "step": 1215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974447524922551e-05,
+      "loss": 1.3682,
+      "step": 1216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743876946396715e-05,
+      "loss": 1.3643,
+      "step": 1217
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743277953019854e-05,
+      "loss": 1.3301,
+      "step": 1218
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974267826913738e-05,
+      "loss": 1.4043,
+      "step": 1219
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974207789479179e-05,
+      "loss": 1.3672,
+      "step": 1220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741476830025635e-05,
+      "loss": 1.3428,
+      "step": 1221
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740875074881508e-05,
+      "loss": 1.3789,
+      "step": 1222
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740272629402063e-05,
+      "loss": 1.3691,
+      "step": 1223
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739669493629984e-05,
+      "loss": 1.3525,
+      "step": 1224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739065667608025e-05,
+      "loss": 1.3506,
+      "step": 1225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738461151378978e-05,
+      "loss": 1.3564,
+      "step": 1226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973785594498568e-05,
+      "loss": 1.3682,
+      "step": 1227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737250048471025e-05,
+      "loss": 1.3936,
+      "step": 1228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736643461877953e-05,
+      "loss": 1.3984,
+      "step": 1229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736036185249454e-05,
+      "loss": 1.3623,
+      "step": 1230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735428218628562e-05,
+      "loss": 1.335,
+      "step": 1231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973481956205837e-05,
+      "loss": 1.3125,
+      "step": 1232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734210215582003e-05,
+      "loss": 1.3486,
+      "step": 1233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733600179242655e-05,
+      "loss": 1.3672,
+      "step": 1234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732989453083557e-05,
+      "loss": 1.4043,
+      "step": 1235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732378037147986e-05,
+      "loss": 1.3486,
+      "step": 1236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973176593147928e-05,
+      "loss": 1.3906,
+      "step": 1237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731153136120813e-05,
+      "loss": 1.3965,
+      "step": 1238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973053965111602e-05,
+      "loss": 1.4199,
+      "step": 1239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972992547650837e-05,
+      "loss": 1.3721,
+      "step": 1240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97293106123414e-05,
+      "loss": 1.4043,
+      "step": 1241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728695058658678e-05,
+      "loss": 1.332,
+      "step": 1242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972807881550383e-05,
+      "loss": 1.3662,
+      "step": 1243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972746188292053e-05,
+      "loss": 1.376,
+      "step": 1244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97268442609525e-05,
+      "loss": 1.374,
+      "step": 1245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972622594964351e-05,
+      "loss": 1.3809,
+      "step": 1246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972560694903738e-05,
+      "loss": 1.374,
+      "step": 1247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972498725917798e-05,
+      "loss": 1.4082,
+      "step": 1248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724366880109225e-05,
+      "loss": 1.3799,
+      "step": 1249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972374581187508e-05,
+      "loss": 1.332,
+      "step": 1250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723124054519567e-05,
+      "loss": 1.375,
+      "step": 1251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972250160808674e-05,
+      "loss": 1.4082,
+      "step": 1252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972187847262072e-05,
+      "loss": 1.4053,
+      "step": 1253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721254648165665e-05,
+      "loss": 1.3301,
+      "step": 1254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720630134765786e-05,
+      "loss": 1.3545,
+      "step": 1255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972000493246534e-05,
+      "loss": 1.3486,
+      "step": 1256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971937904130864e-05,
+      "loss": 1.3252,
+      "step": 1257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971875246134004e-05,
+      "loss": 1.3936,
+      "step": 1258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718125192603944e-05,
+      "loss": 1.4102,
+      "step": 1259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971749723514481e-05,
+      "loss": 1.3828,
+      "step": 1260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716868589007135e-05,
+      "loss": 1.334,
+      "step": 1261
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971623925423548e-05,
+      "loss": 1.3379,
+      "step": 1262
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971560923087444e-05,
+      "loss": 1.3838,
+      "step": 1263
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714978518968667e-05,
+      "loss": 1.374,
+      "step": 1264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971434711856286e-05,
+      "loss": 1.3428,
+      "step": 1265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713715029701764e-05,
+      "loss": 1.3682,
+      "step": 1266
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713082252430177e-05,
+      "loss": 1.3564,
+      "step": 1267
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712448786792942e-05,
+      "loss": 1.4092,
+      "step": 1268
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971181463283496e-05,
+      "loss": 1.3271,
+      "step": 1269
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711179790601157e-05,
+      "loss": 1.3779,
+      "step": 1270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971054426013654e-05,
+      "loss": 1.3135,
+      "step": 1271
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709908041486147e-05,
+      "loss": 1.3672,
+      "step": 1272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709271134695056e-05,
+      "loss": 1.3643,
+      "step": 1273
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970863353980842e-05,
+      "loss": 1.3643,
+      "step": 1274
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970799525687141e-05,
+      "loss": 1.3877,
+      "step": 1275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970735628592927e-05,
+      "loss": 1.375,
+      "step": 1276
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706716627027282e-05,
+      "loss": 1.3672,
+      "step": 1277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970607628021078e-05,
+      "loss": 1.3701,
+      "step": 1278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705435245525143e-05,
+      "loss": 1.4014,
+      "step": 1279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704793523015803e-05,
+      "loss": 1.3789,
+      "step": 1280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970415111272824e-05,
+      "loss": 1.3613,
+      "step": 1281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703508014707976e-05,
+      "loss": 1.2861,
+      "step": 1282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702864229000594e-05,
+      "loss": 1.4004,
+      "step": 1283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702219755651714e-05,
+      "loss": 1.334,
+      "step": 1284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701574594707013e-05,
+      "loss": 1.3711,
+      "step": 1285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970092874621221e-05,
+      "loss": 1.373,
+      "step": 1286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700282210213082e-05,
+      "loss": 1.3789,
+      "step": 1287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699634986755443e-05,
+      "loss": 1.3154,
+      "step": 1288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698987075885167e-05,
+      "loss": 1.2783,
+      "step": 1289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698338477648162e-05,
+      "loss": 1.375,
+      "step": 1290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969768919209041e-05,
+      "loss": 1.3057,
+      "step": 1291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969703921925791e-05,
+      "loss": 1.3955,
+      "step": 1292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696388559196732e-05,
+      "loss": 1.3994,
+      "step": 1293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695737211952993e-05,
+      "loss": 1.3398,
+      "step": 1294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695085177572847e-05,
+      "loss": 1.3789,
+      "step": 1295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694432456102502e-05,
+      "loss": 1.3369,
+      "step": 1296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693779047588226e-05,
+      "loss": 1.3896,
+      "step": 1297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969312495207632e-05,
+      "loss": 1.3828,
+      "step": 1298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692470169613136e-05,
+      "loss": 1.3633,
+      "step": 1299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691814700245084e-05,
+      "loss": 1.3457,
+      "step": 1300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691158544018617e-05,
+      "loss": 1.4307,
+      "step": 1301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690501700980235e-05,
+      "loss": 1.3594,
+      "step": 1302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689844171176488e-05,
+      "loss": 1.335,
+      "step": 1303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689185954653975e-05,
+      "loss": 1.3438,
+      "step": 1304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968852705145935e-05,
+      "loss": 1.3711,
+      "step": 1305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687867461639297e-05,
+      "loss": 1.3965,
+      "step": 1306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687207185240572e-05,
+      "loss": 1.3643,
+      "step": 1307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968654622230996e-05,
+      "loss": 1.3379,
+      "step": 1308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685884572894315e-05,
+      "loss": 1.3467,
+      "step": 1309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685222237040516e-05,
+      "loss": 1.3525,
+      "step": 1310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968455921479551e-05,
+      "loss": 1.3691,
+      "step": 1311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683895506206286e-05,
+      "loss": 1.377,
+      "step": 1312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683231111319875e-05,
+      "loss": 1.3867,
+      "step": 1313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968256603018337e-05,
+      "loss": 1.3428,
+      "step": 1314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681900262843897e-05,
+      "loss": 1.3838,
+      "step": 1315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681233809348645e-05,
+      "loss": 1.333,
+      "step": 1316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680566669744845e-05,
+      "loss": 1.3848,
+      "step": 1317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679898844079776e-05,
+      "loss": 1.3633,
+      "step": 1318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967923033240076e-05,
+      "loss": 1.3408,
+      "step": 1319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678561134755188e-05,
+      "loss": 1.3359,
+      "step": 1320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967789125119048e-05,
+      "loss": 1.3535,
+      "step": 1321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677220681754105e-05,
+      "loss": 1.3721,
+      "step": 1322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676549426493592e-05,
+      "loss": 1.3389,
+      "step": 1323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675877485456513e-05,
+      "loss": 1.334,
+      "step": 1324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675204858690487e-05,
+      "loss": 1.3027,
+      "step": 1325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674531546243183e-05,
+      "loss": 1.3242,
+      "step": 1326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673857548162316e-05,
+      "loss": 1.3369,
+      "step": 1327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673182864495657e-05,
+      "loss": 1.3359,
+      "step": 1328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967250749529102e-05,
+      "loss": 1.3389,
+      "step": 1329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671831440596268e-05,
+      "loss": 1.3828,
+      "step": 1330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967115470045931e-05,
+      "loss": 1.3936,
+      "step": 1331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967047727492811e-05,
+      "loss": 1.3447,
+      "step": 1332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669799164050675e-05,
+      "loss": 1.3857,
+      "step": 1333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966912036787506e-05,
+      "loss": 1.3398,
+      "step": 1334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966844088644938e-05,
+      "loss": 1.3506,
+      "step": 1335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667760719821784e-05,
+      "loss": 1.3633,
+      "step": 1336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667079868040475e-05,
+      "loss": 1.3311,
+      "step": 1337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666398331153704e-05,
+      "loss": 1.3848,
+      "step": 1338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665716109209775e-05,
+      "loss": 1.3369,
+      "step": 1339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665033202257034e-05,
+      "loss": 1.3457,
+      "step": 1340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664349610343883e-05,
+      "loss": 1.3438,
+      "step": 1341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663665333518763e-05,
+      "loss": 1.3506,
+      "step": 1342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966298037183017e-05,
+      "loss": 1.3662,
+      "step": 1343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966229472532665e-05,
+      "loss": 1.3438,
+      "step": 1344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661608394056786e-05,
+      "loss": 1.3867,
+      "step": 1345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660921378069233e-05,
+      "loss": 1.3525,
+      "step": 1346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966023367741267e-05,
+      "loss": 1.3887,
+      "step": 1347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659545292135833e-05,
+      "loss": 1.3379,
+      "step": 1348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658856222287516e-05,
+      "loss": 1.4023,
+      "step": 1349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658166467916543e-05,
+      "loss": 1.3887,
+      "step": 1350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657476029071804e-05,
+      "loss": 1.3799,
+      "step": 1351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965678490580223e-05,
+      "loss": 1.4082,
+      "step": 1352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656093098156798e-05,
+      "loss": 1.3379,
+      "step": 1353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655400606184534e-05,
+      "loss": 1.4111,
+      "step": 1354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654707429934526e-05,
+      "loss": 1.3301,
+      "step": 1355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654013569455887e-05,
+      "loss": 1.3965,
+      "step": 1356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653319024797793e-05,
+      "loss": 1.3477,
+      "step": 1357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652623796009477e-05,
+      "loss": 1.3633,
+      "step": 1358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651927883140194e-05,
+      "loss": 1.3467,
+      "step": 1359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651231286239278e-05,
+      "loss": 1.3115,
+      "step": 1360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650534005356085e-05,
+      "loss": 1.3467,
+      "step": 1361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649836040540042e-05,
+      "loss": 1.3262,
+      "step": 1362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96491373918406e-05,
+      "loss": 1.3623,
+      "step": 1363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648438059307286e-05,
+      "loss": 1.3838,
+      "step": 1364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647738042989655e-05,
+      "loss": 1.373,
+      "step": 1365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647037342937313e-05,
+      "loss": 1.3662,
+      "step": 1366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964633595919993e-05,
+      "loss": 1.3545,
+      "step": 1367
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645633891827205e-05,
+      "loss": 1.4023,
+      "step": 1368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644931140868894e-05,
+      "loss": 1.3184,
+      "step": 1369
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96442277063748e-05,
+      "loss": 1.4092,
+      "step": 1370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643523588394783e-05,
+      "loss": 1.375,
+      "step": 1371
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642818786978735e-05,
+      "loss": 1.3682,
+      "step": 1372
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642113302176607e-05,
+      "loss": 1.3604,
+      "step": 1373
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96414071340384e-05,
+      "loss": 1.3721,
+      "step": 1374
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964070028261416e-05,
+      "loss": 1.3457,
+      "step": 1375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639992747953975e-05,
+      "loss": 1.3486,
+      "step": 1376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639284530107997e-05,
+      "loss": 1.3691,
+      "step": 1377
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963857562912641e-05,
+      "loss": 1.3477,
+      "step": 1378
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637866045059457e-05,
+      "loss": 1.3398,
+      "step": 1379
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637155777957427e-05,
+      "loss": 1.3701,
+      "step": 1380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636444827870654e-05,
+      "loss": 1.3408,
+      "step": 1381
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635733194849527e-05,
+      "loss": 1.3838,
+      "step": 1382
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963502087894448e-05,
+      "loss": 1.3975,
+      "step": 1383
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634307880205985e-05,
+      "loss": 1.3174,
+      "step": 1384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963359419868458e-05,
+      "loss": 1.3799,
+      "step": 1385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632879834430848e-05,
+      "loss": 1.3281,
+      "step": 1386
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632164787495404e-05,
+      "loss": 1.3672,
+      "step": 1387
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631449057928933e-05,
+      "loss": 1.377,
+      "step": 1388
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630732645782157e-05,
+      "loss": 1.3252,
+      "step": 1389
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630015551105844e-05,
+      "loss": 1.3047,
+      "step": 1390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629297773950816e-05,
+      "loss": 1.332,
+      "step": 1391
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628579314367947e-05,
+      "loss": 1.3711,
+      "step": 1392
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627860172408145e-05,
+      "loss": 1.377,
+      "step": 1393
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627140348122386e-05,
+      "loss": 1.3252,
+      "step": 1394
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626419841561674e-05,
+      "loss": 1.3545,
+      "step": 1395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962569865277708e-05,
+      "loss": 1.3779,
+      "step": 1396
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624976781819708e-05,
+      "loss": 1.3301,
+      "step": 1397
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962425422874072e-05,
+      "loss": 1.3281,
+      "step": 1398
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962353099359132e-05,
+      "loss": 1.3799,
+      "step": 1399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622807076422772e-05,
+      "loss": 1.3516,
+      "step": 1400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962208247728637e-05,
+      "loss": 1.3662,
+      "step": 1401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621357196233472e-05,
+      "loss": 1.3604,
+      "step": 1402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620631233315477e-05,
+      "loss": 1.3652,
+      "step": 1403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619904588583834e-05,
+      "loss": 1.3447,
+      "step": 1404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961917726209004e-05,
+      "loss": 1.377,
+      "step": 1405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618449253885643e-05,
+      "loss": 1.3242,
+      "step": 1406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617720564022233e-05,
+      "loss": 1.3467,
+      "step": 1407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616991192551457e-05,
+      "loss": 1.3203,
+      "step": 1408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616261139525e-05,
+      "loss": 1.3643,
+      "step": 1409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96155304049946e-05,
+      "loss": 1.3477,
+      "step": 1410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614798989012056e-05,
+      "loss": 1.3135,
+      "step": 1411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961406689162919e-05,
+      "loss": 1.3408,
+      "step": 1412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961333411289789e-05,
+      "loss": 1.333,
+      "step": 1413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612600652870095e-05,
+      "loss": 1.3662,
+      "step": 1414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611866511597774e-05,
+      "loss": 1.3418,
+      "step": 1415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961113168913296e-05,
+      "loss": 1.3877,
+      "step": 1416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961039618552773e-05,
+      "loss": 1.3623,
+      "step": 1417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609660000834216e-05,
+      "loss": 1.3115,
+      "step": 1418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608923135104578e-05,
+      "loss": 1.2949,
+      "step": 1419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608185588391046e-05,
+      "loss": 1.3428,
+      "step": 1420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960744736074589e-05,
+      "loss": 1.3223,
+      "step": 1421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606708452221426e-05,
+      "loss": 1.3398,
+      "step": 1422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605968862870018e-05,
+      "loss": 1.2949,
+      "step": 1423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960522859274409e-05,
+      "loss": 1.3018,
+      "step": 1424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604487641896095e-05,
+      "loss": 1.3496,
+      "step": 1425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603746010378545e-05,
+      "loss": 1.4131,
+      "step": 1426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603003698244007e-05,
+      "loss": 1.4004,
+      "step": 1427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602260705545083e-05,
+      "loss": 1.3936,
+      "step": 1428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601517032334427e-05,
+      "loss": 1.335,
+      "step": 1429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600772678664746e-05,
+      "loss": 1.3828,
+      "step": 1430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600027644588792e-05,
+      "loss": 1.3496,
+      "step": 1431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959928193015937e-05,
+      "loss": 1.3838,
+      "step": 1432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959853553542932e-05,
+      "loss": 1.3457,
+      "step": 1433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959778846045154e-05,
+      "loss": 1.3496,
+      "step": 1434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597040705278983e-05,
+      "loss": 1.3926,
+      "step": 1435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959629226996464e-05,
+      "loss": 1.3711,
+      "step": 1436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595543154561544e-05,
+      "loss": 1.3545,
+      "step": 1437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594793359122798e-05,
+      "loss": 1.3359,
+      "step": 1438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959404288370153e-05,
+      "loss": 1.3779,
+      "step": 1439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959329172835093e-05,
+      "loss": 1.3574,
+      "step": 1440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592539893124227e-05,
+      "loss": 1.3691,
+      "step": 1441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959178737807471e-05,
+      "loss": 1.374,
+      "step": 1442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959103418325571e-05,
+      "loss": 1.3271,
+      "step": 1443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590280308720602e-05,
+      "loss": 1.3408,
+      "step": 1444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589525754522815e-05,
+      "loss": 1.3545,
+      "step": 1445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588770520715824e-05,
+      "loss": 1.3604,
+      "step": 1446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588014607353155e-05,
+      "loss": 1.3506,
+      "step": 1447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587258014488376e-05,
+      "loss": 1.3896,
+      "step": 1448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586500742175106e-05,
+      "loss": 1.334,
+      "step": 1449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585742790467012e-05,
+      "loss": 1.3623,
+      "step": 1450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584984159417817e-05,
+      "loss": 1.377,
+      "step": 1451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584224849081276e-05,
+      "loss": 1.3564,
+      "step": 1452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958346485951121e-05,
+      "loss": 1.3545,
+      "step": 1453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958270419076147e-05,
+      "loss": 1.3721,
+      "step": 1454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581942842885973e-05,
+      "loss": 1.333,
+      "step": 1455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581180815938676e-05,
+      "loss": 1.3262,
+      "step": 1456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580418109973576e-05,
+      "loss": 1.3594,
+      "step": 1457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579654725044732e-05,
+      "loss": 1.3154,
+      "step": 1458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578890661206244e-05,
+      "loss": 1.3799,
+      "step": 1459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578125918512255e-05,
+      "loss": 1.3701,
+      "step": 1460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577360497016972e-05,
+      "loss": 1.3486,
+      "step": 1461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957659439677464e-05,
+      "loss": 1.3311,
+      "step": 1462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575827617839544e-05,
+      "loss": 1.3633,
+      "step": 1463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957506016026603e-05,
+      "loss": 1.3906,
+      "step": 1464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574292024108487e-05,
+      "loss": 1.3672,
+      "step": 1465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573523209421355e-05,
+      "loss": 1.332,
+      "step": 1466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957275371625912e-05,
+      "loss": 1.335,
+      "step": 1467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571983544676312e-05,
+      "loss": 1.3545,
+      "step": 1468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957121269472752e-05,
+      "loss": 1.4082,
+      "step": 1469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570441166467362e-05,
+      "loss": 1.3408,
+      "step": 1470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956966895995053e-05,
+      "loss": 1.3525,
+      "step": 1471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956889607523174e-05,
+      "loss": 1.3799,
+      "step": 1472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956812251236577e-05,
+      "loss": 1.3213,
+      "step": 1473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567348271407445e-05,
+      "loss": 1.3662,
+      "step": 1474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956657335241163e-05,
+      "loss": 1.3096,
+      "step": 1475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565797755433253e-05,
+      "loss": 1.3369,
+      "step": 1476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956502148052727e-05,
+      "loss": 1.332,
+      "step": 1477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564244527748698e-05,
+      "loss": 1.334,
+      "step": 1478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563466897152605e-05,
+      "loss": 1.3623,
+      "step": 1479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562688588794098e-05,
+      "loss": 1.3564,
+      "step": 1480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956190960272833e-05,
+      "loss": 1.376,
+      "step": 1481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561129939010518e-05,
+      "loss": 1.374,
+      "step": 1482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560349597695908e-05,
+      "loss": 1.3535,
+      "step": 1483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955956857883981e-05,
+      "loss": 1.3555,
+      "step": 1484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955878688249757e-05,
+      "loss": 1.3369,
+      "step": 1485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558004508724592e-05,
+      "loss": 1.3545,
+      "step": 1486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557221457576313e-05,
+      "loss": 1.3984,
+      "step": 1487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556437729108237e-05,
+      "loss": 1.3096,
+      "step": 1488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555653323375903e-05,
+      "loss": 1.3008,
+      "step": 1489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95548682404349e-05,
+      "loss": 1.374,
+      "step": 1490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554082480340875e-05,
+      "loss": 1.3672,
+      "step": 1491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553296043149505e-05,
+      "loss": 1.374,
+      "step": 1492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955250892891653e-05,
+      "loss": 1.3291,
+      "step": 1493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955172113769773e-05,
+      "loss": 1.3477,
+      "step": 1494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550932669548937e-05,
+      "loss": 1.3643,
+      "step": 1495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550143524526032e-05,
+      "loss": 1.3516,
+      "step": 1496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549353702684937e-05,
+      "loss": 1.3359,
+      "step": 1497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954856320408163e-05,
+      "loss": 1.3652,
+      "step": 1498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547772028772133e-05,
+      "loss": 1.332,
+      "step": 1499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546980176812517e-05,
+      "loss": 1.3096,
+      "step": 1500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546187648258903e-05,
+      "loss": 1.3564,
+      "step": 1501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954539444316745e-05,
+      "loss": 1.333,
+      "step": 1502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544600561594376e-05,
+      "loss": 1.3418,
+      "step": 1503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543806003595947e-05,
+      "loss": 1.3613,
+      "step": 1504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954301076922847e-05,
+      "loss": 1.3701,
+      "step": 1505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95422148585483e-05,
+      "loss": 1.3809,
+      "step": 1506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541418271611853e-05,
+      "loss": 1.3545,
+      "step": 1507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540621008475572e-05,
+      "loss": 1.3496,
+      "step": 1508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539823069195963e-05,
+      "loss": 1.3896,
+      "step": 1509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953902445382958e-05,
+      "loss": 1.3633,
+      "step": 1510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953822516243302e-05,
+      "loss": 1.3711,
+      "step": 1511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953742519506292e-05,
+      "loss": 1.3496,
+      "step": 1512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536624551775986e-05,
+      "loss": 1.3281,
+      "step": 1513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953582323262895e-05,
+      "loss": 1.3477,
+      "step": 1514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535021237678603e-05,
+      "loss": 1.3945,
+      "step": 1515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534218566981784e-05,
+      "loss": 1.3525,
+      "step": 1516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533415220595382e-05,
+      "loss": 1.3594,
+      "step": 1517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532611198576324e-05,
+      "loss": 1.3516,
+      "step": 1518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531806500981595e-05,
+      "loss": 1.3799,
+      "step": 1519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531001127868222e-05,
+      "loss": 1.3584,
+      "step": 1520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530195079293283e-05,
+      "loss": 1.3516,
+      "step": 1521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529388355313903e-05,
+      "loss": 1.3438,
+      "step": 1522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952858095598725e-05,
+      "loss": 1.3857,
+      "step": 1523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952777288137055e-05,
+      "loss": 1.3545,
+      "step": 1524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952696413152107e-05,
+      "loss": 1.3359,
+      "step": 1525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526154706496124e-05,
+      "loss": 1.3555,
+      "step": 1526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525344606353077e-05,
+      "loss": 1.334,
+      "step": 1527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952453383114934e-05,
+      "loss": 1.3486,
+      "step": 1528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523722380942373e-05,
+      "loss": 1.3418,
+      "step": 1529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522910255789686e-05,
+      "loss": 1.3193,
+      "step": 1530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522097455748828e-05,
+      "loss": 1.3809,
+      "step": 1531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521283980877406e-05,
+      "loss": 1.3887,
+      "step": 1532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520469831233075e-05,
+      "loss": 1.3369,
+      "step": 1533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951965500687353e-05,
+      "loss": 1.3838,
+      "step": 1534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518839507856513e-05,
+      "loss": 1.3535,
+      "step": 1535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518023334239822e-05,
+      "loss": 1.3506,
+      "step": 1536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517206486081305e-05,
+      "loss": 1.3057,
+      "step": 1537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516388963438844e-05,
+      "loss": 1.3066,
+      "step": 1538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951557076637038e-05,
+      "loss": 1.3535,
+      "step": 1539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514751894933898e-05,
+      "loss": 1.3516,
+      "step": 1540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513932349187432e-05,
+      "loss": 1.3506,
+      "step": 1541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513112129189063e-05,
+      "loss": 1.3047,
+      "step": 1542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512291234996916e-05,
+      "loss": 1.3184,
+      "step": 1543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511469666669177e-05,
+      "loss": 1.3301,
+      "step": 1544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951064742426406e-05,
+      "loss": 1.3096,
+      "step": 1545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509824507839847e-05,
+      "loss": 1.4141,
+      "step": 1546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950900091745485e-05,
+      "loss": 1.3584,
+      "step": 1547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950817665316744e-05,
+      "loss": 1.2939,
+      "step": 1548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507351715036034e-05,
+      "loss": 1.3145,
+      "step": 1549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506526103119094e-05,
+      "loss": 1.3193,
+      "step": 1550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950569981747513e-05,
+      "loss": 1.4043,
+      "step": 1551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504872858162703e-05,
+      "loss": 1.3301,
+      "step": 1552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504045225240414e-05,
+      "loss": 1.377,
+      "step": 1553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503216918766928e-05,
+      "loss": 1.3203,
+      "step": 1554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502387938800936e-05,
+      "loss": 1.3916,
+      "step": 1555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501558285401197e-05,
+      "loss": 1.3232,
+      "step": 1556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95007279586265e-05,
+      "loss": 1.3516,
+      "step": 1557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499896958535695e-05,
+      "loss": 1.3594,
+      "step": 1558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499065285187674e-05,
+      "loss": 1.3223,
+      "step": 1559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949823293864138e-05,
+      "loss": 1.3916,
+      "step": 1560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497399918955797e-05,
+      "loss": 1.3408,
+      "step": 1561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496566226189967e-05,
+      "loss": 1.334,
+      "step": 1562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495731860402968e-05,
+      "loss": 1.3447,
+      "step": 1563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949489682165393e-05,
+      "loss": 1.3174,
+      "step": 1564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494061110002038e-05,
+      "loss": 1.373,
+      "step": 1565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493224725506515e-05,
+      "loss": 1.3662,
+      "step": 1566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949238766822664e-05,
+      "loss": 1.3125,
+      "step": 1567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949154993822173e-05,
+      "loss": 1.3623,
+      "step": 1568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949071153555116e-05,
+      "loss": 1.2969,
+      "step": 1569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948987246027434e-05,
+      "loss": 1.3516,
+      "step": 1570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489032712450745e-05,
+      "loss": 1.3604,
+      "step": 1571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488192292139877e-05,
+      "loss": 1.3496,
+      "step": 1572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487351199401305e-05,
+      "loss": 1.3887,
+      "step": 1573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486509434294634e-05,
+      "loss": 1.3682,
+      "step": 1574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485666996879522e-05,
+      "loss": 1.3643,
+      "step": 1575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948482388721567e-05,
+      "loss": 1.3271,
+      "step": 1576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483980105362832e-05,
+      "loss": 1.3154,
+      "step": 1577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483135651380804e-05,
+      "loss": 1.3184,
+      "step": 1578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948229052532943e-05,
+      "loss": 1.3467,
+      "step": 1579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481444727268607e-05,
+      "loss": 1.3477,
+      "step": 1580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948059825725828e-05,
+      "loss": 1.376,
+      "step": 1581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479751115358436e-05,
+      "loss": 1.3574,
+      "step": 1582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478903301629106e-05,
+      "loss": 1.3574,
+      "step": 1583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478054816130383e-05,
+      "loss": 1.3154,
+      "step": 1584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477205658922394e-05,
+      "loss": 1.3418,
+      "step": 1585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947635583006532e-05,
+      "loss": 1.3457,
+      "step": 1586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475505329619392e-05,
+      "loss": 1.3418,
+      "step": 1587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474654157644882e-05,
+      "loss": 1.3193,
+      "step": 1588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473802314202107e-05,
+      "loss": 1.3643,
+      "step": 1589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472949799351445e-05,
+      "loss": 1.3545,
+      "step": 1590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472096613153312e-05,
+      "loss": 1.3799,
+      "step": 1591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947124275566817e-05,
+      "loss": 1.3809,
+      "step": 1592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470388226956535e-05,
+      "loss": 1.3408,
+      "step": 1593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946953302707897e-05,
+      "loss": 1.3555,
+      "step": 1594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468677156096076e-05,
+      "loss": 1.3467,
+      "step": 1595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946782061406851e-05,
+      "loss": 1.3594,
+      "step": 1596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466963401056982e-05,
+      "loss": 1.334,
+      "step": 1597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466105517122234e-05,
+      "loss": 1.2959,
+      "step": 1598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465246962325072e-05,
+      "loss": 1.377,
+      "step": 1599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464387736726334e-05,
+      "loss": 1.3242,
+      "step": 1600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463527840386915e-05,
+      "loss": 1.3584,
+      "step": 1601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946266727336776e-05,
+      "loss": 1.3623,
+      "step": 1602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461806035729856e-05,
+      "loss": 1.3369,
+      "step": 1603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946094412753424e-05,
+      "loss": 1.2939,
+      "step": 1604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946008154884199e-05,
+      "loss": 1.3193,
+      "step": 1605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945921829971424e-05,
+      "loss": 1.333,
+      "step": 1606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945835438021217e-05,
+      "loss": 1.3691,
+      "step": 1607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457489790397002e-05,
+      "loss": 1.3535,
+      "step": 1608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456624530330013e-05,
+      "loss": 1.375,
+      "step": 1609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455758600072525e-05,
+      "loss": 1.3262,
+      "step": 1610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454891999685904e-05,
+      "loss": 1.3477,
+      "step": 1611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454024729231565e-05,
+      "loss": 1.3867,
+      "step": 1612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453156788770974e-05,
+      "loss": 1.3525,
+      "step": 1613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945228817836564e-05,
+      "loss": 1.332,
+      "step": 1614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451418898077122e-05,
+      "loss": 1.3398,
+      "step": 1615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450548947967024e-05,
+      "loss": 1.374,
+      "step": 1616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449678328097007e-05,
+      "loss": 1.3457,
+      "step": 1617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448807038528762e-05,
+      "loss": 1.3271,
+      "step": 1618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447935079324042e-05,
+      "loss": 1.3896,
+      "step": 1619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944706245054464e-05,
+      "loss": 1.3398,
+      "step": 1620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94461891522524e-05,
+      "loss": 1.3652,
+      "step": 1621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445315184509213e-05,
+      "loss": 1.3223,
+      "step": 1622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444440547377023e-05,
+      "loss": 1.334,
+      "step": 1623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944356524091781e-05,
+      "loss": 1.3262,
+      "step": 1624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94426892651936e-05,
+      "loss": 1.3379,
+      "step": 1625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944181262026649e-05,
+      "loss": 1.3623,
+      "step": 1626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440935306198593e-05,
+      "loss": 1.3066,
+      "step": 1627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440057323052087e-05,
+      "loss": 1.3311,
+      "step": 1628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439178670889202e-05,
+      "loss": 1.3262,
+      "step": 1629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94382993497722e-05,
+      "loss": 1.3057,
+      "step": 1630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437419359763404e-05,
+      "loss": 1.3779,
+      "step": 1631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436538700925175e-05,
+      "loss": 1.3066,
+      "step": 1632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435657373319926e-05,
+      "loss": 1.3252,
+      "step": 1633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943477537701012e-05,
+      "loss": 1.3486,
+      "step": 1634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943389271205826e-05,
+      "loss": 1.3486,
+      "step": 1635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433009378526898e-05,
+      "loss": 1.3164,
+      "step": 1636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432125376478645e-05,
+      "loss": 1.2998,
+      "step": 1637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943124070597614e-05,
+      "loss": 1.3574,
+      "step": 1638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430355367082087e-05,
+      "loss": 1.3223,
+      "step": 1639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429469359859226e-05,
+      "loss": 1.3848,
+      "step": 1640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428582684370353e-05,
+      "loss": 1.3213,
+      "step": 1641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94276953406783e-05,
+      "loss": 1.3096,
+      "step": 1642
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426807328845957e-05,
+      "loss": 1.3643,
+      "step": 1643
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425918648936257e-05,
+      "loss": 1.3379,
+      "step": 1644
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942502930101218e-05,
+      "loss": 1.3213,
+      "step": 1645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942413928513675e-05,
+      "loss": 1.3389,
+      "step": 1646
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423248601373052e-05,
+      "loss": 1.3311,
+      "step": 1647
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422357249784203e-05,
+      "loss": 1.3096,
+      "step": 1648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421465230433373e-05,
+      "loss": 1.3174,
+      "step": 1649
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420572543383777e-05,
+      "loss": 1.376,
+      "step": 1650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419679188698682e-05,
+      "loss": 1.3193,
+      "step": 1651
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418785166441405e-05,
+      "loss": 1.3516,
+      "step": 1652
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417890476675298e-05,
+      "loss": 1.2715,
+      "step": 1653
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416995119463772e-05,
+      "loss": 1.3555,
+      "step": 1654
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416099094870278e-05,
+      "loss": 1.3408,
+      "step": 1655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941520240295832e-05,
+      "loss": 1.333,
+      "step": 1656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414305043791443e-05,
+      "loss": 1.3945,
+      "step": 1657
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413407017433245e-05,
+      "loss": 1.3271,
+      "step": 1658
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941250832394737e-05,
+      "loss": 1.3789,
+      "step": 1659
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411608963397504e-05,
+      "loss": 1.3369,
+      "step": 1660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941070893584739e-05,
+      "loss": 1.3721,
+      "step": 1661
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409808241360815e-05,
+      "loss": 1.332,
+      "step": 1662
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94089068800016e-05,
+      "loss": 1.3496,
+      "step": 1663
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408004851833637e-05,
+      "loss": 1.3369,
+      "step": 1664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407102156920844e-05,
+      "loss": 1.3203,
+      "step": 1665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406198795327202e-05,
+      "loss": 1.3398,
+      "step": 1666
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405294767116723e-05,
+      "loss": 1.3574,
+      "step": 1667
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940439007235348e-05,
+      "loss": 1.2734,
+      "step": 1668
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403484711101592e-05,
+      "loss": 1.4004,
+      "step": 1669
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940257868342522e-05,
+      "loss": 1.3652,
+      "step": 1670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401671989388572e-05,
+      "loss": 1.3262,
+      "step": 1671
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400764629055907e-05,
+      "loss": 1.3555,
+      "step": 1672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939985660249153e-05,
+      "loss": 1.3203,
+      "step": 1673
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398947909759793e-05,
+      "loss": 1.3271,
+      "step": 1674
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398038550925093e-05,
+      "loss": 1.3535,
+      "step": 1675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397128526051876e-05,
+      "loss": 1.3604,
+      "step": 1676
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939621783520464e-05,
+      "loss": 1.2725,
+      "step": 1677
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939530647844792e-05,
+      "loss": 1.376,
+      "step": 1678
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394394455846307e-05,
+      "loss": 1.2861,
+      "step": 1679
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393481767464432e-05,
+      "loss": 1.3604,
+      "step": 1680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392568413366986e-05,
+      "loss": 1.2998,
+      "step": 1681
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391654393618688e-05,
+      "loss": 1.3271,
+      "step": 1682
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939073970828432e-05,
+      "loss": 1.3916,
+      "step": 1683
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389824357428708e-05,
+      "loss": 1.3115,
+      "step": 1684
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388908341116717e-05,
+      "loss": 1.3438,
+      "step": 1685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938799165941327e-05,
+      "loss": 1.3418,
+      "step": 1686
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387074312383327e-05,
+      "loss": 1.3252,
+      "step": 1687
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938615630009191e-05,
+      "loss": 1.3457,
+      "step": 1688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385237622604063e-05,
+      "loss": 1.3594,
+      "step": 1689
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384318279984904e-05,
+      "loss": 1.3447,
+      "step": 1690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383398272299584e-05,
+      "loss": 1.3936,
+      "step": 1691
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938247759961331e-05,
+      "loss": 1.3506,
+      "step": 1692
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938155626199132e-05,
+      "loss": 1.3438,
+      "step": 1693
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380634259498913e-05,
+      "loss": 1.3926,
+      "step": 1694
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379711592201432e-05,
+      "loss": 1.3555,
+      "step": 1695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378788260164264e-05,
+      "loss": 1.3008,
+      "step": 1696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377864263452852e-05,
+      "loss": 1.3936,
+      "step": 1697
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937693960213267e-05,
+      "loss": 1.3057,
+      "step": 1698
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376014276269256e-05,
+      "loss": 1.334,
+      "step": 1699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375088285928186e-05,
+      "loss": 1.3105,
+      "step": 1700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374161631175083e-05,
+      "loss": 1.3428,
+      "step": 1701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937323431207562e-05,
+      "loss": 1.3486,
+      "step": 1702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937230632869552e-05,
+      "loss": 1.3389,
+      "step": 1703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371377681100544e-05,
+      "loss": 1.3369,
+      "step": 1704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370448369356504e-05,
+      "loss": 1.3574,
+      "step": 1705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936951839352927e-05,
+      "loss": 1.2891,
+      "step": 1706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368587753684737e-05,
+      "loss": 1.3027,
+      "step": 1707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367656449888862e-05,
+      "loss": 1.3682,
+      "step": 1708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366724482207655e-05,
+      "loss": 1.3545,
+      "step": 1709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365791850707158e-05,
+      "loss": 1.3564,
+      "step": 1710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364858555453466e-05,
+      "loss": 1.3369,
+      "step": 1711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363924596512723e-05,
+      "loss": 1.3047,
+      "step": 1712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362989973951118e-05,
+      "loss": 1.3506,
+      "step": 1713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362054687834887e-05,
+      "loss": 1.3359,
+      "step": 1714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361118738230318e-05,
+      "loss": 1.3115,
+      "step": 1715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936018212520374e-05,
+      "loss": 1.3584,
+      "step": 1716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359244848821525e-05,
+      "loss": 1.3535,
+      "step": 1717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358306909150105e-05,
+      "loss": 1.3027,
+      "step": 1718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357368306255947e-05,
+      "loss": 1.3535,
+      "step": 1719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356429040205573e-05,
+      "loss": 1.334,
+      "step": 1720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355489111065546e-05,
+      "loss": 1.2979,
+      "step": 1721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354548518902477e-05,
+      "loss": 1.3555,
+      "step": 1722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353607263783034e-05,
+      "loss": 1.3506,
+      "step": 1723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935266534577392e-05,
+      "loss": 1.3643,
+      "step": 1724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351722764941882e-05,
+      "loss": 1.3555,
+      "step": 1725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350779521353726e-05,
+      "loss": 1.3398,
+      "step": 1726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93498356150763e-05,
+      "loss": 1.3037,
+      "step": 1727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348891046176498e-05,
+      "loss": 1.375,
+      "step": 1728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347945814721264e-05,
+      "loss": 1.3145,
+      "step": 1729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934699992077758e-05,
+      "loss": 1.3564,
+      "step": 1730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346053364412486e-05,
+      "loss": 1.3896,
+      "step": 1731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345106145693067e-05,
+      "loss": 1.3574,
+      "step": 1732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344158264686448e-05,
+      "loss": 1.3623,
+      "step": 1733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343209721459807e-05,
+      "loss": 1.3213,
+      "step": 1734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342260516080363e-05,
+      "loss": 1.3506,
+      "step": 1735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341310648615396e-05,
+      "loss": 1.3193,
+      "step": 1736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340360119132212e-05,
+      "loss": 1.334,
+      "step": 1737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339408927698184e-05,
+      "loss": 1.3193,
+      "step": 1738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338457074380715e-05,
+      "loss": 1.3604,
+      "step": 1739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933750455924727e-05,
+      "loss": 1.291,
+      "step": 1740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336551382365346e-05,
+      "loss": 1.3535,
+      "step": 1741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93355975438025e-05,
+      "loss": 1.3516,
+      "step": 1742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933464304362633e-05,
+      "loss": 1.3662,
+      "step": 1743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333687881904483e-05,
+      "loss": 1.3398,
+      "step": 1744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332732058704645e-05,
+      "loss": 1.3662,
+      "step": 1745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331775574094563e-05,
+      "loss": 1.3291,
+      "step": 1746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330818428142016e-05,
+      "loss": 1.3672,
+      "step": 1747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329860620914842e-05,
+      "loss": 1.3613,
+      "step": 1748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932890215248092e-05,
+      "loss": 1.3096,
+      "step": 1749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327943022908168e-05,
+      "loss": 1.3398,
+      "step": 1750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326983232264576e-05,
+      "loss": 1.3428,
+      "step": 1751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932602278061815e-05,
+      "loss": 1.3281,
+      "step": 1752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325061668036965e-05,
+      "loss": 1.3584,
+      "step": 1753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932409989458913e-05,
+      "loss": 1.3369,
+      "step": 1754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932313746034281e-05,
+      "loss": 1.2998,
+      "step": 1755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932217436536621e-05,
+      "loss": 1.3262,
+      "step": 1756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321210609727585e-05,
+      "loss": 1.3643,
+      "step": 1757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320246193495235e-05,
+      "loss": 1.3486,
+      "step": 1758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319281116737513e-05,
+      "loss": 1.3271,
+      "step": 1759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931831537952281e-05,
+      "loss": 1.4053,
+      "step": 1760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931734898191957e-05,
+      "loss": 1.3779,
+      "step": 1761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316381923996273e-05,
+      "loss": 1.3467,
+      "step": 1762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315414205821467e-05,
+      "loss": 1.3154,
+      "step": 1763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931444582746373e-05,
+      "loss": 1.3008,
+      "step": 1764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313476788991687e-05,
+      "loss": 1.335,
+      "step": 1765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312507090474016e-05,
+      "loss": 1.3477,
+      "step": 1766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931153673197944e-05,
+      "loss": 1.332,
+      "step": 1767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931056571357673e-05,
+      "loss": 1.3037,
+      "step": 1768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309594035334697e-05,
+      "loss": 1.3262,
+      "step": 1769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308621697322206e-05,
+      "loss": 1.3154,
+      "step": 1770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307648699608166e-05,
+      "loss": 1.3545,
+      "step": 1771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930667504226154e-05,
+      "loss": 1.332,
+      "step": 1772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305700725351322e-05,
+      "loss": 1.3584,
+      "step": 1773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304725748946566e-05,
+      "loss": 1.4229,
+      "step": 1774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303750113116367e-05,
+      "loss": 1.3359,
+      "step": 1775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302773817929868e-05,
+      "loss": 1.375,
+      "step": 1776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301796863456267e-05,
+      "loss": 1.2988,
+      "step": 1777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300819249764787e-05,
+      "loss": 1.3213,
+      "step": 1778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299840976924722e-05,
+      "loss": 1.3643,
+      "step": 1779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298862045005397e-05,
+      "loss": 1.3203,
+      "step": 1780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297882454076187e-05,
+      "loss": 1.3525,
+      "step": 1781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296902204206525e-05,
+      "loss": 1.3193,
+      "step": 1782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295921295465868e-05,
+      "loss": 1.3359,
+      "step": 1783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294939727923747e-05,
+      "loss": 1.3203,
+      "step": 1784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293957501649718e-05,
+      "loss": 1.3398,
+      "step": 1785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292974616713388e-05,
+      "loss": 1.3662,
+      "step": 1786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929199107318442e-05,
+      "loss": 1.3184,
+      "step": 1787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291006871132514e-05,
+      "loss": 1.332,
+      "step": 1788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290022010627424e-05,
+      "loss": 1.3516,
+      "step": 1789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289036491738943e-05,
+      "loss": 1.3115,
+      "step": 1790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928805031453692e-05,
+      "loss": 1.3633,
+      "step": 1791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928706347909124e-05,
+      "loss": 1.3584,
+      "step": 1792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286075985471843e-05,
+      "loss": 1.3369,
+      "step": 1793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928508783374871e-05,
+      "loss": 1.3721,
+      "step": 1794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284099023991874e-05,
+      "loss": 1.3633,
+      "step": 1795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283109556271412e-05,
+      "loss": 1.3311,
+      "step": 1796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282119430657444e-05,
+      "loss": 1.4014,
+      "step": 1797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281128647220142e-05,
+      "loss": 1.3672,
+      "step": 1798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280137206029726e-05,
+      "loss": 1.3398,
+      "step": 1799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279145107156457e-05,
+      "loss": 1.3584,
+      "step": 1800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278152350670644e-05,
+      "loss": 1.3076,
+      "step": 1801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927715893664264e-05,
+      "loss": 1.2998,
+      "step": 1802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927616486514286e-05,
+      "loss": 1.3906,
+      "step": 1803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275170136241742e-05,
+      "loss": 1.3457,
+      "step": 1804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274174750009787e-05,
+      "loss": 1.3018,
+      "step": 1805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273178706517537e-05,
+      "loss": 1.3467,
+      "step": 1806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927218200583558e-05,
+      "loss": 1.3174,
+      "step": 1807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927118464803456e-05,
+      "loss": 1.375,
+      "step": 1808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927018663318515e-05,
+      "loss": 1.3662,
+      "step": 1809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269187961358085e-05,
+      "loss": 1.29,
+      "step": 1810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268188632624138e-05,
+      "loss": 1.3027,
+      "step": 1811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267188647054135e-05,
+      "loss": 1.3291,
+      "step": 1812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266188004718938e-05,
+      "loss": 1.3516,
+      "step": 1813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926518670568947e-05,
+      "loss": 1.3398,
+      "step": 1814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926418475003669e-05,
+      "loss": 1.3066,
+      "step": 1815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263182137831603e-05,
+      "loss": 1.3389,
+      "step": 1816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926217886914527e-05,
+      "loss": 1.334,
+      "step": 1817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926117494404879e-05,
+      "loss": 1.3652,
+      "step": 1818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260170362613314e-05,
+      "loss": 1.3623,
+      "step": 1819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259165124910032e-05,
+      "loss": 1.373,
+      "step": 1820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258159231010184e-05,
+      "loss": 1.3418,
+      "step": 1821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257152680985065e-05,
+      "loss": 1.3301,
+      "step": 1822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256145474906e-05,
+      "loss": 1.4014,
+      "step": 1823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925513761284438e-05,
+      "loss": 1.333,
+      "step": 1824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254129094871622e-05,
+      "loss": 1.3105,
+      "step": 1825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253119921059204e-05,
+      "loss": 1.3066,
+      "step": 1826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925211009147865e-05,
+      "loss": 1.3672,
+      "step": 1827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925109960620152e-05,
+      "loss": 1.373,
+      "step": 1828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925008846529943e-05,
+      "loss": 1.3438,
+      "step": 1829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249076668844042e-05,
+      "loss": 1.3506,
+      "step": 1830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924806421690706e-05,
+      "loss": 1.3486,
+      "step": 1831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247051109560232e-05,
+      "loss": 1.3311,
+      "step": 1832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924603734687536e-05,
+      "loss": 1.3623,
+      "step": 1833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245022928924292e-05,
+      "loss": 1.3467,
+      "step": 1834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244007855778918e-05,
+      "loss": 1.2783,
+      "step": 1835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242992127511178e-05,
+      "loss": 1.2979,
+      "step": 1836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241975744193052e-05,
+      "loss": 1.3633,
+      "step": 1837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240958705896573e-05,
+      "loss": 1.3379,
+      "step": 1838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923994101269382e-05,
+      "loss": 1.3184,
+      "step": 1839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238922664656914e-05,
+      "loss": 1.3027,
+      "step": 1840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237903661858032e-05,
+      "loss": 1.3203,
+      "step": 1841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236884004369377e-05,
+      "loss": 1.3486,
+      "step": 1842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235863692263228e-05,
+      "loss": 1.2646,
+      "step": 1843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234842725611888e-05,
+      "loss": 1.2959,
+      "step": 1844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923382110448771e-05,
+      "loss": 1.3496,
+      "step": 1845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232798828963096e-05,
+      "loss": 1.3262,
+      "step": 1846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92317758991105e-05,
+      "loss": 1.3428,
+      "step": 1847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230752315002413e-05,
+      "loss": 1.3252,
+      "step": 1848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922972807671138e-05,
+      "loss": 1.335,
+      "step": 1849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922870318430998e-05,
+      "loss": 1.3408,
+      "step": 1850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227677637870862e-05,
+      "loss": 1.3154,
+      "step": 1851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922665143746669e-05,
+      "loss": 1.3389,
+      "step": 1852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92256245831702e-05,
+      "loss": 1.3223,
+      "step": 1853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224597075054167e-05,
+      "loss": 1.3486,
+      "step": 1854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223568913191403e-05,
+      "loss": 1.3408,
+      "step": 1855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922254009765478e-05,
+      "loss": 1.418,
+      "step": 1856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221510628517206e-05,
+      "loss": 1.2979,
+      "step": 1857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220480505851646e-05,
+      "loss": 1.3545,
+      "step": 1858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219449729731093e-05,
+      "loss": 1.3369,
+      "step": 1859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921841830022861e-05,
+      "loss": 1.3408,
+      "step": 1860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217386217417286e-05,
+      "loss": 1.3633,
+      "step": 1861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921635348137027e-05,
+      "loss": 1.3281,
+      "step": 1862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215320092160753e-05,
+      "loss": 1.3135,
+      "step": 1863
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921428604986196e-05,
+      "loss": 1.3037,
+      "step": 1864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213251354547192e-05,
+      "loss": 1.3584,
+      "step": 1865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921221600628976e-05,
+      "loss": 1.3301,
+      "step": 1866
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211180005163047e-05,
+      "loss": 1.3594,
+      "step": 1867
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921014335124048e-05,
+      "loss": 1.3516,
+      "step": 1868
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209106044595514e-05,
+      "loss": 1.3389,
+      "step": 1869
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208068085301672e-05,
+      "loss": 1.3008,
+      "step": 1870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207029473432512e-05,
+      "loss": 1.3252,
+      "step": 1871
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920599020906164e-05,
+      "loss": 1.3428,
+      "step": 1872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204950292262708e-05,
+      "loss": 1.2988,
+      "step": 1873
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203909723109414e-05,
+      "loss": 1.3174,
+      "step": 1874
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920286850167551e-05,
+      "loss": 1.3242,
+      "step": 1875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201826628034775e-05,
+      "loss": 1.3633,
+      "step": 1876
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200784102261058e-05,
+      "loss": 1.3379,
+      "step": 1877
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199740924428236e-05,
+      "loss": 1.3213,
+      "step": 1878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919869709461024e-05,
+      "loss": 1.3574,
+      "step": 1879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919765261288105e-05,
+      "loss": 1.3438,
+      "step": 1880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196607479314685e-05,
+      "loss": 1.3408,
+      "step": 1881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195561693985217e-05,
+      "loss": 1.3203,
+      "step": 1882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194515256966752e-05,
+      "loss": 1.332,
+      "step": 1883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919346816833346e-05,
+      "loss": 1.335,
+      "step": 1884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192420428159543e-05,
+      "loss": 1.3535,
+      "step": 1885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191372036519257e-05,
+      "loss": 1.3232,
+      "step": 1886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190322993486904e-05,
+      "loss": 1.3516,
+      "step": 1887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189273299136823e-05,
+      "loss": 1.3301,
+      "step": 1888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918822295354341e-05,
+      "loss": 1.3232,
+      "step": 1889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187171956781104e-05,
+      "loss": 1.333,
+      "step": 1890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186120308924382e-05,
+      "loss": 1.3652,
+      "step": 1891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185068010047785e-05,
+      "loss": 1.2861,
+      "step": 1892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918401506022588e-05,
+      "loss": 1.3486,
+      "step": 1893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182961459533293e-05,
+      "loss": 1.3535,
+      "step": 1894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181907208044698e-05,
+      "loss": 1.3477,
+      "step": 1895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91808523058348e-05,
+      "loss": 1.3613,
+      "step": 1896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179796752978366e-05,
+      "loss": 1.3623,
+      "step": 1897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91787405495502e-05,
+      "loss": 1.3359,
+      "step": 1898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917768369562516e-05,
+      "loss": 1.3291,
+      "step": 1899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917662619127814e-05,
+      "loss": 1.3096,
+      "step": 1900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175568036584084e-05,
+      "loss": 1.3125,
+      "step": 1901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917450923161799e-05,
+      "loss": 1.3281,
+      "step": 1902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917344977645489e-05,
+      "loss": 1.3389,
+      "step": 1903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172389671169872e-05,
+      "loss": 1.3877,
+      "step": 1904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917132891583806e-05,
+      "loss": 1.2861,
+      "step": 1905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170267510534636e-05,
+      "loss": 1.3486,
+      "step": 1906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169205455334815e-05,
+      "loss": 1.3662,
+      "step": 1907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916814275031387e-05,
+      "loss": 1.3848,
+      "step": 1908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916707939554711e-05,
+      "loss": 1.3584,
+      "step": 1909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91660153911099e-05,
+      "loss": 1.3633,
+      "step": 1910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916495073707765e-05,
+      "loss": 1.333,
+      "step": 1911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163885433525798e-05,
+      "loss": 1.3564,
+      "step": 1912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162819480529852e-05,
+      "loss": 1.3008,
+      "step": 1913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161752878165354e-05,
+      "loss": 1.2695,
+      "step": 1914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160685626507892e-05,
+      "loss": 1.3525,
+      "step": 1915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159617725633107e-05,
+      "loss": 1.3066,
+      "step": 1916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915854917561668e-05,
+      "loss": 1.3604,
+      "step": 1917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157479976534334e-05,
+      "loss": 1.3281,
+      "step": 1918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156410128461846e-05,
+      "loss": 1.3516,
+      "step": 1919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915533963147504e-05,
+      "loss": 1.2842,
+      "step": 1920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154268485649776e-05,
+      "loss": 1.3076,
+      "step": 1921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153196691061965e-05,
+      "loss": 1.3242,
+      "step": 1922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152124247787574e-05,
+      "loss": 1.3184,
+      "step": 1923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151051155902597e-05,
+      "loss": 1.333,
+      "step": 1924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149977415483093e-05,
+      "loss": 1.3252,
+      "step": 1925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148903026605153e-05,
+      "loss": 1.3174,
+      "step": 1926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147827989344915e-05,
+      "loss": 1.3174,
+      "step": 1927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146752303778575e-05,
+      "loss": 1.3291,
+      "step": 1928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914567596998236e-05,
+      "loss": 1.3555,
+      "step": 1929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144598988032553e-05,
+      "loss": 1.3477,
+      "step": 1930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143521358005483e-05,
+      "loss": 1.3242,
+      "step": 1931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142443079977514e-05,
+      "loss": 1.2969,
+      "step": 1932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141364154025072e-05,
+      "loss": 1.3125,
+      "step": 1933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914028458022461e-05,
+      "loss": 1.3467,
+      "step": 1934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139204358652648e-05,
+      "loss": 1.3525,
+      "step": 1935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138123489385735e-05,
+      "loss": 1.376,
+      "step": 1936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913704197250047e-05,
+      "loss": 1.3545,
+      "step": 1937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913595980807351e-05,
+      "loss": 1.3496,
+      "step": 1938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134876996181537e-05,
+      "loss": 1.2969,
+      "step": 1939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133793536901294e-05,
+      "loss": 1.3271,
+      "step": 1940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132709430309568e-05,
+      "loss": 1.3525,
+      "step": 1941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131624676483186e-05,
+      "loss": 1.3555,
+      "step": 1942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913053927549903e-05,
+      "loss": 1.3066,
+      "step": 1943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129453227434017e-05,
+      "loss": 1.3691,
+      "step": 1944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912836653236511e-05,
+      "loss": 1.3271,
+      "step": 1945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912727919036934e-05,
+      "loss": 1.2959,
+      "step": 1946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126191201523753e-05,
+      "loss": 1.3564,
+      "step": 1947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125102565905453e-05,
+      "loss": 1.377,
+      "step": 1948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124013283591602e-05,
+      "loss": 1.2998,
+      "step": 1949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912292335465939e-05,
+      "loss": 1.334,
+      "step": 1950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121832779186064e-05,
+      "loss": 1.2949,
+      "step": 1951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912074155724891e-05,
+      "loss": 1.3418,
+      "step": 1952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119649688925262e-05,
+      "loss": 1.2803,
+      "step": 1953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911855717429251e-05,
+      "loss": 1.292,
+      "step": 1954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117464013428067e-05,
+      "loss": 1.3223,
+      "step": 1955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116370206409414e-05,
+      "loss": 1.2617,
+      "step": 1956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115275753314066e-05,
+      "loss": 1.3594,
+      "step": 1957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114180654219584e-05,
+      "loss": 1.2959,
+      "step": 1958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113084909203588e-05,
+      "loss": 1.3242,
+      "step": 1959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111988518343723e-05,
+      "loss": 1.3672,
+      "step": 1960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110891481717692e-05,
+      "loss": 1.3369,
+      "step": 1961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109793799403243e-05,
+      "loss": 1.3096,
+      "step": 1962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108695471478175e-05,
+      "loss": 1.3105,
+      "step": 1963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107596498020315e-05,
+      "loss": 1.335,
+      "step": 1964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106496879107557e-05,
+      "loss": 1.2969,
+      "step": 1965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105396614817825e-05,
+      "loss": 1.3633,
+      "step": 1966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104295705229097e-05,
+      "loss": 1.3408,
+      "step": 1967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910319415041939e-05,
+      "loss": 1.3145,
+      "step": 1968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910209195046678e-05,
+      "loss": 1.3447,
+      "step": 1969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100989105449373e-05,
+      "loss": 1.332,
+      "step": 1970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909988561544533e-05,
+      "loss": 1.293,
+      "step": 1971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098781480532855e-05,
+      "loss": 1.3115,
+      "step": 1972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097676700790198e-05,
+      "loss": 1.3516,
+      "step": 1973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096571276295656e-05,
+      "loss": 1.3184,
+      "step": 1974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095465207127563e-05,
+      "loss": 1.3613,
+      "step": 1975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094358493364318e-05,
+      "loss": 1.3945,
+      "step": 1976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093251135084342e-05,
+      "loss": 1.3047,
+      "step": 1977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092143132366124e-05,
+      "loss": 1.3535,
+      "step": 1978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909103448528818e-05,
+      "loss": 1.3242,
+      "step": 1979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089925193929084e-05,
+      "loss": 1.374,
+      "step": 1980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908881525836745e-05,
+      "loss": 1.3555,
+      "step": 1981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908770467868194e-05,
+      "loss": 1.3291,
+      "step": 1982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908659345495126e-05,
+      "loss": 1.4297,
+      "step": 1983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085481587254163e-05,
+      "loss": 1.3252,
+      "step": 1984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908436907566945e-05,
+      "loss": 1.3477,
+      "step": 1985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083255920275955e-05,
+      "loss": 1.3271,
+      "step": 1986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082142121152576e-05,
+      "loss": 1.3271,
+      "step": 1987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081027678378245e-05,
+      "loss": 1.334,
+      "step": 1988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079912592031943e-05,
+      "loss": 1.373,
+      "step": 1989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9078796862192696e-05,
+      "loss": 1.3594,
+      "step": 1990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077680488939575e-05,
+      "loss": 1.3564,
+      "step": 1991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076563472351696e-05,
+      "loss": 1.332,
+      "step": 1992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075445812508227e-05,
+      "loss": 1.3447,
+      "step": 1993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907432750948837e-05,
+      "loss": 1.3652,
+      "step": 1994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073208563371385e-05,
+      "loss": 1.3604,
+      "step": 1995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907208897423657e-05,
+      "loss": 1.3555,
+      "step": 1996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907096874216327e-05,
+      "loss": 1.333,
+      "step": 1997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906984786723087e-05,
+      "loss": 1.3359,
+      "step": 1998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906872634951882e-05,
+      "loss": 1.3525,
+      "step": 1999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067604189106585e-05,
+      "loss": 1.3438,
+      "step": 2000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066481386073703e-05,
+      "loss": 1.3574,
+      "step": 2001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065357940499748e-05,
+      "loss": 1.292,
+      "step": 2002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064233852464332e-05,
+      "loss": 1.3271,
+      "step": 2003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063109122047122e-05,
+      "loss": 1.2822,
+      "step": 2004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906198374932783e-05,
+      "loss": 1.3408,
+      "step": 2005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906085773438621e-05,
+      "loss": 1.3662,
+      "step": 2006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.905973107730206e-05,
+      "loss": 1.3379,
+      "step": 2007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058603778155225e-05,
+      "loss": 1.3408,
+      "step": 2008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057475837025605e-05,
+      "loss": 1.3418,
+      "step": 2009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056347253993127e-05,
+      "loss": 1.2734,
+      "step": 2010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055218029137784e-05,
+      "loss": 1.3594,
+      "step": 2011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054088162539593e-05,
+      "loss": 1.3271,
+      "step": 2012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052957654278638e-05,
+      "loss": 1.3154,
+      "step": 2013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051826504435026e-05,
+      "loss": 1.4014,
+      "step": 2014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050694713088934e-05,
+      "loss": 1.3564,
+      "step": 2015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049562280320564e-05,
+      "loss": 1.3389,
+      "step": 2016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048429206210172e-05,
+      "loss": 1.3457,
+      "step": 2017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047295490838066e-05,
+      "loss": 1.3232,
+      "step": 2018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046161134284586e-05,
+      "loss": 1.29,
+      "step": 2019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045026136630124e-05,
+      "loss": 1.3447,
+      "step": 2020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043890497955117e-05,
+      "loss": 1.3359,
+      "step": 2021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904275421834005e-05,
+      "loss": 1.3301,
+      "step": 2022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904161729786545e-05,
+      "loss": 1.3574,
+      "step": 2023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904047973661189e-05,
+      "loss": 1.3164,
+      "step": 2024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039341534659988e-05,
+      "loss": 1.3223,
+      "step": 2025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903820269209041e-05,
+      "loss": 1.3301,
+      "step": 2026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903706320898387e-05,
+      "loss": 1.3096,
+      "step": 2027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035923085421113e-05,
+      "loss": 1.3477,
+      "step": 2028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034782321482946e-05,
+      "loss": 1.3389,
+      "step": 2029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033640917250216e-05,
+      "loss": 1.3438,
+      "step": 2030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032498872803812e-05,
+      "loss": 1.3232,
+      "step": 2031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903135618822467e-05,
+      "loss": 1.3438,
+      "step": 2032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030212863593775e-05,
+      "loss": 1.3164,
+      "step": 2033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902906889899215e-05,
+      "loss": 1.2861,
+      "step": 2034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902792429450087e-05,
+      "loss": 1.3721,
+      "step": 2035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026779050201056e-05,
+      "loss": 1.3125,
+      "step": 2036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902563316617386e-05,
+      "loss": 1.3008,
+      "step": 2037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024486642500507e-05,
+      "loss": 1.3242,
+      "step": 2038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902333947926224e-05,
+      "loss": 1.3369,
+      "step": 2039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902219167654036e-05,
+      "loss": 1.293,
+      "step": 2040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021043234416216e-05,
+      "loss": 1.332,
+      "step": 2041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019894152971193e-05,
+      "loss": 1.3428,
+      "step": 2042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018744432286732e-05,
+      "loss": 1.3467,
+      "step": 2043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017594072444308e-05,
+      "loss": 1.3564,
+      "step": 2044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901644307352545e-05,
+      "loss": 1.3125,
+      "step": 2045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015291435611727e-05,
+      "loss": 1.3008,
+      "step": 2046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901413915878476e-05,
+      "loss": 1.3545,
+      "step": 2047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012986243126205e-05,
+      "loss": 1.3076,
+      "step": 2048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011832688717774e-05,
+      "loss": 1.3281,
+      "step": 2049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010678495641218e-05,
+      "loss": 1.3379,
+      "step": 2050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900952366397833e-05,
+      "loss": 1.3379,
+      "step": 2051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900836819381096e-05,
+      "loss": 1.3174,
+      "step": 2052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900721208522099e-05,
+      "loss": 1.3057,
+      "step": 2053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006055338290357e-05,
+      "loss": 1.333,
+      "step": 2054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900489795310104e-05,
+      "loss": 1.3398,
+      "step": 2055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003739929735065e-05,
+      "loss": 1.293,
+      "step": 2056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002581268274492e-05,
+      "loss": 1.334,
+      "step": 2057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900142196880144e-05,
+      "loss": 1.3408,
+      "step": 2058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000262031398072e-05,
+      "loss": 1.3281,
+      "step": 2059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899910145614659e-05,
+      "loss": 1.3203,
+      "step": 2060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997940243129246e-05,
+      "loss": 1.3398,
+      "step": 2061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899677839242833e-05,
+      "loss": 1.3223,
+      "step": 2062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899561590412619e-05,
+      "loss": 1.3398,
+      "step": 2063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89944527783052e-05,
+      "loss": 1.293,
+      "step": 2064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89932890150478e-05,
+      "loss": 1.3594,
+      "step": 2065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992124614436466e-05,
+      "loss": 1.3066,
+      "step": 2066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990959576553717e-05,
+      "loss": 1.3662,
+      "step": 2067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989793901482116e-05,
+      "loss": 1.3408,
+      "step": 2068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988627589304274e-05,
+      "loss": 1.3574,
+      "step": 2069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987460640102854e-05,
+      "loss": 1.3428,
+      "step": 2070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986293053960555e-05,
+      "loss": 1.3398,
+      "step": 2071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985124830960122e-05,
+      "loss": 1.2969,
+      "step": 2072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983955971184345e-05,
+      "loss": 1.332,
+      "step": 2073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982786474716067e-05,
+      "loss": 1.3467,
+      "step": 2074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981616341638168e-05,
+      "loss": 1.3525,
+      "step": 2075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980445572033572e-05,
+      "loss": 1.3652,
+      "step": 2076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979274165985253e-05,
+      "loss": 1.3271,
+      "step": 2077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897810212357623e-05,
+      "loss": 1.2666,
+      "step": 2078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897692944488957e-05,
+      "loss": 1.3594,
+      "step": 2079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975756130008368e-05,
+      "loss": 1.3271,
+      "step": 2080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974582179015786e-05,
+      "loss": 1.3311,
+      "step": 2081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897340759199502e-05,
+      "loss": 1.3027,
+      "step": 2082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897223236902931e-05,
+      "loss": 1.3203,
+      "step": 2083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897105651020195e-05,
+      "loss": 1.3096,
+      "step": 2084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896988001559627e-05,
+      "loss": 1.3066,
+      "step": 2085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896870288529565e-05,
+      "loss": 1.3281,
+      "step": 2086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967525119383506e-05,
+      "loss": 1.3018,
+      "step": 2087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8966346717943314e-05,
+      "loss": 1.2959,
+      "step": 2088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896516768105859e-05,
+      "loss": 1.3525,
+      "step": 2089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896398800881288e-05,
+      "loss": 1.3154,
+      "step": 2090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962807701289795e-05,
+      "loss": 1.2939,
+      "step": 2091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961626758572983e-05,
+      "loss": 1.2852,
+      "step": 2092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896044518074614e-05,
+      "loss": 1.3301,
+      "step": 2093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959262967892998e-05,
+      "loss": 1.2881,
+      "step": 2094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958080120097345e-05,
+      "loss": 1.3291,
+      "step": 2095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956896637443007e-05,
+      "loss": 1.3555,
+      "step": 2096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895571252001386e-05,
+      "loss": 1.3623,
+      "step": 2097
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895452776789382e-05,
+      "loss": 1.2988,
+      "step": 2098
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953342381166847e-05,
+      "loss": 1.2949,
+      "step": 2099
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895215635991696e-05,
+      "loss": 1.3525,
+      "step": 2100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89509697042282e-05,
+      "loss": 1.2949,
+      "step": 2101
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949782414184678e-05,
+      "loss": 1.3457,
+      "step": 2102
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948594489870525e-05,
+      "loss": 1.3369,
+      "step": 2103
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947405931369935e-05,
+      "loss": 1.3184,
+      "step": 2104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894621673876714e-05,
+      "loss": 1.3848,
+      "step": 2105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945026912146417e-05,
+      "loss": 1.3242,
+      "step": 2106
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943836451592094e-05,
+      "loss": 1.2646,
+      "step": 2107
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8942645357188532e-05,
+      "loss": 1.3418,
+      "step": 2108
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941453629020144e-05,
+      "loss": 1.3389,
+      "step": 2109
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940261267171395e-05,
+      "loss": 1.3018,
+      "step": 2110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939068271726782e-05,
+      "loss": 1.3193,
+      "step": 2111
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937874642770856e-05,
+      "loss": 1.3809,
+      "step": 2112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936680380388205e-05,
+      "loss": 1.334,
+      "step": 2113
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935485484663468e-05,
+      "loss": 1.3291,
+      "step": 2114
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934289955681327e-05,
+      "loss": 1.3262,
+      "step": 2115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933093793526514e-05,
+      "loss": 1.3428,
+      "step": 2116
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931896998283795e-05,
+      "loss": 1.3887,
+      "step": 2117
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930699570037987e-05,
+      "loss": 1.3105,
+      "step": 2118
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929501508873954e-05,
+      "loss": 1.3623,
+      "step": 2119
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89283028148766e-05,
+      "loss": 1.3535,
+      "step": 2120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892710348813088e-05,
+      "loss": 1.3096,
+      "step": 2121
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925903528721787e-05,
+      "loss": 1.3555,
+      "step": 2122
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924702936734364e-05,
+      "loss": 1.2637,
+      "step": 2123
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923501712253697e-05,
+      "loss": 1.3096,
+      "step": 2124
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922299855364916e-05,
+      "loss": 1.3223,
+      "step": 2125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921097366153196e-05,
+      "loss": 1.3398,
+      "step": 2126
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891989424470376e-05,
+      "loss": 1.2793,
+      "step": 2127
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891869049110187e-05,
+      "loss": 1.3174,
+      "step": 2128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8917486105432835e-05,
+      "loss": 1.3467,
+      "step": 2129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891628108778201e-05,
+      "loss": 1.3516,
+      "step": 2130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915075438234802e-05,
+      "loss": 1.3721,
+      "step": 2131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891386915687665e-05,
+      "loss": 1.3164,
+      "step": 2132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8912662243793035e-05,
+      "loss": 1.3662,
+      "step": 2133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911454699069506e-05,
+      "loss": 1.3164,
+      "step": 2134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910246522791632e-05,
+      "loss": 1.3252,
+      "step": 2135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909037715045037e-05,
+      "loss": 1.3545,
+      "step": 2136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890782827591539e-05,
+      "loss": 1.3477,
+      "step": 2137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890661820548841e-05,
+      "loss": 1.2656,
+      "step": 2138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905407503849847e-05,
+      "loss": 1.3525,
+      "step": 2139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904196171085504e-05,
+      "loss": 1.3691,
+      "step": 2140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902984207281232e-05,
+      "loss": 1.3008,
+      "step": 2141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901771612522917e-05,
+      "loss": 1.2832,
+      "step": 2142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89005583868965e-05,
+      "loss": 1.3193,
+      "step": 2143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899344530487964e-05,
+      "loss": 1.3242,
+      "step": 2144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889813004338333e-05,
+      "loss": 1.3877,
+      "step": 2145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896914925668674e-05,
+      "loss": 1.3457,
+      "step": 2146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895699177430106e-05,
+      "loss": 1.3652,
+      "step": 2147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894482798753788e-05,
+      "loss": 1.2773,
+      "step": 2148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893265789725922e-05,
+      "loss": 1.3301,
+      "step": 2149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892048150432766e-05,
+      "loss": 1.3145,
+      "step": 2150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890829880960606e-05,
+      "loss": 1.334,
+      "step": 2151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888961098139578e-05,
+      "loss": 1.3643,
+      "step": 2152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888391451824683e-05,
+      "loss": 1.3467,
+      "step": 2153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887171292333728e-05,
+      "loss": 1.3232,
+      "step": 2154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885950503009393e-05,
+      "loss": 1.3848,
+      "step": 2155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88847290839382e-05,
+      "loss": 1.3301,
+      "step": 2156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883507035206703e-05,
+      "loss": 1.3477,
+      "step": 2157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882284356901517e-05,
+      "loss": 1.3555,
+      "step": 2158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881061049109288e-05,
+      "loss": 1.3242,
+      "step": 2159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887983711191671e-05,
+      "loss": 1.2969,
+      "step": 2160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878612545410527e-05,
+      "loss": 1.3516,
+      "step": 2161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877387349677526e-05,
+      "loss": 1.3379,
+      "step": 2162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876161524804532e-05,
+      "loss": 1.3242,
+      "step": 2163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874935070878416e-05,
+      "loss": 1.3008,
+      "step": 2164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887370798798611e-05,
+      "loss": 1.376,
+      "step": 2165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872480276214565e-05,
+      "loss": 1.2803,
+      "step": 2166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887125193565079e-05,
+      "loss": 1.3438,
+      "step": 2167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870022966381842e-05,
+      "loss": 1.3262,
+      "step": 2168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868793368494818e-05,
+      "loss": 1.3271,
+      "step": 2169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867563142076855e-05,
+      "loss": 1.3174,
+      "step": 2170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886633228721514e-05,
+      "loss": 1.292,
+      "step": 2171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886510080399691e-05,
+      "loss": 1.2939,
+      "step": 2172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886386869250943e-05,
+      "loss": 1.3496,
+      "step": 2173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886263595284003e-05,
+      "loss": 1.3262,
+      "step": 2174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861402585076066e-05,
+      "loss": 1.3457,
+      "step": 2175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860168589304954e-05,
+      "loss": 1.332,
+      "step": 2176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858933965614138e-05,
+      "loss": 1.3496,
+      "step": 2177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857698714091125e-05,
+      "loss": 1.3145,
+      "step": 2178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856462834823453e-05,
+      "loss": 1.3398,
+      "step": 2179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885522632789871e-05,
+      "loss": 1.3438,
+      "step": 2180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853989193404523e-05,
+      "loss": 1.2998,
+      "step": 2181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885275143142857e-05,
+      "loss": 1.3203,
+      "step": 2182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851513042058577e-05,
+      "loss": 1.2617,
+      "step": 2183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850274025382295e-05,
+      "loss": 1.2998,
+      "step": 2184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849034381487547e-05,
+      "loss": 1.3154,
+      "step": 2185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884779411046218e-05,
+      "loss": 1.3018,
+      "step": 2186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846553212394096e-05,
+      "loss": 1.2969,
+      "step": 2187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845311687371232e-05,
+      "loss": 1.3086,
+      "step": 2188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844069535481578e-05,
+      "loss": 1.3262,
+      "step": 2189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884282675681316e-05,
+      "loss": 1.3594,
+      "step": 2190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884158335145406e-05,
+      "loss": 1.3076,
+      "step": 2191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88403393194924e-05,
+      "loss": 1.3086,
+      "step": 2192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883909466101634e-05,
+      "loss": 1.3613,
+      "step": 2193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837849376114086e-05,
+      "loss": 1.2764,
+      "step": 2194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836603464873894e-05,
+      "loss": 1.2822,
+      "step": 2195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835356927384064e-05,
+      "loss": 1.3564,
+      "step": 2196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883410976373294e-05,
+      "loss": 1.3496,
+      "step": 2197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88328619740089e-05,
+      "loss": 1.3174,
+      "step": 2198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831613558300385e-05,
+      "loss": 1.3057,
+      "step": 2199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883036451669586e-05,
+      "loss": 1.3125,
+      "step": 2200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829114849283853e-05,
+      "loss": 1.2988,
+      "step": 2201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882786455615292e-05,
+      "loss": 1.3271,
+      "step": 2202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826613637391675e-05,
+      "loss": 1.3555,
+      "step": 2203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882536209308877e-05,
+      "loss": 1.3359,
+      "step": 2204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824109923332902e-05,
+      "loss": 1.3213,
+      "step": 2205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822857128212807e-05,
+      "loss": 1.3271,
+      "step": 2206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821603707817276e-05,
+      "loss": 1.2988,
+      "step": 2207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882034966223514e-05,
+      "loss": 1.3398,
+      "step": 2208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881909499155527e-05,
+      "loss": 1.3838,
+      "step": 2209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817839695866583e-05,
+      "loss": 1.3281,
+      "step": 2210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816583775258044e-05,
+      "loss": 1.2949,
+      "step": 2211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881532722981866e-05,
+      "loss": 1.333,
+      "step": 2212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814070059637487e-05,
+      "loss": 1.335,
+      "step": 2213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812812264803607e-05,
+      "loss": 1.2754,
+      "step": 2214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811553845406176e-05,
+      "loss": 1.3418,
+      "step": 2215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810294801534368e-05,
+      "loss": 1.3818,
+      "step": 2216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809035133277414e-05,
+      "loss": 1.335,
+      "step": 2217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807774840724587e-05,
+      "loss": 1.3018,
+      "step": 2218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806513923965202e-05,
+      "loss": 1.291,
+      "step": 2219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880525238308862e-05,
+      "loss": 1.3281,
+      "step": 2220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803990218184253e-05,
+      "loss": 1.3623,
+      "step": 2221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802727429341546e-05,
+      "loss": 1.3613,
+      "step": 2222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801464016649985e-05,
+      "loss": 1.3184,
+      "step": 2223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800199980199124e-05,
+      "loss": 1.3623,
+      "step": 2224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879893532007853e-05,
+      "loss": 1.3525,
+      "step": 2225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797670036377838e-05,
+      "loss": 1.2881,
+      "step": 2226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879640412918672e-05,
+      "loss": 1.3271,
+      "step": 2227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795137598594884e-05,
+      "loss": 1.3203,
+      "step": 2228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879387044469209e-05,
+      "loss": 1.3418,
+      "step": 2229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879260266756815e-05,
+      "loss": 1.2627,
+      "step": 2230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87913342673129e-05,
+      "loss": 1.3691,
+      "step": 2231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790065244016236e-05,
+      "loss": 1.3174,
+      "step": 2232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788795597768093e-05,
+      "loss": 1.3242,
+      "step": 2233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787525328658454e-05,
+      "loss": 1.3564,
+      "step": 2234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878625443677734e-05,
+      "loss": 1.4072,
+      "step": 2235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784982922214817e-05,
+      "loss": 1.333,
+      "step": 2236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783710785061003e-05,
+      "loss": 1.3213,
+      "step": 2237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878243802540605e-05,
+      "loss": 1.2607,
+      "step": 2238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878116464334016e-05,
+      "loss": 1.292,
+      "step": 2239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779890638953577e-05,
+      "loss": 1.3242,
+      "step": 2240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778616012336588e-05,
+      "loss": 1.3076,
+      "step": 2241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777340763579528e-05,
+      "loss": 1.3154,
+      "step": 2242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776064892772773e-05,
+      "loss": 1.3945,
+      "step": 2243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877478840000674e-05,
+      "loss": 1.291,
+      "step": 2244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773511285371904e-05,
+      "loss": 1.3047,
+      "step": 2245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772233548958767e-05,
+      "loss": 1.3213,
+      "step": 2246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877095519085788e-05,
+      "loss": 1.3105,
+      "step": 2247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769676211159847e-05,
+      "loss": 1.3408,
+      "step": 2248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768396609955304e-05,
+      "loss": 1.335,
+      "step": 2249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767116387334935e-05,
+      "loss": 1.2686,
+      "step": 2250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765835543389476e-05,
+      "loss": 1.3213,
+      "step": 2251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764554078209697e-05,
+      "loss": 1.3164,
+      "step": 2252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876327199188641e-05,
+      "loss": 1.3232,
+      "step": 2253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761989284510484e-05,
+      "loss": 1.29,
+      "step": 2254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760705956172822e-05,
+      "loss": 1.3203,
+      "step": 2255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875942200696437e-05,
+      "loss": 1.3096,
+      "step": 2256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875813743697613e-05,
+      "loss": 1.3252,
+      "step": 2257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756852246299128e-05,
+      "loss": 1.3291,
+      "step": 2258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755566435024457e-05,
+      "loss": 1.3223,
+      "step": 2259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875428000324323e-05,
+      "loss": 1.3291,
+      "step": 2260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752992951046628e-05,
+      "loss": 1.334,
+      "step": 2261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875170527852586e-05,
+      "loss": 1.2842,
+      "step": 2262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875041698577218e-05,
+      "loss": 1.3467,
+      "step": 2263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749128072876888e-05,
+      "loss": 1.3096,
+      "step": 2264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874783853993134e-05,
+      "loss": 1.3682,
+      "step": 2265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746548387026916e-05,
+      "loss": 1.3506,
+      "step": 2266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874525761425505e-05,
+      "loss": 1.3672,
+      "step": 2267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874396622170722e-05,
+      "loss": 1.3281,
+      "step": 2268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874267420947495e-05,
+      "loss": 1.334,
+      "step": 2269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87413815776498e-05,
+      "loss": 1.3096,
+      "step": 2270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874008832632338e-05,
+      "loss": 1.3047,
+      "step": 2271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738794455587344e-05,
+      "loss": 1.2969,
+      "step": 2272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737499965533387e-05,
+      "loss": 1.3691,
+      "step": 2273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736204856253253e-05,
+      "loss": 1.3359,
+      "step": 2274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734909127838722e-05,
+      "loss": 1.3457,
+      "step": 2275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733612780381625e-05,
+      "loss": 1.3467,
+      "step": 2276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732315813973835e-05,
+      "loss": 1.3311,
+      "step": 2277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873101822870726e-05,
+      "loss": 1.3223,
+      "step": 2278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872972002467387e-05,
+      "loss": 1.3379,
+      "step": 2279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728421201965664e-05,
+      "loss": 1.3242,
+      "step": 2280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727121760674685e-05,
+      "loss": 1.2764,
+      "step": 2281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725821700893035e-05,
+      "loss": 1.291,
+      "step": 2282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724521022712844e-05,
+      "loss": 1.3271,
+      "step": 2283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872321972622629e-05,
+      "loss": 1.332,
+      "step": 2284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872191781152559e-05,
+      "loss": 1.3008,
+      "step": 2285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720615278703024e-05,
+      "loss": 1.3164,
+      "step": 2286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719312127850894e-05,
+      "loss": 1.3076,
+      "step": 2287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718008359061554e-05,
+      "loss": 1.3457,
+      "step": 2288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87167039724274e-05,
+      "loss": 1.3223,
+      "step": 2289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715398968040882e-05,
+      "loss": 1.3154,
+      "step": 2290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714093345994482e-05,
+      "loss": 1.3623,
+      "step": 2291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712787106380727e-05,
+      "loss": 1.2939,
+      "step": 2292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711480249292192e-05,
+      "loss": 1.3496,
+      "step": 2293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871017277482149e-05,
+      "loss": 1.3311,
+      "step": 2294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870886468306129e-05,
+      "loss": 1.3711,
+      "step": 2295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870755597410429e-05,
+      "loss": 1.3145,
+      "step": 2296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706246648043237e-05,
+      "loss": 1.3379,
+      "step": 2297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704936704970925e-05,
+      "loss": 1.3447,
+      "step": 2298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870362614498019e-05,
+      "loss": 1.377,
+      "step": 2299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702314968163914e-05,
+      "loss": 1.3193,
+      "step": 2300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701003174615016e-05,
+      "loss": 1.3398,
+      "step": 2301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699690764426465e-05,
+      "loss": 1.3252,
+      "step": 2302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869837773769127e-05,
+      "loss": 1.3887,
+      "step": 2303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697064094502486e-05,
+      "loss": 1.3408,
+      "step": 2304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869574983495321e-05,
+      "loss": 1.3252,
+      "step": 2305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694434959136584e-05,
+      "loss": 1.3271,
+      "step": 2306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693119467145797e-05,
+      "loss": 1.3662,
+      "step": 2307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869180335907407e-05,
+      "loss": 1.3076,
+      "step": 2308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869048663501468e-05,
+      "loss": 1.3604,
+      "step": 2309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868916929506094e-05,
+      "loss": 1.3037,
+      "step": 2310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687851339306217e-05,
+      "loss": 1.2783,
+      "step": 2311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686532767843907e-05,
+      "loss": 1.3184,
+      "step": 2312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685213580767463e-05,
+      "loss": 1.3311,
+      "step": 2313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868389377817037e-05,
+      "loss": 1.2832,
+      "step": 2314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868257336014617e-05,
+      "loss": 1.3242,
+      "step": 2315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868125232678843e-05,
+      "loss": 1.3193,
+      "step": 2316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679930678190783e-05,
+      "loss": 1.3232,
+      "step": 2317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678608414446885e-05,
+      "loss": 1.3213,
+      "step": 2318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867728553565045e-05,
+      "loss": 1.3213,
+      "step": 2319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867596204189523e-05,
+      "loss": 1.2939,
+      "step": 2320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867463793327502e-05,
+      "loss": 1.3633,
+      "step": 2321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867331320988366e-05,
+      "loss": 1.3271,
+      "step": 2322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867198787181503e-05,
+      "loss": 1.335,
+      "step": 2323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867066191916306e-05,
+      "loss": 1.3125,
+      "step": 2324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866933535202172e-05,
+      "loss": 1.293,
+      "step": 2325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866800817048502e-05,
+      "loss": 1.3105,
+      "step": 2326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666680374647024e-05,
+      "loss": 1.3076,
+      "step": 2327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665351964601827e-05,
+      "loss": 1.3193,
+      "step": 2328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664022940443578e-05,
+      "loss": 1.3008,
+      "step": 2329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662693302266463e-05,
+      "loss": 1.3115,
+      "step": 2330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866136305016471e-05,
+      "loss": 1.3535,
+      "step": 2331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86600321842326e-05,
+      "loss": 1.3389,
+      "step": 2332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865870070456444e-05,
+      "loss": 1.2852,
+      "step": 2333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865736861125461e-05,
+      "loss": 1.3311,
+      "step": 2334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86560359043975e-05,
+      "loss": 1.3105,
+      "step": 2335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654702584087566e-05,
+      "loss": 1.3389,
+      "step": 2336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653368650419296e-05,
+      "loss": 1.333,
+      "step": 2337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652034103487233e-05,
+      "loss": 1.3496,
+      "step": 2338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865069894338595e-05,
+      "loss": 1.3457,
+      "step": 2339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649363170210068e-05,
+      "loss": 1.334,
+      "step": 2340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864802678405426e-05,
+      "loss": 1.3477,
+      "step": 2341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864668978501323e-05,
+      "loss": 1.29,
+      "step": 2342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645352173181735e-05,
+      "loss": 1.3467,
+      "step": 2343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644013948654575e-05,
+      "loss": 1.3203,
+      "step": 2344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642675111526577e-05,
+      "loss": 1.3223,
+      "step": 2345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641335661892636e-05,
+      "loss": 1.3203,
+      "step": 2346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863999559984767e-05,
+      "loss": 1.29,
+      "step": 2347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638654925486656e-05,
+      "loss": 1.3066,
+      "step": 2348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637313638904607e-05,
+      "loss": 1.3408,
+      "step": 2349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863597174019658e-05,
+      "loss": 1.3164,
+      "step": 2350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634629229457667e-05,
+      "loss": 1.3389,
+      "step": 2351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633286106783023e-05,
+      "loss": 1.2793,
+      "step": 2352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863194237226783e-05,
+      "loss": 1.3301,
+      "step": 2353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630598026007322e-05,
+      "loss": 1.2666,
+      "step": 2354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629253068096763e-05,
+      "loss": 1.25,
+      "step": 2355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862790749863148e-05,
+      "loss": 1.3174,
+      "step": 2356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626561317706828e-05,
+      "loss": 1.2842,
+      "step": 2357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625214525418217e-05,
+      "loss": 1.3457,
+      "step": 2358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623867121861085e-05,
+      "loss": 1.2979,
+      "step": 2359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862251910713093e-05,
+      "loss": 1.3174,
+      "step": 2360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862117048132328e-05,
+      "loss": 1.3115,
+      "step": 2361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861982124453372e-05,
+      "loss": 1.2744,
+      "step": 2362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618471396857862e-05,
+      "loss": 1.2881,
+      "step": 2363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617120938391373e-05,
+      "loss": 1.3369,
+      "step": 2364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861576986922996e-05,
+      "loss": 1.3223,
+      "step": 2365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614418189469373e-05,
+      "loss": 1.3662,
+      "step": 2366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613065899205407e-05,
+      "loss": 1.3174,
+      "step": 2367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611712998533897e-05,
+      "loss": 1.2803,
+      "step": 2368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610359487550723e-05,
+      "loss": 1.3408,
+      "step": 2369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860900536635181e-05,
+      "loss": 1.3066,
+      "step": 2370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860765063503312e-05,
+      "loss": 1.3096,
+      "step": 2371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8606295293690668e-05,
+      "loss": 1.3096,
+      "step": 2372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604939342420505e-05,
+      "loss": 1.2676,
+      "step": 2373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8603582781318725e-05,
+      "loss": 1.3125,
+      "step": 2374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860222561048147e-05,
+      "loss": 1.334,
+      "step": 2375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600867830004925e-05,
+      "loss": 1.2939,
+      "step": 2376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859950943998531e-05,
+      "loss": 1.249,
+      "step": 2377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85981504405189e-05,
+      "loss": 1.332,
+      "step": 2378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596790831702e-05,
+      "loss": 1.3281,
+      "step": 2379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595430613630972e-05,
+      "loss": 1.2939,
+      "step": 2380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594069786402214e-05,
+      "loss": 1.2793,
+      "step": 2381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8592708350112162e-05,
+      "loss": 1.3252,
+      "step": 2382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591346304857307e-05,
+      "loss": 1.334,
+      "step": 2383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589983650734176e-05,
+      "loss": 1.2979,
+      "step": 2384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858862038783934e-05,
+      "loss": 1.3262,
+      "step": 2385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587256516269407e-05,
+      "loss": 1.3076,
+      "step": 2386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585892036121045e-05,
+      "loss": 1.3125,
+      "step": 2387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584526947490946e-05,
+      "loss": 1.3281,
+      "step": 2388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858316125047586e-05,
+      "loss": 1.3135,
+      "step": 2389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858179494517257e-05,
+      "loss": 1.2754,
+      "step": 2390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858042803167791e-05,
+      "loss": 1.3301,
+      "step": 2391
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579060510088747e-05,
+      "loss": 1.3223,
+      "step": 2392
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577692380502e-05,
+      "loss": 1.3242,
+      "step": 2393
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857632364301463e-05,
+      "loss": 1.293,
+      "step": 2394
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574954297723643e-05,
+      "loss": 1.3379,
+      "step": 2395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8573584344726073e-05,
+      "loss": 1.2715,
+      "step": 2396
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857221378411902e-05,
+      "loss": 1.3115,
+      "step": 2397
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857084261599961e-05,
+      "loss": 1.3643,
+      "step": 2398
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569470840465013e-05,
+      "loss": 1.25,
+      "step": 2399
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568098457612457e-05,
+      "loss": 1.3213,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566725467539196e-05,
+      "loss": 1.3096,
+      "step": 2401
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565351870342538e-05,
+      "loss": 1.3203,
+      "step": 2402
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563977666119823e-05,
+      "loss": 1.3496,
+      "step": 2403
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562602854968446e-05,
+      "loss": 1.3545,
+      "step": 2404
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856122743698584e-05,
+      "loss": 1.3584,
+      "step": 2405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855985141226948e-05,
+      "loss": 1.2861,
+      "step": 2406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558474780916883e-05,
+      "loss": 1.3184,
+      "step": 2407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557097543025613e-05,
+      "loss": 1.3379,
+      "step": 2408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555719698693275e-05,
+      "loss": 1.3203,
+      "step": 2409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554341248017513e-05,
+      "loss": 1.3281,
+      "step": 2410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552962191096023e-05,
+      "loss": 1.335,
+      "step": 2411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551582528026534e-05,
+      "loss": 1.3242,
+      "step": 2412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550202258906824e-05,
+      "loss": 1.3359,
+      "step": 2413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548821383834717e-05,
+      "loss": 1.3135,
+      "step": 2414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854743990290807e-05,
+      "loss": 1.3252,
+      "step": 2415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854605781622479e-05,
+      "loss": 1.3262,
+      "step": 2416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544675123882828e-05,
+      "loss": 1.3359,
+      "step": 2417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543291825980168e-05,
+      "loss": 1.3086,
+      "step": 2418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541907922614854e-05,
+      "loss": 1.3379,
+      "step": 2419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540523413884956e-05,
+      "loss": 1.3184,
+      "step": 2420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85391382998886e-05,
+      "loss": 1.2734,
+      "step": 2421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853775258072394e-05,
+      "loss": 1.2969,
+      "step": 2422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536366256489194e-05,
+      "loss": 1.333,
+      "step": 2423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85349793272826e-05,
+      "loss": 1.3359,
+      "step": 2424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533591793202455e-05,
+      "loss": 1.3203,
+      "step": 2425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853220365434709e-05,
+      "loss": 1.2959,
+      "step": 2426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530814910814888e-05,
+      "loss": 1.2549,
+      "step": 2427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529425562704266e-05,
+      "loss": 1.3047,
+      "step": 2428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528035610113683e-05,
+      "loss": 1.3467,
+      "step": 2429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852664505314165e-05,
+      "loss": 1.293,
+      "step": 2430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525253891886715e-05,
+      "loss": 1.3086,
+      "step": 2431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852386212644747e-05,
+      "loss": 1.2734,
+      "step": 2432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522469756922546e-05,
+      "loss": 1.3076,
+      "step": 2433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521076783410628e-05,
+      "loss": 1.3574,
+      "step": 2434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519683206010422e-05,
+      "loss": 1.3604,
+      "step": 2435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518289024820707e-05,
+      "loss": 1.3027,
+      "step": 2436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516894239940276e-05,
+      "loss": 1.3193,
+      "step": 2437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851549885146798e-05,
+      "loss": 1.3008,
+      "step": 2438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514102859502717e-05,
+      "loss": 1.3164,
+      "step": 2439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512706264143412e-05,
+      "loss": 1.3057,
+      "step": 2440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851130906548905e-05,
+      "loss": 1.3047,
+      "step": 2441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509911263638643e-05,
+      "loss": 1.3105,
+      "step": 2442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508512858691258e-05,
+      "loss": 1.3516,
+      "step": 2443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507113850745994e-05,
+      "loss": 1.3213,
+      "step": 2444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505714239902007e-05,
+      "loss": 1.3467,
+      "step": 2445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850431402625848e-05,
+      "loss": 1.3242,
+      "step": 2446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850291320991465e-05,
+      "loss": 1.3057,
+      "step": 2447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501511790969794e-05,
+      "loss": 1.3799,
+      "step": 2448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500109769523226e-05,
+      "loss": 1.3555,
+      "step": 2449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498707145674307e-05,
+      "loss": 1.3115,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849730391952245e-05,
+      "loss": 1.3467,
+      "step": 2451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849590009116709e-05,
+      "loss": 1.2754,
+      "step": 2452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494495660707725e-05,
+      "loss": 1.3125,
+      "step": 2453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849309062824388e-05,
+      "loss": 1.2998,
+      "step": 2454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491684993875135e-05,
+      "loss": 1.3154,
+      "step": 2455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8490278757701103e-05,
+      "loss": 1.3164,
+      "step": 2456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488871919821445e-05,
+      "loss": 1.2803,
+      "step": 2457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487464480335868e-05,
+      "loss": 1.3057,
+      "step": 2458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848605643934411e-05,
+      "loss": 1.3027,
+      "step": 2459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484647796945966e-05,
+      "loss": 1.2969,
+      "step": 2460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848323855324126e-05,
+      "loss": 1.2725,
+      "step": 2461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848182870832987e-05,
+      "loss": 1.2988,
+      "step": 2462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480418262311707e-05,
+      "loss": 1.2959,
+      "step": 2463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479007215286732e-05,
+      "loss": 1.3301,
+      "step": 2464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477595567354947e-05,
+      "loss": 1.2842,
+      "step": 2465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847618331861639e-05,
+      "loss": 1.3252,
+      "step": 2466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474770469171156e-05,
+      "loss": 1.3154,
+      "step": 2467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473357019119367e-05,
+      "loss": 1.2959,
+      "step": 2468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471942968561192e-05,
+      "loss": 1.3369,
+      "step": 2469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470528317596856e-05,
+      "loss": 1.3457,
+      "step": 2470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84691130663266e-05,
+      "loss": 1.2939,
+      "step": 2471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846769721485073e-05,
+      "loss": 1.3408,
+      "step": 2472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466280763269597e-05,
+      "loss": 1.3359,
+      "step": 2473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464863711683566e-05,
+      "loss": 1.3008,
+      "step": 2474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463446060193078e-05,
+      "loss": 1.3213,
+      "step": 2475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462027808898597e-05,
+      "loss": 1.3057,
+      "step": 2476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460608957900628e-05,
+      "loss": 1.292,
+      "step": 2477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459189507299736e-05,
+      "loss": 1.3086,
+      "step": 2478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457769457196513e-05,
+      "loss": 1.3584,
+      "step": 2479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456348807691597e-05,
+      "loss": 1.3232,
+      "step": 2480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454927558885666e-05,
+      "loss": 1.2969,
+      "step": 2481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453505710879454e-05,
+      "loss": 1.3057,
+      "step": 2482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452083263773716e-05,
+      "loss": 1.3291,
+      "step": 2483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450660217669266e-05,
+      "loss": 1.3369,
+      "step": 2484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449236572666956e-05,
+      "loss": 1.2998,
+      "step": 2485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8447812328867676e-05,
+      "loss": 1.3164,
+      "step": 2486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446387486372367e-05,
+      "loss": 1.3486,
+      "step": 2487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444962045282005e-05,
+      "loss": 1.3506,
+      "step": 2488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844353600569761e-05,
+      "loss": 1.3203,
+      "step": 2489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844210936772025e-05,
+      "loss": 1.3115,
+      "step": 2490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844068213145102e-05,
+      "loss": 1.3564,
+      "step": 2491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439254296991082e-05,
+      "loss": 1.3525,
+      "step": 2492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437825864441616e-05,
+      "loss": 1.3154,
+      "step": 2493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843639683390386e-05,
+      "loss": 1.2988,
+      "step": 2494
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434967205479085e-05,
+      "loss": 1.2656,
+      "step": 2495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433536979268618e-05,
+      "loss": 1.3057,
+      "step": 2496
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432106155373807e-05,
+      "loss": 1.333,
+      "step": 2497
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430674733896063e-05,
+      "loss": 1.3545,
+      "step": 2498
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842924271493683e-05,
+      "loss": 1.2959,
+      "step": 2499
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842781009859759e-05,
+      "loss": 1.333,
+      "step": 2500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426376884979876e-05,
+      "loss": 1.3281,
+      "step": 2501
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842494307418526e-05,
+      "loss": 1.2939,
+      "step": 2502
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423508666315354e-05,
+      "loss": 1.3301,
+      "step": 2503
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422073661471817e-05,
+      "loss": 1.3008,
+      "step": 2504
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842063805975635e-05,
+      "loss": 1.3047,
+      "step": 2505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419201861270684e-05,
+      "loss": 1.3945,
+      "step": 2506
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841776506611661e-05,
+      "loss": 1.2939,
+      "step": 2507
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416327674395956e-05,
+      "loss": 1.3076,
+      "step": 2508
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414889686210585e-05,
+      "loss": 1.2998,
+      "step": 2509
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413451101662407e-05,
+      "loss": 1.3232,
+      "step": 2510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841201192085338e-05,
+      "loss": 1.3223,
+      "step": 2511
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841057214388549e-05,
+      "loss": 1.3457,
+      "step": 2512
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409131770860783e-05,
+      "loss": 1.3066,
+      "step": 2513
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840769080188133e-05,
+      "loss": 1.3057,
+      "step": 2514
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840624923704926e-05,
+      "loss": 1.335,
+      "step": 2515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840480707646673e-05,
+      "loss": 1.2939,
+      "step": 2516
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403364320235946e-05,
+      "loss": 1.2949,
+      "step": 2517
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401920968459164e-05,
+      "loss": 1.3555,
+      "step": 2518
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400477021238665e-05,
+      "loss": 1.3389,
+      "step": 2519
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8399032478676786e-05,
+      "loss": 1.2959,
+      "step": 2520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397587340875903e-05,
+      "loss": 1.2617,
+      "step": 2521
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839614160793843e-05,
+      "loss": 1.3164,
+      "step": 2522
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394695279966823e-05,
+      "loss": 1.3105,
+      "step": 2523
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393248357063595e-05,
+      "loss": 1.3154,
+      "step": 2524
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391800839331275e-05,
+      "loss": 1.3115,
+      "step": 2525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390352726872455e-05,
+      "loss": 1.3311,
+      "step": 2526
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388904019789766e-05,
+      "loss": 1.3203,
+      "step": 2527
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387454718185874e-05,
+      "loss": 1.3047,
+      "step": 2528
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386004822163488e-05,
+      "loss": 1.291,
+      "step": 2529
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384554331825368e-05,
+      "loss": 1.3564,
+      "step": 2530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8383103247274306e-05,
+      "loss": 1.3057,
+      "step": 2531
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381651568613143e-05,
+      "loss": 1.2969,
+      "step": 2532
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380199295944758e-05,
+      "loss": 1.3193,
+      "step": 2533
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378746429372073e-05,
+      "loss": 1.2949,
+      "step": 2534
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377292968998057e-05,
+      "loss": 1.3486,
+      "step": 2535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837583891492571e-05,
+      "loss": 1.3184,
+      "step": 2536
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374384267258083e-05,
+      "loss": 1.3271,
+      "step": 2537
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837292902609827e-05,
+      "loss": 1.2861,
+      "step": 2538
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371473191549397e-05,
+      "loss": 1.2852,
+      "step": 2539
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370016763714652e-05,
+      "loss": 1.3105,
+      "step": 2540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368559742697235e-05,
+      "loss": 1.3213,
+      "step": 2541
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367102128600416e-05,
+      "loss": 1.3447,
+      "step": 2542
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8365643921527493e-05,
+      "loss": 1.3213,
+      "step": 2543
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836418512158181e-05,
+      "loss": 1.3516,
+      "step": 2544
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836272572886675e-05,
+      "loss": 1.3145,
+      "step": 2545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361265743485742e-05,
+      "loss": 1.3213,
+      "step": 2546
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359805165542256e-05,
+      "loss": 1.2988,
+      "step": 2547
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83583439951398e-05,
+      "loss": 1.3096,
+      "step": 2548
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835688223238193e-05,
+      "loss": 1.3203,
+      "step": 2549
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835541987737224e-05,
+      "loss": 1.3369,
+      "step": 2550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353956930214368e-05,
+      "loss": 1.2656,
+      "step": 2551
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835249339101199e-05,
+      "loss": 1.3447,
+      "step": 2552
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351029259868828e-05,
+      "loss": 1.2861,
+      "step": 2553
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349564536888648e-05,
+      "loss": 1.2881,
+      "step": 2554
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834809922217525e-05,
+      "loss": 1.3242,
+      "step": 2555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346633315832486e-05,
+      "loss": 1.2842,
+      "step": 2556
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834516681796424e-05,
+      "loss": 1.3105,
+      "step": 2557
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343699728674444e-05,
+      "loss": 1.3281,
+      "step": 2558
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8342232048067076e-05,
+      "loss": 1.3467,
+      "step": 2559
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834076377624614e-05,
+      "loss": 1.2422,
+      "step": 2560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83392949133157e-05,
+      "loss": 1.3076,
+      "step": 2561
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337825459379853e-05,
+      "loss": 1.2852,
+      "step": 2562
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336355414542735e-05,
+      "loss": 1.333,
+      "step": 2563
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334884778908536e-05,
+      "loss": 1.2783,
+      "step": 2564
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333413552581474e-05,
+      "loss": 1.2832,
+      "step": 2565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331941735665816e-05,
+      "loss": 1.3027,
+      "step": 2566
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833046932826587e-05,
+      "loss": 1.3018,
+      "step": 2567
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328996330485984e-05,
+      "loss": 1.2998,
+      "step": 2568
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832752274243055e-05,
+      "loss": 1.3604,
+      "step": 2569
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326048564204003e-05,
+      "loss": 1.3281,
+      "step": 2570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324573795910816e-05,
+      "loss": 1.3467,
+      "step": 2571
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323098437655505e-05,
+      "loss": 1.291,
+      "step": 2572
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832162248954263e-05,
+      "loss": 1.3506,
+      "step": 2573
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320145951676787e-05,
+      "loss": 1.2803,
+      "step": 2574
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831866882416263e-05,
+      "loss": 1.2578,
+      "step": 2575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317191107104833e-05,
+      "loss": 1.3047,
+      "step": 2576
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831571280060812e-05,
+      "loss": 1.3018,
+      "step": 2577
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314233904777265e-05,
+      "loss": 1.2695,
+      "step": 2578
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312754419717073e-05,
+      "loss": 1.3066,
+      "step": 2579
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83112743455324e-05,
+      "loss": 1.3477,
+      "step": 2580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309793682328136e-05,
+      "loss": 1.3242,
+      "step": 2581
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830831243020921e-05,
+      "loss": 1.3115,
+      "step": 2582
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306830589280606e-05,
+      "loss": 1.3545,
+      "step": 2583
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305348159647337e-05,
+      "loss": 1.2852,
+      "step": 2584
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830386514141447e-05,
+      "loss": 1.3291,
+      "step": 2585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302381534687096e-05,
+      "loss": 1.3203,
+      "step": 2586
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300897339570367e-05,
+      "loss": 1.2607,
+      "step": 2587
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299412556169462e-05,
+      "loss": 1.3477,
+      "step": 2588
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829792718458961e-05,
+      "loss": 1.2988,
+      "step": 2589
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8296441224936083e-05,
+      "loss": 1.3291,
+      "step": 2590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294954677314187e-05,
+      "loss": 1.3057,
+      "step": 2591
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829346754182927e-05,
+      "loss": 1.2861,
+      "step": 2592
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291979818586728e-05,
+      "loss": 1.2842,
+      "step": 2593
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290491507691998e-05,
+      "loss": 1.2559,
+      "step": 2594
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8289002609250554e-05,
+      "loss": 1.3633,
+      "step": 2595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828751312336792e-05,
+      "loss": 1.335,
+      "step": 2596
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286023050149646e-05,
+      "loss": 1.29,
+      "step": 2597
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828453238970134e-05,
+      "loss": 1.2988,
+      "step": 2598
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283041142128642e-05,
+      "loss": 1.3066,
+      "step": 2599
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8281549307537245e-05,
+      "loss": 1.2686,
+      "step": 2600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280056886032863e-05,
+      "loss": 1.3369,
+      "step": 2601
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278563877721266e-05,
+      "loss": 1.2832,
+      "step": 2602
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277070282708274e-05,
+      "loss": 1.3525,
+      "step": 2603
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275576101099724e-05,
+      "loss": 1.3086,
+      "step": 2604
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827408133300152e-05,
+      "loss": 1.3408,
+      "step": 2605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272585978519587e-05,
+      "loss": 1.2969,
+      "step": 2606
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827109003775991e-05,
+      "loss": 1.3271,
+      "step": 2607
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8269593510828497e-05,
+      "loss": 1.2988,
+      "step": 2608
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268096397831417e-05,
+      "loss": 1.3105,
+      "step": 2609
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826659869887476e-05,
+      "loss": 1.3066,
+      "step": 2610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265100414064673e-05,
+      "loss": 1.3271,
+      "step": 2611
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826360154350734e-05,
+      "loss": 1.334,
+      "step": 2612
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262102087308983e-05,
+      "loss": 1.2822,
+      "step": 2613
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826060204557587e-05,
+      "loss": 1.3223,
+      "step": 2614
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259101418414305e-05,
+      "loss": 1.3643,
+      "step": 2615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257600205930646e-05,
+      "loss": 1.291,
+      "step": 2616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256098408231277e-05,
+      "loss": 1.3467,
+      "step": 2617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254596025422636e-05,
+      "loss": 1.29,
+      "step": 2618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825309305761119e-05,
+      "loss": 1.2725,
+      "step": 2619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251589504903458e-05,
+      "loss": 1.291,
+      "step": 2620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250085367405995e-05,
+      "loss": 1.3047,
+      "step": 2621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82485806452254e-05,
+      "loss": 1.2891,
+      "step": 2622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247075338468316e-05,
+      "loss": 1.3242,
+      "step": 2623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824556944724142e-05,
+      "loss": 1.3027,
+      "step": 2624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244062971651436e-05,
+      "loss": 1.3086,
+      "step": 2625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242555911805124e-05,
+      "loss": 1.3135,
+      "step": 2626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241048267809295e-05,
+      "loss": 1.3027,
+      "step": 2627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823954003977079e-05,
+      "loss": 1.3398,
+      "step": 2628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238031227796503e-05,
+      "loss": 1.2695,
+      "step": 2629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823652183199336e-05,
+      "loss": 1.2871,
+      "step": 2630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235011852468328e-05,
+      "loss": 1.3066,
+      "step": 2631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823350128932843e-05,
+      "loss": 1.2832,
+      "step": 2632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823199014268071e-05,
+      "loss": 1.3252,
+      "step": 2633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230478412632267e-05,
+      "loss": 1.3184,
+      "step": 2634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228966099290234e-05,
+      "loss": 1.334,
+      "step": 2635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227453202761794e-05,
+      "loss": 1.29,
+      "step": 2636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822593972315416e-05,
+      "loss": 1.2764,
+      "step": 2637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224425660574595e-05,
+      "loss": 1.3477,
+      "step": 2638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222911015130402e-05,
+      "loss": 1.2803,
+      "step": 2639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822139578692892e-05,
+      "loss": 1.3213,
+      "step": 2640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219879976077537e-05,
+      "loss": 1.3184,
+      "step": 2641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821836358268368e-05,
+      "loss": 1.2793,
+      "step": 2642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216846606854804e-05,
+      "loss": 1.3037,
+      "step": 2643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215329048698432e-05,
+      "loss": 1.3379,
+      "step": 2644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213810908322106e-05,
+      "loss": 1.3457,
+      "step": 2645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212292185833414e-05,
+      "loss": 1.2383,
+      "step": 2646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210772881339996e-05,
+      "loss": 1.3174,
+      "step": 2647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820925299494952e-05,
+      "loss": 1.3066,
+      "step": 2648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8207732526769698e-05,
+      "loss": 1.2656,
+      "step": 2649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206211476908286e-05,
+      "loss": 1.2773,
+      "step": 2650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204689845473084e-05,
+      "loss": 1.3457,
+      "step": 2651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203167632571932e-05,
+      "loss": 1.3643,
+      "step": 2652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82016448383127e-05,
+      "loss": 1.2949,
+      "step": 2653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200121462803318e-05,
+      "loss": 1.2744,
+      "step": 2654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819859750615174e-05,
+      "loss": 1.292,
+      "step": 2655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197072968465978e-05,
+      "loss": 1.3496,
+      "step": 2656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195547849854066e-05,
+      "loss": 1.2939,
+      "step": 2657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81940221504241e-05,
+      "loss": 1.293,
+      "step": 2658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192495870284193e-05,
+      "loss": 1.3223,
+      "step": 2659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819096900954252e-05,
+      "loss": 1.2686,
+      "step": 2660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189441568307287e-05,
+      "loss": 1.3076,
+      "step": 2661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187913546686748e-05,
+      "loss": 1.3145,
+      "step": 2662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818638494478919e-05,
+      "loss": 1.335,
+      "step": 2663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818485576272295e-05,
+      "loss": 1.2881,
+      "step": 2664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183326000596393e-05,
+      "loss": 1.3174,
+      "step": 2665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181795658517937e-05,
+      "loss": 1.3027,
+      "step": 2666
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180264736596037e-05,
+      "loss": 1.334,
+      "step": 2667
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817873323493919e-05,
+      "loss": 1.3555,
+      "step": 2668
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177201153655936e-05,
+      "loss": 1.3203,
+      "step": 2669
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175668492854847e-05,
+      "loss": 1.3193,
+      "step": 2670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817413525264455e-05,
+      "loss": 1.3164,
+      "step": 2671
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172601433133698e-05,
+      "loss": 1.3428,
+      "step": 2672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171067034431e-05,
+      "loss": 1.2402,
+      "step": 2673
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169532056645197e-05,
+      "loss": 1.3076,
+      "step": 2674
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167996499885068e-05,
+      "loss": 1.3496,
+      "step": 2675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816646036425944e-05,
+      "loss": 1.3398,
+      "step": 2676
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816492364987718e-05,
+      "loss": 1.2969,
+      "step": 2677
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81633863568472e-05,
+      "loss": 1.3389,
+      "step": 2678
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816184848527844e-05,
+      "loss": 1.3027,
+      "step": 2679
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160310035279888e-05,
+      "loss": 1.3252,
+      "step": 2680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158771006960582e-05,
+      "loss": 1.3115,
+      "step": 2681
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815723140042959e-05,
+      "loss": 1.3154,
+      "step": 2682
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155691215796018e-05,
+      "loss": 1.3477,
+      "step": 2683
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154150453169023e-05,
+      "loss": 1.3301,
+      "step": 2684
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152609112657802e-05,
+      "loss": 1.2988,
+      "step": 2685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151067194371584e-05,
+      "loss": 1.3389,
+      "step": 2686
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8149524698419646e-05,
+      "loss": 1.3369,
+      "step": 2687
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147981624911307e-05,
+      "loss": 1.3271,
+      "step": 2688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146437973955926e-05,
+      "loss": 1.3271,
+      "step": 2689
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8144893745662894e-05,
+      "loss": 1.2363,
+      "step": 2690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814334894014166e-05,
+      "loss": 1.3379,
+      "step": 2691
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141803557501695e-05,
+      "loss": 1.2988,
+      "step": 2692
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140257597852526e-05,
+      "loss": 1.3184,
+      "step": 2693
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138711061303715e-05,
+      "loss": 1.3145,
+      "step": 2694
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137163947964863e-05,
+      "loss": 1.2676,
+      "step": 2695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135616257945615e-05,
+      "loss": 1.3477,
+      "step": 2696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134067991355655e-05,
+      "loss": 1.2705,
+      "step": 2697
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132519148304707e-05,
+      "loss": 1.2988,
+      "step": 2698
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130969728902542e-05,
+      "loss": 1.375,
+      "step": 2699
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129419733258965e-05,
+      "loss": 1.2871,
+      "step": 2700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812786916148382e-05,
+      "loss": 1.3262,
+      "step": 2701
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126318013687004e-05,
+      "loss": 1.3076,
+      "step": 2702
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812476628997844e-05,
+      "loss": 1.3154,
+      "step": 2703
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123213990468102e-05,
+      "loss": 1.293,
+      "step": 2704
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121661115265998e-05,
+      "loss": 1.2979,
+      "step": 2705
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8120107664482184e-05,
+      "loss": 1.3174,
+      "step": 2706
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118553638226753e-05,
+      "loss": 1.3359,
+      "step": 2707
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116999036609835e-05,
+      "loss": 1.3232,
+      "step": 2708
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115443859741606e-05,
+      "loss": 1.2969,
+      "step": 2709
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113888107732284e-05,
+      "loss": 1.2568,
+      "step": 2710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811233178069212e-05,
+      "loss": 1.3389,
+      "step": 2711
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811077487873142e-05,
+      "loss": 1.293,
+      "step": 2712
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810921740196051e-05,
+      "loss": 1.3213,
+      "step": 2713
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107659350489776e-05,
+      "loss": 1.3301,
+      "step": 2714
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106100724429634e-05,
+      "loss": 1.2832,
+      "step": 2715
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104541523890543e-05,
+      "loss": 1.3076,
+      "step": 2716
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810298174898301e-05,
+      "loss": 1.3457,
+      "step": 2717
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810142139981757e-05,
+      "loss": 1.2646,
+      "step": 2718
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099860476504805e-05,
+      "loss": 1.3057,
+      "step": 2719
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809829897915534e-05,
+      "loss": 1.2939,
+      "step": 2720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096736907879834e-05,
+      "loss": 1.3018,
+      "step": 2721
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095174262788992e-05,
+      "loss": 1.2441,
+      "step": 2722
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8093611043993566e-05,
+      "loss": 1.2969,
+      "step": 2723
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809204725160433e-05,
+      "loss": 1.3281,
+      "step": 2724
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809048288573212e-05,
+      "loss": 1.3252,
+      "step": 2725
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088917946487797e-05,
+      "loss": 1.2393,
+      "step": 2726
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087352433982268e-05,
+      "loss": 1.3535,
+      "step": 2727
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808578634832648e-05,
+      "loss": 1.2988,
+      "step": 2728
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084219689631428e-05,
+      "loss": 1.3037,
+      "step": 2729
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082652458008137e-05,
+      "loss": 1.2891,
+      "step": 2730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081084653567673e-05,
+      "loss": 1.2393,
+      "step": 2731
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807951627642115e-05,
+      "loss": 1.2998,
+      "step": 2732
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077947326679717e-05,
+      "loss": 1.2939,
+      "step": 2733
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807637780445457e-05,
+      "loss": 1.3184,
+      "step": 2734
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807480770985693e-05,
+      "loss": 1.2842,
+      "step": 2735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073237042998086e-05,
+      "loss": 1.291,
+      "step": 2736
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071665803989334e-05,
+      "loss": 1.3174,
+      "step": 2737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070093992942042e-05,
+      "loss": 1.292,
+      "step": 2738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068521609967598e-05,
+      "loss": 1.2793,
+      "step": 2739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8066948655177433e-05,
+      "loss": 1.3379,
+      "step": 2740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065375128683027e-05,
+      "loss": 1.3535,
+      "step": 2741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063801030595895e-05,
+      "loss": 1.3311,
+      "step": 2742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806222636102759e-05,
+      "loss": 1.3301,
+      "step": 2743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060651120089713e-05,
+      "loss": 1.2744,
+      "step": 2744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80590753078939e-05,
+      "loss": 1.373,
+      "step": 2745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805749892455183e-05,
+      "loss": 1.3496,
+      "step": 2746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805592197017522e-05,
+      "loss": 1.2383,
+      "step": 2747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054344444875824e-05,
+      "loss": 1.29,
+      "step": 2748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8052766348765448e-05,
+      "loss": 1.3262,
+      "step": 2749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805118768195593e-05,
+      "loss": 1.2988,
+      "step": 2750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804960844455915e-05,
+      "loss": 1.3184,
+      "step": 2751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048028636687027e-05,
+      "loss": 1.3174,
+      "step": 2752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804644825845152e-05,
+      "loss": 1.2832,
+      "step": 2753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044867309964636e-05,
+      "loss": 1.3408,
+      "step": 2754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043285791338413e-05,
+      "loss": 1.3203,
+      "step": 2755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804170370268493e-05,
+      "loss": 1.2949,
+      "step": 2756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8040121044116316e-05,
+      "loss": 1.2939,
+      "step": 2757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038537815744734e-05,
+      "loss": 1.334,
+      "step": 2758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8036954017682383e-05,
+      "loss": 1.2988,
+      "step": 2759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035369650041505e-05,
+      "loss": 1.3418,
+      "step": 2760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803378471293439e-05,
+      "loss": 1.3232,
+      "step": 2761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032199206473358e-05,
+      "loss": 1.3115,
+      "step": 2762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030613130770776e-05,
+      "loss": 1.3408,
+      "step": 2763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802902648593905e-05,
+      "loss": 1.2852,
+      "step": 2764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8027439272090625e-05,
+      "loss": 1.2871,
+      "step": 2765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025851489337983e-05,
+      "loss": 1.3184,
+      "step": 2766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024263137793655e-05,
+      "loss": 1.3057,
+      "step": 2767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022674217570202e-05,
+      "loss": 1.2783,
+      "step": 2768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8021084728780235e-05,
+      "loss": 1.2471,
+      "step": 2769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80194946715364e-05,
+      "loss": 1.2959,
+      "step": 2770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017904045951385e-05,
+      "loss": 1.2588,
+      "step": 2771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801631285213792e-05,
+      "loss": 1.2773,
+      "step": 2772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014721090208765e-05,
+      "loss": 1.2812,
+      "step": 2773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801312876027673e-05,
+      "loss": 1.3066,
+      "step": 2774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801153586245467e-05,
+      "loss": 1.3379,
+      "step": 2775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800994239685547e-05,
+      "loss": 1.2734,
+      "step": 2776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800834836359206e-05,
+      "loss": 1.2969,
+      "step": 2777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006753762777404e-05,
+      "loss": 1.291,
+      "step": 2778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005158594524515e-05,
+      "loss": 1.29,
+      "step": 2779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003562858946443e-05,
+      "loss": 1.2656,
+      "step": 2780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800196655615628e-05,
+      "loss": 1.3193,
+      "step": 2781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800036968626715e-05,
+      "loss": 1.2725,
+      "step": 2782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998772249392224e-05,
+      "loss": 1.252,
+      "step": 2783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997174245644715e-05,
+      "loss": 1.3086,
+      "step": 2784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995575675137872e-05,
+      "loss": 1.2988,
+      "step": 2785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799397653798499e-05,
+      "loss": 1.3564,
+      "step": 2786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992376834299394e-05,
+      "loss": 1.3369,
+      "step": 2787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7990776564194455e-05,
+      "loss": 1.2822,
+      "step": 2788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798917572778359e-05,
+      "loss": 1.3066,
+      "step": 2789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987574325180244e-05,
+      "loss": 1.3174,
+      "step": 2790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798597235649791e-05,
+      "loss": 1.291,
+      "step": 2791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984369821850124e-05,
+      "loss": 1.2852,
+      "step": 2792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982766721350446e-05,
+      "loss": 1.2734,
+      "step": 2793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79811630551125e-05,
+      "loss": 1.2705,
+      "step": 2794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797955882324993e-05,
+      "loss": 1.3203,
+      "step": 2795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977954025876437e-05,
+      "loss": 1.3174,
+      "step": 2796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797634866310574e-05,
+      "loss": 1.3535,
+      "step": 2797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797474273505162e-05,
+      "loss": 1.3076,
+      "step": 2798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973136241827887e-05,
+      "loss": 1.3047,
+      "step": 2799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971529183548392e-05,
+      "loss": 1.2979,
+      "step": 2800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796992156032703e-05,
+      "loss": 1.3223,
+      "step": 2801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796831337227773e-05,
+      "loss": 1.2998,
+      "step": 2802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966704619514463e-05,
+      "loss": 1.3252,
+      "step": 2803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965095302151246e-05,
+      "loss": 1.332,
+      "step": 2804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7963485420302127e-05,
+      "loss": 1.3193,
+      "step": 2805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961874974081198e-05,
+      "loss": 1.2822,
+      "step": 2806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960263963602598e-05,
+      "loss": 1.3242,
+      "step": 2807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958652388980494e-05,
+      "loss": 1.2959,
+      "step": 2808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957040250329095e-05,
+      "loss": 1.3135,
+      "step": 2809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795542754776266e-05,
+      "loss": 1.3184,
+      "step": 2810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953814281395478e-05,
+      "loss": 1.3262,
+      "step": 2811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952200451341883e-05,
+      "loss": 1.29,
+      "step": 2812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950586057716238e-05,
+      "loss": 1.3164,
+      "step": 2813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948971100632967e-05,
+      "loss": 1.3076,
+      "step": 2814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947355580206517e-05,
+      "loss": 1.3369,
+      "step": 2815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794573949655138e-05,
+      "loss": 1.2852,
+      "step": 2816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7944122849782088e-05,
+      "loss": 1.291,
+      "step": 2817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794250564001321e-05,
+      "loss": 1.29,
+      "step": 2818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794088786735936e-05,
+      "loss": 1.3301,
+      "step": 2819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793926953193519e-05,
+      "loss": 1.3145,
+      "step": 2820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793765063385539e-05,
+      "loss": 1.2637,
+      "step": 2821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936031173234698e-05,
+      "loss": 1.3164,
+      "step": 2822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934411150187875e-05,
+      "loss": 1.2979,
+      "step": 2823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932790564829733e-05,
+      "loss": 1.3643,
+      "step": 2824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793116941727513e-05,
+      "loss": 1.3057,
+      "step": 2825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792954770763895e-05,
+      "loss": 1.332,
+      "step": 2826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792792543603613e-05,
+      "loss": 1.3164,
+      "step": 2827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792630260258163e-05,
+      "loss": 1.3418,
+      "step": 2828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792467920739047e-05,
+      "loss": 1.3193,
+      "step": 2829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923055250577697e-05,
+      "loss": 1.3232,
+      "step": 2830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921430732258397e-05,
+      "loss": 1.2773,
+      "step": 2831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919805652547706e-05,
+      "loss": 1.2949,
+      "step": 2832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791818001156079e-05,
+      "loss": 1.2793,
+      "step": 2833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791655380941285e-05,
+      "loss": 1.2676,
+      "step": 2834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914927046219156e-05,
+      "loss": 1.3311,
+      "step": 2835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7913299722094975e-05,
+      "loss": 1.2764,
+      "step": 2836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911671837155642e-05,
+      "loss": 1.2725,
+      "step": 2837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910043391516535e-05,
+      "loss": 1.3027,
+      "step": 2838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790841438529305e-05,
+      "loss": 1.3037,
+      "step": 2839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790678481860064e-05,
+      "loss": 1.291,
+      "step": 2840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7905154691554788e-05,
+      "loss": 1.2695,
+      "step": 2841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903524004271023e-05,
+      "loss": 1.2744,
+      "step": 2842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901892756864915e-05,
+      "loss": 1.2871,
+      "step": 2843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900260949452064e-05,
+      "loss": 1.3193,
+      "step": 2844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898628582148125e-05,
+      "loss": 1.3271,
+      "step": 2845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896995655068777e-05,
+      "loss": 1.292,
+      "step": 2846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7895362168329744e-05,
+      "loss": 1.3086,
+      "step": 2847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78937281220468e-05,
+      "loss": 1.3076,
+      "step": 2848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789209351633574e-05,
+      "loss": 1.335,
+      "step": 2849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890458351312413e-05,
+      "loss": 1.3398,
+      "step": 2850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7888822627092705e-05,
+      "loss": 1.3369,
+      "step": 2851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887186343792535e-05,
+      "loss": 1.293,
+      "step": 2852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788554950152787e-05,
+      "loss": 1.3418,
+      "step": 2853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883912100414712e-05,
+      "loss": 1.3252,
+      "step": 2854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882274140569104e-05,
+      "loss": 1.2734,
+      "step": 2855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880635622107124e-05,
+      "loss": 1.3027,
+      "step": 2856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7878996545144898e-05,
+      "loss": 1.3154,
+      "step": 2857
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787735690979859e-05,
+      "loss": 1.2793,
+      "step": 2858
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787571671618439e-05,
+      "loss": 1.2822,
+      "step": 2859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787407596441855e-05,
+      "loss": 1.3359,
+      "step": 2860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872434654617346e-05,
+      "loss": 1.3242,
+      "step": 2861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870792786897098e-05,
+      "loss": 1.3203,
+      "step": 2862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869150361374164e-05,
+      "loss": 1.293,
+      "step": 2863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7867507378164938e-05,
+      "loss": 1.3105,
+      "step": 2864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786586383738587e-05,
+      "loss": 1.3096,
+      "step": 2865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864219739153424e-05,
+      "loss": 1.2988,
+      "step": 2866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786257508358413e-05,
+      "loss": 1.3291,
+      "step": 2867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860929870794535e-05,
+      "loss": 1.3262,
+      "step": 2868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859284100901238e-05,
+      "loss": 1.3105,
+      "step": 2869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7857637774020873e-05,
+      "loss": 1.292,
+      "step": 2870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7855990890270123e-05,
+      "loss": 1.335,
+      "step": 2871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7854343449765693e-05,
+      "loss": 1.2832,
+      "step": 2872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852695452624343e-05,
+      "loss": 1.3584,
+      "step": 2873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851046898962865e-05,
+      "loss": 1.3105,
+      "step": 2874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784939778889809e-05,
+      "loss": 1.2695,
+      "step": 2875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784774812254689e-05,
+      "loss": 1.292,
+      "step": 2876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846097900026178e-05,
+      "loss": 1.2998,
+      "step": 2877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844447121452907e-05,
+      "loss": 1.3057,
+      "step": 2878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784279578694407e-05,
+      "loss": 1.2939,
+      "step": 2879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841143896616686e-05,
+      "loss": 1.2598,
+      "step": 2880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839491450587836e-05,
+      "loss": 1.3213,
+      "step": 2881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837838448974626e-05,
+      "loss": 1.3086,
+      "step": 2882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78361848918942e-05,
+      "loss": 1.3115,
+      "step": 2883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7834530779463745e-05,
+      "loss": 1.3174,
+      "step": 2884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832876111800493e-05,
+      "loss": 1.2686,
+      "step": 2885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7831220889021713e-05,
+      "loss": 1.292,
+      "step": 2886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78295651112447e-05,
+      "loss": 1.2842,
+      "step": 2887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7827908778586808e-05,
+      "loss": 1.2803,
+      "step": 2888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7826251891165415e-05,
+      "loss": 1.3164,
+      "step": 2889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782459444909795e-05,
+      "loss": 1.3359,
+      "step": 2890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782293645250187e-05,
+      "loss": 1.3281,
+      "step": 2891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821277901494678e-05,
+      "loss": 1.3135,
+      "step": 2892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7819618796193922e-05,
+      "loss": 1.3203,
+      "step": 2893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817959136717178e-05,
+      "loss": 1.3154,
+      "step": 2894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7816298923182065e-05,
+      "loss": 1.3252,
+      "step": 2895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814638155706243e-05,
+      "loss": 1.3252,
+      "step": 2896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812976834407414e-05,
+      "loss": 1.3223,
+      "step": 2897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781131495940331e-05,
+      "loss": 1.3467,
+      "step": 2898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780965253081171e-05,
+      "loss": 1.3379,
+      "step": 2899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807989548750437e-05,
+      "loss": 1.2871,
+      "step": 2900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806326013337334e-05,
+      "loss": 1.2793,
+      "step": 2901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804661924690303e-05,
+      "loss": 1.3086,
+      "step": 2902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780299728292728e-05,
+      "loss": 1.2549,
+      "step": 2903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801332088166236e-05,
+      "loss": 1.3623,
+      "step": 2904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799666340525177e-05,
+      "loss": 1.3457,
+      "step": 2905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798000040122163e-05,
+      "loss": 1.293,
+      "step": 2906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779633318707528e-05,
+      "loss": 1.3379,
+      "step": 2907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794665781502662e-05,
+      "loss": 1.3203,
+      "step": 2908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7792997823522476e-05,
+      "loss": 1.3135,
+      "step": 2909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791329313252926e-05,
+      "loss": 1.3232,
+      "step": 2910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789660250812267e-05,
+      "loss": 1.3281,
+      "step": 2911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7787990636318776e-05,
+      "loss": 1.3193,
+      "step": 2912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786320469890786e-05,
+      "loss": 1.333,
+      "step": 2913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7784649751646656e-05,
+      "loss": 1.3066,
+      "step": 2914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782978481704797e-05,
+      "loss": 1.2598,
+      "step": 2915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781306660183644e-05,
+      "loss": 1.3037,
+      "step": 2916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779634287201687e-05,
+      "loss": 1.3154,
+      "step": 2917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777961362877437e-05,
+      "loss": 1.3164,
+      "step": 2918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777628788732946e-05,
+      "loss": 1.3105,
+      "step": 2919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777461386067636e-05,
+      "loss": 1.3184,
+      "step": 2920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7772939283036765e-05,
+      "loss": 1.3486,
+      "step": 2921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777126415452936e-05,
+      "loss": 1.3184,
+      "step": 2922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769588475272854e-05,
+      "loss": 1.2959,
+      "step": 2923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776791224538601e-05,
+      "loss": 1.2861,
+      "step": 2924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776623546498762e-05,
+      "loss": 1.333,
+      "step": 2925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764558134196515e-05,
+      "loss": 1.3184,
+      "step": 2926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762880253131565e-05,
+      "loss": 1.292,
+      "step": 2927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776120182191169e-05,
+      "loss": 1.293,
+      "step": 2928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759522840655835e-05,
+      "loss": 1.3076,
+      "step": 2929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757843309482988e-05,
+      "loss": 1.3408,
+      "step": 2930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775616322851218e-05,
+      "loss": 1.3027,
+      "step": 2931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754482597862476e-05,
+      "loss": 1.3076,
+      "step": 2932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752801417652986e-05,
+      "loss": 1.3037,
+      "step": 2933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775111968800285e-05,
+      "loss": 1.3135,
+      "step": 2934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749437409031256e-05,
+      "loss": 1.2803,
+      "step": 2935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7747754580857425e-05,
+      "loss": 1.334,
+      "step": 2936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774607120360062e-05,
+      "loss": 1.2783,
+      "step": 2937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774438727738014e-05,
+      "loss": 1.3193,
+      "step": 2938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742702802315328e-05,
+      "loss": 1.3936,
+      "step": 2939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741017778525558e-05,
+      "loss": 1.3379,
+      "step": 2940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7739332206130254e-05,
+      "loss": 1.3291,
+      "step": 2941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773764608524887e-05,
+      "loss": 1.3008,
+      "step": 2942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735959416000895e-05,
+      "loss": 1.2686,
+      "step": 2943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734272198505872e-05,
+      "loss": 1.3232,
+      "step": 2944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773258443288337e-05,
+      "loss": 1.3428,
+      "step": 2945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7730896119253002e-05,
+      "loss": 1.291,
+      "step": 2946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729207257734415e-05,
+      "loss": 1.2822,
+      "step": 2947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77275178484473e-05,
+      "loss": 1.334,
+      "step": 2948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725827891511392e-05,
+      "loss": 1.3096,
+      "step": 2949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772413738704645e-05,
+      "loss": 1.2764,
+      "step": 2950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722446335172286e-05,
+      "loss": 1.2979,
+      "step": 2951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772075473600874e-05,
+      "loss": 1.2822,
+      "step": 2952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77190625896757e-05,
+      "loss": 1.3105,
+      "step": 2953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771736989629308e-05,
+      "loss": 1.3242,
+      "step": 2954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771567665598085e-05,
+      "loss": 1.333,
+      "step": 2955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713982868859005e-05,
+      "loss": 1.3262,
+      "step": 2956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712288535047586e-05,
+      "loss": 1.2451,
+      "step": 2957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771059365466667e-05,
+      "loss": 1.3154,
+      "step": 2958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708898227836373e-05,
+      "loss": 1.29,
+      "step": 2959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707202254676845e-05,
+      "loss": 1.3115,
+      "step": 2960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705505735308287e-05,
+      "loss": 1.2959,
+      "step": 2961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770380866985093e-05,
+      "loss": 1.3154,
+      "step": 2962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770211105842504e-05,
+      "loss": 1.3164,
+      "step": 2963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700412901150933e-05,
+      "loss": 1.3086,
+      "step": 2964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698714198148954e-05,
+      "loss": 1.2822,
+      "step": 2965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697014949539487e-05,
+      "loss": 1.2734,
+      "step": 2966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695315155442964e-05,
+      "loss": 1.2949,
+      "step": 2967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693614815979846e-05,
+      "loss": 1.2891,
+      "step": 2968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691913931270633e-05,
+      "loss": 1.3066,
+      "step": 2969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690212501435873e-05,
+      "loss": 1.2793,
+      "step": 2970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688510526596137e-05,
+      "loss": 1.2822,
+      "step": 2971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686808006872053e-05,
+      "loss": 1.2998,
+      "step": 2972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768510494238428e-05,
+      "loss": 1.2871,
+      "step": 2973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683401333253502e-05,
+      "loss": 1.2861,
+      "step": 2974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681697179600464e-05,
+      "loss": 1.293,
+      "step": 2975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767999248154594e-05,
+      "loss": 1.3545,
+      "step": 2976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678287239210735e-05,
+      "loss": 1.3213,
+      "step": 2977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76765814527157e-05,
+      "loss": 1.3252,
+      "step": 2978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767487512218173e-05,
+      "loss": 1.2783,
+      "step": 2979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767316824772975e-05,
+      "loss": 1.333,
+      "step": 2980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767146082948072e-05,
+      "loss": 1.291,
+      "step": 2981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669752867555658e-05,
+      "loss": 1.2959,
+      "step": 2982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766804436207559e-05,
+      "loss": 1.2793,
+      "step": 2983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666335313161612e-05,
+      "loss": 1.3086,
+      "step": 2984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766462572093484e-05,
+      "loss": 1.3271,
+      "step": 2985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7662915585516432e-05,
+      "loss": 1.3164,
+      "step": 2986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766120490702758e-05,
+      "loss": 1.3203,
+      "step": 2987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659493685589528e-05,
+      "loss": 1.3086,
+      "step": 2988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657781921323548e-05,
+      "loss": 1.2998,
+      "step": 2989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656069614350947e-05,
+      "loss": 1.375,
+      "step": 2990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654356764793083e-05,
+      "loss": 1.3262,
+      "step": 2991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652643372771345e-05,
+      "loss": 1.332,
+      "step": 2992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650929438407157e-05,
+      "loss": 1.333,
+      "step": 2993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649214961821987e-05,
+      "loss": 1.2764,
+      "step": 2994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647499943137343e-05,
+      "loss": 1.3564,
+      "step": 2995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645784382474765e-05,
+      "loss": 1.2979,
+      "step": 2996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644068279955836e-05,
+      "loss": 1.252,
+      "step": 2997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642351635702172e-05,
+      "loss": 1.2988,
+      "step": 2998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764063444983544e-05,
+      "loss": 1.2715,
+      "step": 2999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638916722477327e-05,
+      "loss": 1.2539,
+      "step": 3000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763719845374958e-05,
+      "loss": 1.3281,
+      "step": 3001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635479643773958e-05,
+      "loss": 1.2891,
+      "step": 3002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633760292672283e-05,
+      "loss": 1.3154,
+      "step": 3003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7632040400566403e-05,
+      "loss": 1.3193,
+      "step": 3004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763031996757821e-05,
+      "loss": 1.3213,
+      "step": 3005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628598993829626e-05,
+      "loss": 1.3555,
+      "step": 3006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626877479442616e-05,
+      "loss": 1.2939,
+      "step": 3007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762515542453919e-05,
+      "loss": 1.3174,
+      "step": 3008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762343282924138e-05,
+      "loss": 1.2949,
+      "step": 3009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7621709693671275e-05,
+      "loss": 1.3105,
+      "step": 3010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761998601795099e-05,
+      "loss": 1.2676,
+      "step": 3011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618261802202682e-05,
+      "loss": 1.2588,
+      "step": 3012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616537046548544e-05,
+      "loss": 1.3057,
+      "step": 3013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614811751110816e-05,
+      "loss": 1.3574,
+      "step": 3014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761308591601176e-05,
+      "loss": 1.2676,
+      "step": 3015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76113595413737e-05,
+      "loss": 1.2803,
+      "step": 3016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609632627318966e-05,
+      "loss": 1.2979,
+      "step": 3017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760790517396996e-05,
+      "loss": 1.3281,
+      "step": 3018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76061771814491e-05,
+      "loss": 1.3555,
+      "step": 3019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760444864987884e-05,
+      "loss": 1.2783,
+      "step": 3020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76027195793817e-05,
+      "loss": 1.2666,
+      "step": 3021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600989970080203e-05,
+      "loss": 1.2812,
+      "step": 3022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759925982209693e-05,
+      "loss": 1.2783,
+      "step": 3023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7597529135554502e-05,
+      "loss": 1.3369,
+      "step": 3024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7595797910575574e-05,
+      "loss": 1.3428,
+      "step": 3025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594066147282826e-05,
+      "loss": 1.293,
+      "step": 3026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592333845798998e-05,
+      "loss": 1.2627,
+      "step": 3027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590601006246854e-05,
+      "loss": 1.2666,
+      "step": 3028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588867628749206e-05,
+      "loss": 1.3232,
+      "step": 3029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758713371342889e-05,
+      "loss": 1.2588,
+      "step": 3030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758539926040879e-05,
+      "loss": 1.293,
+      "step": 3031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583664269811833e-05,
+      "loss": 1.3154,
+      "step": 3032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758192874176097e-05,
+      "loss": 1.3125,
+      "step": 3033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758019267637921e-05,
+      "loss": 1.2607,
+      "step": 3034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757845607378957e-05,
+      "loss": 1.3027,
+      "step": 3035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576718934115137e-05,
+      "loss": 1.2852,
+      "step": 3036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574981257479017e-05,
+      "loss": 1.3154,
+      "step": 3037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573243044004355e-05,
+      "loss": 1.334,
+      "step": 3038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571504293814344e-05,
+      "loss": 1.332,
+      "step": 3039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756976500703221e-05,
+      "loss": 1.2959,
+      "step": 3040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568025183781213e-05,
+      "loss": 1.3027,
+      "step": 3041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566284824184655e-05,
+      "loss": 1.2979,
+      "step": 3042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564543928365876e-05,
+      "loss": 1.2871,
+      "step": 3043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756280249644825e-05,
+      "loss": 1.2715,
+      "step": 3044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561060528555197e-05,
+      "loss": 1.3428,
+      "step": 3045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7559318024810163e-05,
+      "loss": 1.3486,
+      "step": 3046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557574985336648e-05,
+      "loss": 1.2627,
+      "step": 3047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555831410258176e-05,
+      "loss": 1.3125,
+      "step": 3048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554087299698315e-05,
+      "loss": 1.3213,
+      "step": 3049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755234265378067e-05,
+      "loss": 1.2871,
+      "step": 3050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550597472628883e-05,
+      "loss": 1.3203,
+      "step": 3051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548851756366635e-05,
+      "loss": 1.2979,
+      "step": 3052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754710550511765e-05,
+      "loss": 1.2354,
+      "step": 3053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545358719005677e-05,
+      "loss": 1.2969,
+      "step": 3054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543611398154514e-05,
+      "loss": 1.3066,
+      "step": 3055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7541863542687996e-05,
+      "loss": 1.3232,
+      "step": 3056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540115152729987e-05,
+      "loss": 1.2979,
+      "step": 3057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538366228404404e-05,
+      "loss": 1.2949,
+      "step": 3058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536616769835186e-05,
+      "loss": 1.3115,
+      "step": 3059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753486677714632e-05,
+      "loss": 1.3008,
+      "step": 3060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753311625046183e-05,
+      "loss": 1.2676,
+      "step": 3061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753136518990577e-05,
+      "loss": 1.2959,
+      "step": 3062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752961359560224e-05,
+      "loss": 1.3369,
+      "step": 3063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527861467675376e-05,
+      "loss": 1.3047,
+      "step": 3064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7526108806249354e-05,
+      "loss": 1.2764,
+      "step": 3065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524355611448383e-05,
+      "loss": 1.2773,
+      "step": 3066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522601883396712e-05,
+      "loss": 1.2871,
+      "step": 3067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520847622218624e-05,
+      "loss": 1.2881,
+      "step": 3068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7519092828038448e-05,
+      "loss": 1.2197,
+      "step": 3069
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517337500980544e-05,
+      "loss": 1.3115,
+      "step": 3070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751558164116931e-05,
+      "loss": 1.3096,
+      "step": 3071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513825248729192e-05,
+      "loss": 1.2842,
+      "step": 3072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512068323784653e-05,
+      "loss": 1.3174,
+      "step": 3073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510310866460215e-05,
+      "loss": 1.2852,
+      "step": 3074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508552876880426e-05,
+      "loss": 1.2676,
+      "step": 3075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750679435516988e-05,
+      "loss": 1.333,
+      "step": 3076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7505035301453193e-05,
+      "loss": 1.2549,
+      "step": 3077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503275715855036e-05,
+      "loss": 1.3027,
+      "step": 3078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750151559850011e-05,
+      "loss": 1.3184,
+      "step": 3079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499754949513152e-05,
+      "loss": 1.2461,
+      "step": 3080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497993769018944e-05,
+      "loss": 1.2773,
+      "step": 3081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7496232057142297e-05,
+      "loss": 1.292,
+      "step": 3082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7494469814008063e-05,
+      "loss": 1.3076,
+      "step": 3083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492707039741134e-05,
+      "loss": 1.3115,
+      "step": 3084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749094373446644e-05,
+      "loss": 1.2715,
+      "step": 3085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489179898308938e-05,
+      "loss": 1.2773,
+      "step": 3086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487415531393637e-05,
+      "loss": 1.3203,
+      "step": 3087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748565063384558e-05,
+      "loss": 1.2861,
+      "step": 3088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748388520578984e-05,
+      "loss": 1.3203,
+      "step": 3089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482119247351536e-05,
+      "loss": 1.3398,
+      "step": 3090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480352758655818e-05,
+      "loss": 1.3018,
+      "step": 3091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747858573982788e-05,
+      "loss": 1.2822,
+      "step": 3092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747681819099295e-05,
+      "loss": 1.2939,
+      "step": 3093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475050112276295e-05,
+      "loss": 1.2627,
+      "step": 3094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473281503803215e-05,
+      "loss": 1.2715,
+      "step": 3095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7471512365699058e-05,
+      "loss": 1.3398,
+      "step": 3096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469742698089195e-05,
+      "loss": 1.3174,
+      "step": 3097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746797250109905e-05,
+      "loss": 1.3164,
+      "step": 3098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466201774854066e-05,
+      "loss": 1.292,
+      "step": 3099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7464430519479744e-05,
+      "loss": 1.3535,
+      "step": 3100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746265873510161e-05,
+      "loss": 1.292,
+      "step": 3101
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7460886421845226e-05,
+      "loss": 1.3154,
+      "step": 3102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7459113579836203e-05,
+      "loss": 1.3193,
+      "step": 3103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745734020920018e-05,
+      "loss": 1.3018,
+      "step": 3104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745556631006283e-05,
+      "loss": 1.2627,
+      "step": 3105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453791882549876e-05,
+      "loss": 1.291,
+      "step": 3106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745201692678707e-05,
+      "loss": 1.2764,
+      "step": 3107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450241442900206e-05,
+      "loss": 1.2793,
+      "step": 3108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74484654310151e-05,
+      "loss": 1.3135,
+      "step": 3109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446688891257632e-05,
+      "loss": 1.2568,
+      "step": 3110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7444911823753697e-05,
+      "loss": 1.2998,
+      "step": 3111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744313422862924e-05,
+      "loss": 1.2803,
+      "step": 3112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441356106010237e-05,
+      "loss": 1.3057,
+      "step": 3113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439577456022707e-05,
+      "loss": 1.2783,
+      "step": 3114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437798278792698e-05,
+      "loss": 1.3213,
+      "step": 3115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74360185744463e-05,
+      "loss": 1.3037,
+      "step": 3116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434238343109646e-05,
+      "loss": 1.3096,
+      "step": 3117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432457584908893e-05,
+      "loss": 1.3203,
+      "step": 3118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743067629997025e-05,
+      "loss": 1.3018,
+      "step": 3119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7428894488419955e-05,
+      "loss": 1.2598,
+      "step": 3120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427112150384283e-05,
+      "loss": 1.2734,
+      "step": 3121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742532928598955e-05,
+      "loss": 1.3135,
+      "step": 3122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742354589536211e-05,
+      "loss": 1.3066,
+      "step": 3123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421761978628345e-05,
+      "loss": 1.3066,
+      "step": 3124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7419977535914686e-05,
+      "loss": 1.292,
+      "step": 3125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7418192567347595e-05,
+      "loss": 1.2275,
+      "step": 3126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741640707305357e-05,
+      "loss": 1.3262,
+      "step": 3127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414621053159156e-05,
+      "loss": 1.3467,
+      "step": 3128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7412834507790922e-05,
+      "loss": 1.3105,
+      "step": 3129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411047437075484e-05,
+      "loss": 1.292,
+      "step": 3130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740925984113949e-05,
+      "loss": 1.2959,
+      "step": 3131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407471720109624e-05,
+      "loss": 1.29,
+      "step": 3132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740568307411261e-05,
+      "loss": 1.2979,
+      "step": 3133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740389390327522e-05,
+      "loss": 1.3232,
+      "step": 3134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740210420772424e-05,
+      "loss": 1.3291,
+      "step": 3135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7400313987586507e-05,
+      "loss": 1.335,
+      "step": 3136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73985232429889e-05,
+      "loss": 1.2588,
+      "step": 3137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739673197405833e-05,
+      "loss": 1.3086,
+      "step": 3138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394940180921733e-05,
+      "loss": 1.3018,
+      "step": 3139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393147863706103e-05,
+      "loss": 1.3184,
+      "step": 3140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739135502253846e-05,
+      "loss": 1.3555,
+      "step": 3141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738956165754586e-05,
+      "loss": 1.3213,
+      "step": 3142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73877677688554e-05,
+      "loss": 1.3105,
+      "step": 3143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385973356594212e-05,
+      "loss": 1.3203,
+      "step": 3144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7384178420889467e-05,
+      "loss": 1.3193,
+      "step": 3145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382382961868374e-05,
+      "loss": 1.3076,
+      "step": 3146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380586979658174e-05,
+      "loss": 1.3076,
+      "step": 3147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378790474386148e-05,
+      "loss": 1.293,
+      "step": 3148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376993446179613e-05,
+      "loss": 1.2627,
+      "step": 3149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737519589516593e-05,
+      "loss": 1.292,
+      "step": 3150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373397821472485e-05,
+      "loss": 1.3115,
+      "step": 3151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371599225226713e-05,
+      "loss": 1.3252,
+      "step": 3152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369800106556075e-05,
+      "loss": 1.2705,
+      "step": 3153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736800046558808e-05,
+      "loss": 1.332,
+      "step": 3154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736620030245026e-05,
+      "loss": 1.3057,
+      "step": 3155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364399617270202e-05,
+      "loss": 1.3027,
+      "step": 3156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736259841017552e-05,
+      "loss": 1.2939,
+      "step": 3157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360796681293855e-05,
+      "loss": 1.3174,
+      "step": 3158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7358994430752908e-05,
+      "loss": 1.3389,
+      "step": 3159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357191658680395e-05,
+      "loss": 1.2969,
+      "step": 3160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7355388365204083e-05,
+      "loss": 1.2656,
+      "step": 3161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735358455045177e-05,
+      "loss": 1.2842,
+      "step": 3162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735178021455129e-05,
+      "loss": 1.2979,
+      "step": 3163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349975357630527e-05,
+      "loss": 1.2949,
+      "step": 3164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734816997981738e-05,
+      "loss": 1.3311,
+      "step": 3165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346364081239796e-05,
+      "loss": 1.293,
+      "step": 3166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734455766202576e-05,
+      "loss": 1.2588,
+      "step": 3167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73427507223033e-05,
+      "loss": 1.2686,
+      "step": 3168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340943262200468e-05,
+      "loss": 1.3115,
+      "step": 3169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339135281845355e-05,
+      "loss": 1.2666,
+      "step": 3170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337326781366103e-05,
+      "loss": 1.2842,
+      "step": 3171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733551776089087e-05,
+      "loss": 1.2734,
+      "step": 3172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7333708220547865e-05,
+      "loss": 1.2842,
+      "step": 3173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733189816046533e-05,
+      "loss": 1.3047,
+      "step": 3174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7330087580771545e-05,
+      "loss": 1.2764,
+      "step": 3175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328276481594825e-05,
+      "loss": 1.3164,
+      "step": 3176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326464863063523e-05,
+      "loss": 1.2793,
+      "step": 3177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324652725306024e-05,
+      "loss": 1.3457,
+      "step": 3178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732284006845076e-05,
+      "loss": 1.3135,
+      "step": 3179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321026892626193e-05,
+      "loss": 1.29,
+      "step": 3180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731921319796082e-05,
+      "loss": 1.3066,
+      "step": 3181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731739898458318e-05,
+      "loss": 1.3135,
+      "step": 3182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315584252621844e-05,
+      "loss": 1.3057,
+      "step": 3183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313769002205422e-05,
+      "loss": 1.3027,
+      "step": 3184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311953233462563e-05,
+      "loss": 1.3145,
+      "step": 3185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310136946521948e-05,
+      "loss": 1.3516,
+      "step": 3186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308320141512297e-05,
+      "loss": 1.3086,
+      "step": 3187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730650281856237e-05,
+      "loss": 1.3105,
+      "step": 3188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730468497780096e-05,
+      "loss": 1.2969,
+      "step": 3189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730286661935689e-05,
+      "loss": 1.3281,
+      "step": 3190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730104774335904e-05,
+      "loss": 1.3008,
+      "step": 3191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7299228349936302e-05,
+      "loss": 1.2744,
+      "step": 3192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297408439217625e-05,
+      "loss": 1.3154,
+      "step": 3193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295588011331977e-05,
+      "loss": 1.293,
+      "step": 3194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729376706640838e-05,
+      "loss": 1.3145,
+      "step": 3195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729194560457588e-05,
+      "loss": 1.333,
+      "step": 3196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290123625963564e-05,
+      "loss": 1.2881,
+      "step": 3197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288301130700556e-05,
+      "loss": 1.2812,
+      "step": 3198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286478118916017e-05,
+      "loss": 1.2979,
+      "step": 3199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7284654590739146e-05,
+      "loss": 1.334,
+      "step": 3200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7282830546299167e-05,
+      "loss": 1.2734,
+      "step": 3201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728100598572536e-05,
+      "loss": 1.3027,
+      "step": 3202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727918090914703e-05,
+      "loss": 1.2871,
+      "step": 3203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277355316693515e-05,
+      "loss": 1.2998,
+      "step": 3204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7275529208494196e-05,
+      "loss": 1.2998,
+      "step": 3205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273702584678494e-05,
+      "loss": 1.3262,
+      "step": 3206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271875445375858e-05,
+      "loss": 1.2686,
+      "step": 3207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7270047790715777e-05,
+      "loss": 1.292,
+      "step": 3208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268219620827778e-05,
+      "loss": 1.2441,
+      "step": 3209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266390935841425e-05,
+      "loss": 1.2754,
+      "step": 3210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726456173588631e-05,
+      "loss": 1.3027,
+      "step": 3211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262732021092077e-05,
+      "loss": 1.3242,
+      "step": 3212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7260901791588395e-05,
+      "loss": 1.3369,
+      "step": 3213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7259071047504968e-05,
+      "loss": 1.291,
+      "step": 3214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257239788971544e-05,
+      "loss": 1.3135,
+      "step": 3215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7255408016117905e-05,
+      "loss": 1.2949,
+      "step": 3216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253575729073867e-05,
+      "loss": 1.2725,
+      "step": 3217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7251742927969288e-05,
+      "loss": 1.2822,
+      "step": 3218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724990961293405e-05,
+      "loss": 1.2959,
+      "step": 3219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7248075784098088e-05,
+      "loss": 1.2451,
+      "step": 3220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7246241441591364e-05,
+      "loss": 1.2803,
+      "step": 3221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244406585543875e-05,
+      "loss": 1.3516,
+      "step": 3222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242571216085658e-05,
+      "loss": 1.2812,
+      "step": 3223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240735333346786e-05,
+      "loss": 1.2998,
+      "step": 3224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723889893745737e-05,
+      "loss": 1.2959,
+      "step": 3225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723706202854755e-05,
+      "loss": 1.3135,
+      "step": 3226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235224606747514e-05,
+      "loss": 1.3262,
+      "step": 3227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233386672187476e-05,
+      "loss": 1.2734,
+      "step": 3228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723154822499769e-05,
+      "loss": 1.2793,
+      "step": 3229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229709265308446e-05,
+      "loss": 1.2852,
+      "step": 3230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722786979325008e-05,
+      "loss": 1.2852,
+      "step": 3231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722602980895294e-05,
+      "loss": 1.3223,
+      "step": 3232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722418931254744e-05,
+      "loss": 1.3164,
+      "step": 3233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222348304164006e-05,
+      "loss": 1.3008,
+      "step": 3234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220506783933118e-05,
+      "loss": 1.3154,
+      "step": 3235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218664751985276e-05,
+      "loss": 1.2861,
+      "step": 3236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216822208451028e-05,
+      "loss": 1.3066,
+      "step": 3237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214979153460963e-05,
+      "loss": 1.2764,
+      "step": 3238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213135587145684e-05,
+      "loss": 1.3311,
+      "step": 3239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211291509635856e-05,
+      "loss": 1.3066,
+      "step": 3240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720944692106216e-05,
+      "loss": 1.2998,
+      "step": 3241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207601821555328e-05,
+      "loss": 1.2852,
+      "step": 3242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720575621124612e-05,
+      "loss": 1.291,
+      "step": 3243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203910090265337e-05,
+      "loss": 1.248,
+      "step": 3244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720206345874381e-05,
+      "loss": 1.3066,
+      "step": 3245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7200216316812412e-05,
+      "loss": 1.3281,
+      "step": 3246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198368664602044e-05,
+      "loss": 1.3018,
+      "step": 3247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7196520502243658e-05,
+      "loss": 1.3018,
+      "step": 3248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7194671829868232e-05,
+      "loss": 1.3203,
+      "step": 3249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719282264760677e-05,
+      "loss": 1.2725,
+      "step": 3250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190972955590335e-05,
+      "loss": 1.2803,
+      "step": 3251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718912275395001e-05,
+      "loss": 1.3027,
+      "step": 3252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187272042816923e-05,
+      "loss": 1.292,
+      "step": 3253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718542082232223e-05,
+      "loss": 1.2891,
+      "step": 3254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183569092597125e-05,
+      "loss": 1.2617,
+      "step": 3255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718171685377285e-05,
+      "loss": 1.3359,
+      "step": 3256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179864105980657e-05,
+      "loss": 1.2812,
+      "step": 3257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717801084935186e-05,
+      "loss": 1.2666,
+      "step": 3258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71761570840178e-05,
+      "loss": 1.2617,
+      "step": 3259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717430281010985e-05,
+      "loss": 1.248,
+      "step": 3260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172448027759428e-05,
+      "loss": 1.3721,
+      "step": 3261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170592737097974e-05,
+      "loss": 1.3447,
+      "step": 3262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168736938256975e-05,
+      "loss": 1.2676,
+      "step": 3263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166880631367956e-05,
+      "loss": 1.3164,
+      "step": 3264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165023816562467e-05,
+      "loss": 1.2861,
+      "step": 3265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7163166493972104e-05,
+      "loss": 1.2705,
+      "step": 3266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71613086637285e-05,
+      "loss": 1.3018,
+      "step": 3267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159450325963303e-05,
+      "loss": 1.3008,
+      "step": 3268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157591480808232e-05,
+      "loss": 1.3164,
+      "step": 3269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155732128395016e-05,
+      "loss": 1.2793,
+      "step": 3270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153872268855425e-05,
+      "loss": 1.2676,
+      "step": 3271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715201190232127e-05,
+      "loss": 1.3037,
+      "step": 3272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150151028924393e-05,
+      "loss": 1.3662,
+      "step": 3273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7148289648796673e-05,
+      "loss": 1.252,
+      "step": 3274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146427762070033e-05,
+      "loss": 1.2705,
+      "step": 3275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144565368876417e-05,
+      "loss": 1.2686,
+      "step": 3276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142702469347813e-05,
+      "loss": 1.3252,
+      "step": 3277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714083906361625e-05,
+      "loss": 1.2607,
+      "step": 3278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138975151813784e-05,
+      "loss": 1.3154,
+      "step": 3279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7137110734072514e-05,
+      "loss": 1.2979,
+      "step": 3280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135245810524565e-05,
+      "loss": 1.3076,
+      "step": 3281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133380381302106e-05,
+      "loss": 1.3213,
+      "step": 3282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131514446537344e-05,
+      "loss": 1.2471,
+      "step": 3283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7129648006362512e-05,
+      "loss": 1.3164,
+      "step": 3284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712778106090989e-05,
+      "loss": 1.332,
+      "step": 3285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125913610311782e-05,
+      "loss": 1.334,
+      "step": 3286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124045654700537e-05,
+      "loss": 1.3008,
+      "step": 3287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712217719420854e-05,
+      "loss": 1.2852,
+      "step": 3288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712030822896821e-05,
+      "loss": 1.3242,
+      "step": 3289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118438759111994e-05,
+      "loss": 1.29,
+      "step": 3290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116568784772384e-05,
+      "loss": 1.3125,
+      "step": 3291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114698306081903e-05,
+      "loss": 1.3271,
+      "step": 3292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7112827323173116e-05,
+      "loss": 1.3291,
+      "step": 3293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110955836178616e-05,
+      "loss": 1.2754,
+      "step": 3294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109083845231036e-05,
+      "loss": 1.3076,
+      "step": 3295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7107211350463044e-05,
+      "loss": 1.3027,
+      "step": 3296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7105338352007344e-05,
+      "loss": 1.2988,
+      "step": 3297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710346484999668e-05,
+      "loss": 1.2812,
+      "step": 3298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101590844563814e-05,
+      "loss": 1.3379,
+      "step": 3299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099716335841573e-05,
+      "loss": 1.2793,
+      "step": 3300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709784132396279e-05,
+      "loss": 1.3105,
+      "step": 3301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095965809060353e-05,
+      "loss": 1.29,
+      "step": 3302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094089791267177e-05,
+      "loss": 1.2754,
+      "step": 3303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092213270716224e-05,
+      "loss": 1.3398,
+      "step": 3304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709033624754047e-05,
+      "loss": 1.3047,
+      "step": 3305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7088458721872946e-05,
+      "loss": 1.3018,
+      "step": 3306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086580693846712e-05,
+      "loss": 1.3193,
+      "step": 3307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084702163594867e-05,
+      "loss": 1.3232,
+      "step": 3308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082823131250535e-05,
+      "loss": 1.2549,
+      "step": 3309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708094359694689e-05,
+      "loss": 1.293,
+      "step": 3310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079063560817126e-05,
+      "loss": 1.2969,
+      "step": 3311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707718302299449e-05,
+      "loss": 1.2451,
+      "step": 3312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707530198361225e-05,
+      "loss": 1.291,
+      "step": 3313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7073420442803722e-05,
+      "loss": 1.3232,
+      "step": 3314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071538400702242e-05,
+      "loss": 1.293,
+      "step": 3315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706965585744119e-05,
+      "loss": 1.3047,
+      "step": 3316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7067772813153994e-05,
+      "loss": 1.2803,
+      "step": 3317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706588926797409e-05,
+      "loss": 1.2734,
+      "step": 3318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064005222034977e-05,
+      "loss": 1.3389,
+      "step": 3319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706212067547017e-05,
+      "loss": 1.2939,
+      "step": 3320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706023562841323e-05,
+      "loss": 1.2871,
+      "step": 3321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7058350080997745e-05,
+      "loss": 1.3467,
+      "step": 3322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056464033357353e-05,
+      "loss": 1.2832,
+      "step": 3323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7054577485625708e-05,
+      "loss": 1.2861,
+      "step": 3324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7052690437936518e-05,
+      "loss": 1.2803,
+      "step": 3325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050802890423513e-05,
+      "loss": 1.3193,
+      "step": 3326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048914843220462e-05,
+      "loss": 1.3105,
+      "step": 3327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047026296461175e-05,
+      "loss": 1.2949,
+      "step": 3328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045137250279495e-05,
+      "loss": 1.3369,
+      "step": 3329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704324770480929e-05,
+      "loss": 1.2812,
+      "step": 3330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7041357660184477e-05,
+      "loss": 1.2744,
+      "step": 3331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703946711653901e-05,
+      "loss": 1.2871,
+      "step": 3332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703757607400686e-05,
+      "loss": 1.2812,
+      "step": 3333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035684532722053e-05,
+      "loss": 1.3223,
+      "step": 3334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7033792492818637e-05,
+      "loss": 1.2725,
+      "step": 3335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7031899954430707e-05,
+      "loss": 1.2988,
+      "step": 3336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030006917692383e-05,
+      "loss": 1.3037,
+      "step": 3337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028113382737823e-05,
+      "loss": 1.2881,
+      "step": 3338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026219349701227e-05,
+      "loss": 1.3057,
+      "step": 3339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7024324818716817e-05,
+      "loss": 1.291,
+      "step": 3340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702242978991887e-05,
+      "loss": 1.2988,
+      "step": 3341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020534263441674e-05,
+      "loss": 1.332,
+      "step": 3342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018638239419572e-05,
+      "loss": 1.2656,
+      "step": 3343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7016741717986934e-05,
+      "loss": 1.2803,
+      "step": 3344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014844699278165e-05,
+      "loss": 1.332,
+      "step": 3345
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012947183427712e-05,
+      "loss": 1.3203,
+      "step": 3346
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701104917057004e-05,
+      "loss": 1.2852,
+      "step": 3347
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700915066083968e-05,
+      "loss": 1.3105,
+      "step": 3348
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700725165437116e-05,
+      "loss": 1.2646,
+      "step": 3349
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700535215129907e-05,
+      "loss": 1.2783,
+      "step": 3350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700345215175803e-05,
+      "loss": 1.3115,
+      "step": 3351
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700155165588269e-05,
+      "loss": 1.2559,
+      "step": 3352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6999650663807743e-05,
+      "loss": 1.2676,
+      "step": 3353
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6997749175667907e-05,
+      "loss": 1.3086,
+      "step": 3354
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6995847191597937e-05,
+      "loss": 1.2783,
+      "step": 3355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993944711732635e-05,
+      "loss": 1.3037,
+      "step": 3356
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699204173620683e-05,
+      "loss": 1.2842,
+      "step": 3357
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6990138265155372e-05,
+      "loss": 1.3086,
+      "step": 3358
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6988234298713176e-05,
+      "loss": 1.292,
+      "step": 3359
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6986329837015166e-05,
+      "loss": 1.292,
+      "step": 3360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698442488019632e-05,
+      "loss": 1.2832,
+      "step": 3361
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982519428391634e-05,
+      "loss": 1.2773,
+      "step": 3362
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698061348173615e-05,
+      "loss": 1.2744,
+      "step": 3363
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697870704036494e-05,
+      "loss": 1.2822,
+      "step": 3364
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697680010441312e-05,
+      "loss": 1.2705,
+      "step": 3365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697489267401583e-05,
+      "loss": 1.3008,
+      "step": 3366
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6972984749308247e-05,
+      "loss": 1.2861,
+      "step": 3367
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697107633042559e-05,
+      "loss": 1.2969,
+      "step": 3368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969167417503105e-05,
+      "loss": 1.2646,
+      "step": 3369
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696725801067608e-05,
+      "loss": 1.2861,
+      "step": 3370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965348110079832e-05,
+      "loss": 1.3115,
+      "step": 3371
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6963437715849718e-05,
+      "loss": 1.3125,
+      "step": 3372
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961526828121125e-05,
+      "loss": 1.2676,
+      "step": 3373
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695961544702948e-05,
+      "loss": 1.2646,
+      "step": 3374
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695770357271024e-05,
+      "loss": 1.2646,
+      "step": 3375
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955791205298896e-05,
+      "loss": 1.249,
+      "step": 3376
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695387834493099e-05,
+      "loss": 1.2832,
+      "step": 3377
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6951964991742073e-05,
+      "loss": 1.3125,
+      "step": 3378
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950051145867752e-05,
+      "loss": 1.29,
+      "step": 3379
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948136807443657e-05,
+      "loss": 1.3203,
+      "step": 3380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946221976605462e-05,
+      "loss": 1.3057,
+      "step": 3381
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6944306653488866e-05,
+      "loss": 1.2988,
+      "step": 3382
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942390838229607e-05,
+      "loss": 1.2598,
+      "step": 3383
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694047453096346e-05,
+      "loss": 1.3154,
+      "step": 3384
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693855773182624e-05,
+      "loss": 1.2842,
+      "step": 3385
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693664044095378e-05,
+      "loss": 1.2793,
+      "step": 3386
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693472265848197e-05,
+      "loss": 1.2725,
+      "step": 3387
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932804384546713e-05,
+      "loss": 1.3145,
+      "step": 3388
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930885619283965e-05,
+      "loss": 1.29,
+      "step": 3389
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928966362829697e-05,
+      "loss": 1.3105,
+      "step": 3390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927046615319937e-05,
+      "loss": 1.3564,
+      "step": 3391
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6925126376890735e-05,
+      "loss": 1.2715,
+      "step": 3392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692320564767818e-05,
+      "loss": 1.3311,
+      "step": 3393
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921284427818388e-05,
+      "loss": 1.2686,
+      "step": 3394
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919362717447523e-05,
+      "loss": 1.3223,
+      "step": 3395
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691744051670177e-05,
+      "loss": 1.2686,
+      "step": 3396
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691551782571736e-05,
+      "loss": 1.252,
+      "step": 3397
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6913594644630552e-05,
+      "loss": 1.2627,
+      "step": 3398
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6911670973577644e-05,
+      "loss": 1.3184,
+      "step": 3399
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690974681269496e-05,
+      "loss": 1.293,
+      "step": 3400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690782216211887e-05,
+      "loss": 1.2852,
+      "step": 3401
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690589702198578e-05,
+      "loss": 1.3154,
+      "step": 3402
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903971392432113e-05,
+      "loss": 1.2617,
+      "step": 3403
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902045273594344e-05,
+      "loss": 1.2969,
+      "step": 3404
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900118665608976e-05,
+      "loss": 1.2705,
+      "step": 3405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689819156861255e-05,
+      "loss": 1.3135,
+      "step": 3406
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896263982741634e-05,
+      "loss": 1.2881,
+      "step": 3407
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689433590813284e-05,
+      "loss": 1.2842,
+      "step": 3408
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689240734492281e-05,
+      "loss": 1.29,
+      "step": 3409
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890478293248224e-05,
+      "loss": 1.3516,
+      "step": 3410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888548753245787e-05,
+      "loss": 1.2861,
+      "step": 3411
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688661872505225e-05,
+      "loss": 1.3135,
+      "step": 3412
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884688208804394e-05,
+      "loss": 1.291,
+      "step": 3413
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882757204639033e-05,
+      "loss": 1.3604,
+      "step": 3414
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6880825712693015e-05,
+      "loss": 1.2559,
+      "step": 3415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687889373310323e-05,
+      "loss": 1.2969,
+      "step": 3416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68769612660066e-05,
+      "loss": 1.2646,
+      "step": 3417
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6875028311540067e-05,
+      "loss": 1.3018,
+      "step": 3418
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873094869840627e-05,
+      "loss": 1.3311,
+      "step": 3419
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871160941045303e-05,
+      "loss": 1.3047,
+      "step": 3420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6869226525291154e-05,
+      "loss": 1.3076,
+      "step": 3421
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6867291622715264e-05,
+      "loss": 1.2979,
+      "step": 3422
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6865356233454768e-05,
+      "loss": 1.2422,
+      "step": 3423
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6863420357646824e-05,
+      "loss": 1.2881,
+      "step": 3424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6861483995428628e-05,
+      "loss": 1.3086,
+      "step": 3425
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6859547146937406e-05,
+      "loss": 1.2617,
+      "step": 3426
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6857609812310428e-05,
+      "loss": 1.2686,
+      "step": 3427
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685567199168499e-05,
+      "loss": 1.2979,
+      "step": 3428
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6853733685198427e-05,
+      "loss": 1.2588,
+      "step": 3429
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68517948929881e-05,
+      "loss": 1.3125,
+      "step": 3430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684985561519142e-05,
+      "loss": 1.2861,
+      "step": 3431
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684791585194582e-05,
+      "loss": 1.3096,
+      "step": 3432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845975603388768e-05,
+      "loss": 1.2988,
+      "step": 3433
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6844034869657776e-05,
+      "loss": 1.2949,
+      "step": 3434
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684209365089038e-05,
+      "loss": 1.2803,
+      "step": 3435
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684015194722415e-05,
+      "loss": 1.2598,
+      "step": 3436
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68382097587967e-05,
+      "loss": 1.3223,
+      "step": 3437
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683626708574567e-05,
+      "loss": 1.291,
+      "step": 3438
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683432392820874e-05,
+      "loss": 1.3076,
+      "step": 3439
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6832380286323618e-05,
+      "loss": 1.2793,
+      "step": 3440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683043616022805e-05,
+      "loss": 1.2646,
+      "step": 3441
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828491550059823e-05,
+      "loss": 1.3418,
+      "step": 3442
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6826546455956743e-05,
+      "loss": 1.2559,
+      "step": 3443
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824600878056662e-05,
+      "loss": 1.3281,
+      "step": 3444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822654816497462e-05,
+      "loss": 1.29,
+      "step": 3445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820708271417063e-05,
+      "loss": 1.29,
+      "step": 3446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818761242953412e-05,
+      "loss": 1.3076,
+      "step": 3447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68168137312445e-05,
+      "loss": 1.3047,
+      "step": 3448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6814865736428342e-05,
+      "loss": 1.2979,
+      "step": 3449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812917258642994e-05,
+      "loss": 1.3369,
+      "step": 3450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810968298026548e-05,
+      "loss": 1.2568,
+      "step": 3451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809018854717122e-05,
+      "loss": 1.2871,
+      "step": 3452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807068928852877e-05,
+      "loss": 1.334,
+      "step": 3453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805118520572e-05,
+      "loss": 1.2734,
+      "step": 3454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680316763001272e-05,
+      "loss": 1.3066,
+      "step": 3455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801216257313293e-05,
+      "loss": 1.3467,
+      "step": 3456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6799264402612017e-05,
+      "loss": 1.2715,
+      "step": 3457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6797312066047217e-05,
+      "loss": 1.3574,
+      "step": 3458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795359247757255e-05,
+      "loss": 1.3047,
+      "step": 3459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679340594788053e-05,
+      "loss": 1.3115,
+      "step": 3460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6791452166555465e-05,
+      "loss": 1.2588,
+      "step": 3461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678949790392053e-05,
+      "loss": 1.3047,
+      "step": 3462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6787543160114227e-05,
+      "loss": 1.2617,
+      "step": 3463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678558793527508e-05,
+      "loss": 1.3672,
+      "step": 3464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783632229541664e-05,
+      "loss": 1.2451,
+      "step": 3465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6781676043052574e-05,
+      "loss": 1.292,
+      "step": 3466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779719375946443e-05,
+      "loss": 1.3301,
+      "step": 3467
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677776222836195e-05,
+      "loss": 1.2578,
+      "step": 3468
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775804600437787e-05,
+      "loss": 1.2939,
+      "step": 3469
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773846492312697e-05,
+      "loss": 1.3232,
+      "step": 3470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677188790412545e-05,
+      "loss": 1.2734,
+      "step": 3471
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6769928836014848e-05,
+      "loss": 1.2783,
+      "step": 3472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767969288119736e-05,
+      "loss": 1.2373,
+      "step": 3473
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6766009260578983e-05,
+      "loss": 1.291,
+      "step": 3474
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6764048753531493e-05,
+      "loss": 1.3105,
+      "step": 3475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6762087767116217e-05,
+      "loss": 1.2891,
+      "step": 3476
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760126301472118e-05,
+      "loss": 1.3086,
+      "step": 3477
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6758164356738212e-05,
+      "loss": 1.3164,
+      "step": 3478
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675620193305354e-05,
+      "loss": 1.2393,
+      "step": 3479
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675423903055718e-05,
+      "loss": 1.293,
+      "step": 3480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675227564938825e-05,
+      "loss": 1.292,
+      "step": 3481
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750311789685877e-05,
+      "loss": 1.2832,
+      "step": 3482
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674834745158925e-05,
+      "loss": 1.2939,
+      "step": 3483
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6746382635237587e-05,
+      "loss": 1.3164,
+      "step": 3484
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6744417340770125e-05,
+      "loss": 1.3203,
+      "step": 3485
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674245156832615e-05,
+      "loss": 1.2578,
+      "step": 3486
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6740485318044974e-05,
+      "loss": 1.3047,
+      "step": 3487
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6738518590065945e-05,
+      "loss": 1.29,
+      "step": 3488
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736551384528442e-05,
+      "loss": 1.2969,
+      "step": 3489
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673458370157189e-05,
+      "loss": 1.2949,
+      "step": 3490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6732615541335725e-05,
+      "loss": 1.2549,
+      "step": 3491
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730646903959444e-05,
+      "loss": 1.2949,
+      "step": 3492
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728677789582553e-05,
+      "loss": 1.3037,
+      "step": 3493
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726708198344615e-05,
+      "loss": 1.3438,
+      "step": 3494
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724738130385205e-05,
+      "loss": 1.2559,
+      "step": 3495
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722767585843943e-05,
+      "loss": 1.2686,
+      "step": 3496
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6720796564860482e-05,
+      "loss": 1.2666,
+      "step": 3497
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6718825067574515e-05,
+      "loss": 1.335,
+      "step": 3498
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716853094125747e-05,
+      "loss": 1.2183,
+      "step": 3499
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6714880644653947e-05,
+      "loss": 1.292,
+      "step": 3500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712907719298894e-05,
+      "loss": 1.3047,
+      "step": 3501
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710934318200414e-05,
+      "loss": 1.3252,
+      "step": 3502
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6708960441498356e-05,
+      "loss": 1.2832,
+      "step": 3503
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706986089332614e-05,
+      "loss": 1.2324,
+      "step": 3504
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705011261843105e-05,
+      "loss": 1.3115,
+      "step": 3505
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670303595916979e-05,
+      "loss": 1.3379,
+      "step": 3506
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6701060181452654e-05,
+      "loss": 1.2812,
+      "step": 3507
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699083928831726e-05,
+      "loss": 1.2939,
+      "step": 3508
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669710720144706e-05,
+      "loss": 1.3242,
+      "step": 3509
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6695129999438743e-05,
+      "loss": 1.3125,
+      "step": 3510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693152322946903e-05,
+      "loss": 1.2559,
+      "step": 3511
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66911741721117e-05,
+      "loss": 1.2686,
+      "step": 3512
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6689195547073318e-05,
+      "loss": 1.3008,
+      "step": 3513
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668721644797199e-05,
+      "loss": 1.2891,
+      "step": 3514
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668523687494797e-05,
+      "loss": 1.2471,
+      "step": 3515
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683256828141554e-05,
+      "loss": 1.2852,
+      "step": 3516
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681276307693063e-05,
+      "loss": 1.3213,
+      "step": 3517
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679295313742858e-05,
+      "loss": 1.3105,
+      "step": 3518
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677313846431333e-05,
+      "loss": 1.2578,
+      "step": 3519
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6675331905898916e-05,
+      "loss": 1.2998,
+      "step": 3520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673349492286065e-05,
+      "loss": 1.3018,
+      "step": 3521
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6671366605733272e-05,
+      "loss": 1.3018,
+      "step": 3522
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669383246381068e-05,
+      "loss": 1.2891,
+      "step": 3523
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666739941437001e-05,
+      "loss": 1.2432,
+      "step": 3524
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665415109840694e-05,
+      "loss": 1.3027,
+      "step": 3525
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6663430332933746e-05,
+      "loss": 1.3057,
+      "step": 3526
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666144508378983e-05,
+      "loss": 1.2861,
+      "step": 3527
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6659459362549637e-05,
+      "loss": 1.2354,
+      "step": 3528
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6657473169353898e-05,
+      "loss": 1.3213,
+      "step": 3529
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655486504343375e-05,
+      "loss": 1.2871,
+      "step": 3530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6653499367658855e-05,
+      "loss": 1.2539,
+      "step": 3531
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651511759441177e-05,
+      "loss": 1.2803,
+      "step": 3532
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649523679831198e-05,
+      "loss": 1.2773,
+      "step": 3533
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664753512896981e-05,
+      "loss": 1.3174,
+      "step": 3534
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664554610699794e-05,
+      "loss": 1.2754,
+      "step": 3535
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643556614056563e-05,
+      "loss": 1.2666,
+      "step": 3536
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641566650286662e-05,
+      "loss": 1.3438,
+      "step": 3537
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663957621582927e-05,
+      "loss": 1.334,
+      "step": 3538
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663758531082545e-05,
+      "loss": 1.25,
+      "step": 3539
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635593935416293e-05,
+      "loss": 1.3359,
+      "step": 3540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633602089742932e-05,
+      "loss": 1.3164,
+      "step": 3541
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631609773946523e-05,
+      "loss": 1.2998,
+      "step": 3542
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662961698816827e-05,
+      "loss": 1.2832,
+      "step": 3543
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627623732549393e-05,
+      "loss": 1.3047,
+      "step": 3544
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662563000723116e-05,
+      "loss": 1.29,
+      "step": 3545
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6623635812354866e-05,
+      "loss": 1.252,
+      "step": 3546
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621641148061836e-05,
+      "loss": 1.3281,
+      "step": 3547
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661964601449343e-05,
+      "loss": 1.3291,
+      "step": 3548
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617650411791052e-05,
+      "loss": 1.3096,
+      "step": 3549
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615654340096124e-05,
+      "loss": 1.2695,
+      "step": 3550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661365779955011e-05,
+      "loss": 1.2598,
+      "step": 3551
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66116607902945e-05,
+      "loss": 1.2949,
+      "step": 3552
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6609663312470824e-05,
+      "loss": 1.2627,
+      "step": 3553
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607665366220645e-05,
+      "loss": 1.2998,
+      "step": 3554
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605666951685558e-05,
+      "loss": 1.2539,
+      "step": 3555
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660366806900719e-05,
+      "loss": 1.2539,
+      "step": 3556
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66016687183272e-05,
+      "loss": 1.3066,
+      "step": 3557
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659966889978728e-05,
+      "loss": 1.2861,
+      "step": 3558
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659766861352916e-05,
+      "loss": 1.3203,
+      "step": 3559
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65956678596946e-05,
+      "loss": 1.2549,
+      "step": 3560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6593666638425396e-05,
+      "loss": 1.2832,
+      "step": 3561
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6591664949863368e-05,
+      "loss": 1.2939,
+      "step": 3562
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589662794150384e-05,
+      "loss": 1.291,
+      "step": 3563
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587660171428326e-05,
+      "loss": 1.251,
+      "step": 3564
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658565708183913e-05,
+      "loss": 1.3174,
+      "step": 3565
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6583653525524746e-05,
+      "loss": 1.2734,
+      "step": 3566
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581649502627174e-05,
+      "loss": 1.2871,
+      "step": 3567
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579645013288435e-05,
+      "loss": 1.3164,
+      "step": 3568
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657764005765059e-05,
+      "loss": 1.3076,
+      "step": 3569
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657563463585572e-05,
+      "loss": 1.3145,
+      "step": 3570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573628748045963e-05,
+      "loss": 1.2998,
+      "step": 3571
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657162239436347e-05,
+      "loss": 1.2432,
+      "step": 3572
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656961557495043e-05,
+      "loss": 1.3564,
+      "step": 3573
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567608289949063e-05,
+      "loss": 1.3203,
+      "step": 3574
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6565600539501636e-05,
+      "loss": 1.3477,
+      "step": 3575
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656359232375043e-05,
+      "loss": 1.2783,
+      "step": 3576
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6561583642837767e-05,
+      "loss": 1.2812,
+      "step": 3577
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6559574496906002e-05,
+      "loss": 1.2979,
+      "step": 3578
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655756488609753e-05,
+      "loss": 1.3154,
+      "step": 3579
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655555481055476e-05,
+      "loss": 1.2881,
+      "step": 3580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655354427042016e-05,
+      "loss": 1.3154,
+      "step": 3581
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551533265836208e-05,
+      "loss": 1.3086,
+      "step": 3582
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654952179694542e-05,
+      "loss": 1.2773,
+      "step": 3583
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654750986389036e-05,
+      "loss": 1.2891,
+      "step": 3584
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545497466813607e-05,
+      "loss": 1.3047,
+      "step": 3585
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543484605857783e-05,
+      "loss": 1.3086,
+      "step": 3586
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654147128116553e-05,
+      "loss": 1.3223,
+      "step": 3587
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539457492879546e-05,
+      "loss": 1.3115,
+      "step": 3588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6537443241142538e-05,
+      "loss": 1.3086,
+      "step": 3589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535428526097257e-05,
+      "loss": 1.2949,
+      "step": 3590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653341334788649e-05,
+      "loss": 1.2803,
+      "step": 3591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653139770665305e-05,
+      "loss": 1.2998,
+      "step": 3592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529381602539783e-05,
+      "loss": 1.2959,
+      "step": 3593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652736503568958e-05,
+      "loss": 1.2783,
+      "step": 3594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652534800624534e-05,
+      "loss": 1.2812,
+      "step": 3595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523330514350017e-05,
+      "loss": 1.2734,
+      "step": 3596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521312560146598e-05,
+      "loss": 1.292,
+      "step": 3597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519294143778083e-05,
+      "loss": 1.2705,
+      "step": 3598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651727526538753e-05,
+      "loss": 1.3242,
+      "step": 3599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515255925118e-05,
+      "loss": 1.2979,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651323612311262e-05,
+      "loss": 1.2295,
+      "step": 3601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6511215859514516e-05,
+      "loss": 1.2803,
+      "step": 3602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6509195134466884e-05,
+      "loss": 1.2783,
+      "step": 3603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650717394811292e-05,
+      "loss": 1.2959,
+      "step": 3604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650515230059587e-05,
+      "loss": 1.2959,
+      "step": 3605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6503130192059e-05,
+      "loss": 1.335,
+      "step": 3606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650110762264563e-05,
+      "loss": 1.3164,
+      "step": 3607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649908459249909e-05,
+      "loss": 1.2871,
+      "step": 3608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6497061101762753e-05,
+      "loss": 1.292,
+      "step": 3609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649503715058003e-05,
+      "loss": 1.249,
+      "step": 3610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649301273909435e-05,
+      "loss": 1.2695,
+      "step": 3611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6490987867449182e-05,
+      "loss": 1.3066,
+      "step": 3612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488962535788038e-05,
+      "loss": 1.2891,
+      "step": 3613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648693674425445e-05,
+      "loss": 1.2607,
+      "step": 3614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484910492991975e-05,
+      "loss": 1.2998,
+      "step": 3615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648288378214423e-05,
+      "loss": 1.2979,
+      "step": 3616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648085661185484e-05,
+      "loss": 1.2959,
+      "step": 3617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478828982267465e-05,
+      "loss": 1.2783,
+      "step": 3618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476800893525814e-05,
+      "loss": 1.3691,
+      "step": 3619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474772345773608e-05,
+      "loss": 1.2676,
+      "step": 3620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6472743339154614e-05,
+      "loss": 1.2607,
+      "step": 3621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6470713873812633e-05,
+      "loss": 1.2773,
+      "step": 3622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468683949891482e-05,
+      "loss": 1.2969,
+      "step": 3623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646665356753503e-05,
+      "loss": 1.291,
+      "step": 3624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6464622726887167e-05,
+      "loss": 1.2939,
+      "step": 3625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646259142809182e-05,
+      "loss": 1.3369,
+      "step": 3626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6460559671292948e-05,
+      "loss": 1.292,
+      "step": 3627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458527456634536e-05,
+      "loss": 1.3223,
+      "step": 3628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456494784260614e-05,
+      "loss": 1.2344,
+      "step": 3629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6454461654315228e-05,
+      "loss": 1.2861,
+      "step": 3630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6452428066942476e-05,
+      "loss": 1.3027,
+      "step": 3631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645039402228647e-05,
+      "loss": 1.291,
+      "step": 3632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448359520491376e-05,
+      "loss": 1.2334,
+      "step": 3633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446324561701362e-05,
+      "loss": 1.2793,
+      "step": 3634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6444289146060655e-05,
+      "loss": 1.293,
+      "step": 3635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64422532737135e-05,
+      "loss": 1.2529,
+      "step": 3636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440216944804184e-05,
+      "loss": 1.2988,
+      "step": 3637
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438180159477022e-05,
+      "loss": 1.3545,
+      "step": 3638
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643614291787636e-05,
+      "loss": 1.2754,
+      "step": 3639
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643410522014657e-05,
+      "loss": 1.2334,
+      "step": 3640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6432067066432074e-05,
+      "loss": 1.2793,
+      "step": 3641
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6430028456877307e-05,
+      "loss": 1.3135,
+      "step": 3642
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427989391626753e-05,
+      "loss": 1.2676,
+      "step": 3643
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6425949870824912e-05,
+      "loss": 1.2725,
+      "step": 3644
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6423909894616336e-05,
+      "loss": 1.3262,
+      "step": 3645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6421869463145586e-05,
+      "loss": 1.2646,
+      "step": 3646
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419828576557277e-05,
+      "loss": 1.3037,
+      "step": 3647
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417787234996035e-05,
+      "loss": 1.2705,
+      "step": 3648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415745438606544e-05,
+      "loss": 1.2334,
+      "step": 3649
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413703187533494e-05,
+      "loss": 1.2578,
+      "step": 3650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641166048192163e-05,
+      "loss": 1.2334,
+      "step": 3651
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409617321915708e-05,
+      "loss": 1.3057,
+      "step": 3652
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407573707660535e-05,
+      "loss": 1.2939,
+      "step": 3653
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405529639300933e-05,
+      "loss": 1.2451,
+      "step": 3654
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403485116981774e-05,
+      "loss": 1.2793,
+      "step": 3655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640144014084795e-05,
+      "loss": 1.2314,
+      "step": 3656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399394711044384e-05,
+      "loss": 1.3037,
+      "step": 3657
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397348827716043e-05,
+      "loss": 1.2676,
+      "step": 3658
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395302491007914e-05,
+      "loss": 1.2842,
+      "step": 3659
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639325570106502e-05,
+      "loss": 1.291,
+      "step": 3660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639120845803242e-05,
+      "loss": 1.3535,
+      "step": 3661
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389160762055203e-05,
+      "loss": 1.2783,
+      "step": 3662
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638711261327848e-05,
+      "loss": 1.2598,
+      "step": 3663
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6385064011847414e-05,
+      "loss": 1.3428,
+      "step": 3664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638301495790719e-05,
+      "loss": 1.2676,
+      "step": 3665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638096545160301e-05,
+      "loss": 1.3086,
+      "step": 3666
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637891549308014e-05,
+      "loss": 1.2998,
+      "step": 3667
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6376865082483848e-05,
+      "loss": 1.293,
+      "step": 3668
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374814219959453e-05,
+      "loss": 1.3057,
+      "step": 3669
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372762905652296e-05,
+      "loss": 1.2627,
+      "step": 3670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6370711139707757e-05,
+      "loss": 1.2109,
+      "step": 3671
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368658922271238e-05,
+      "loss": 1.3164,
+      "step": 3672
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6366606253488186e-05,
+      "loss": 1.2832,
+      "step": 3673
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364553133504072e-05,
+      "loss": 1.3242,
+      "step": 3674
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6362499562464404e-05,
+      "loss": 1.2305,
+      "step": 3675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636044554051471e-05,
+      "loss": 1.3076,
+      "step": 3676
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358391067800564e-05,
+      "loss": 1.2646,
+      "step": 3677
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6356336144467568e-05,
+      "loss": 1.3076,
+      "step": 3678
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354280770661345e-05,
+      "loss": 1.2881,
+      "step": 3679
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635222494652757e-05,
+      "loss": 1.2803,
+      "step": 3680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6350168672211934e-05,
+      "loss": 1.3359,
+      "step": 3681
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634811194786017e-05,
+      "loss": 1.2617,
+      "step": 3682
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634605477361803e-05,
+      "loss": 1.2705,
+      "step": 3683
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634399714963131e-05,
+      "loss": 1.3408,
+      "step": 3684
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6341939076045836e-05,
+      "loss": 1.3252,
+      "step": 3685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633988055300746e-05,
+      "loss": 1.3076,
+      "step": 3686
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633782158066207e-05,
+      "loss": 1.3125,
+      "step": 3687
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6335762159155586e-05,
+      "loss": 1.3047,
+      "step": 3688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6333702288633962e-05,
+      "loss": 1.2754,
+      "step": 3689
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331641969243173e-05,
+      "loss": 1.3174,
+      "step": 3690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329581201129242e-05,
+      "loss": 1.2881,
+      "step": 3691
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6327519984438213e-05,
+      "loss": 1.3105,
+      "step": 3692
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632545831931616e-05,
+      "loss": 1.2832,
+      "step": 3693
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63233962059092e-05,
+      "loss": 1.2773,
+      "step": 3694
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321333644363465e-05,
+      "loss": 1.2705,
+      "step": 3695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319270634825142e-05,
+      "loss": 1.2666,
+      "step": 3696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6317207177440425e-05,
+      "loss": 1.2832,
+      "step": 3697
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315143272355563e-05,
+      "loss": 1.292,
+      "step": 3698
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631307891971681e-05,
+      "loss": 1.2871,
+      "step": 3699
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6311014119670477e-05,
+      "loss": 1.2783,
+      "step": 3700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6308948872362892e-05,
+      "loss": 1.2236,
+      "step": 3701
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630688317794042e-05,
+      "loss": 1.3438,
+      "step": 3702
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304817036549455e-05,
+      "loss": 1.3057,
+      "step": 3703
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302750448336432e-05,
+      "loss": 1.2822,
+      "step": 3704
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63006834134478e-05,
+      "loss": 1.3018,
+      "step": 3705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298615932030055e-05,
+      "loss": 1.3027,
+      "step": 3706
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629654800422972e-05,
+      "loss": 1.3213,
+      "step": 3707
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294479630193343e-05,
+      "loss": 1.3369,
+      "step": 3708
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629241081006752e-05,
+      "loss": 1.3164,
+      "step": 3709
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290341543998858e-05,
+      "loss": 1.3105,
+      "step": 3710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628827183213401e-05,
+      "loss": 1.2852,
+      "step": 3711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6286201674619655e-05,
+      "loss": 1.29,
+      "step": 3712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284131071602507e-05,
+      "loss": 1.2734,
+      "step": 3713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6282060023229308e-05,
+      "loss": 1.2744,
+      "step": 3714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279988529646833e-05,
+      "loss": 1.2549,
+      "step": 3715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627791659100189e-05,
+      "loss": 1.293,
+      "step": 3716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275844207441318e-05,
+      "loss": 1.2744,
+      "step": 3717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627377137911198e-05,
+      "loss": 1.2754,
+      "step": 3718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627169810616079e-05,
+      "loss": 1.2559,
+      "step": 3719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6269624388734668e-05,
+      "loss": 1.292,
+      "step": 3720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267550226980584e-05,
+      "loss": 1.2402,
+      "step": 3721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265475621045536e-05,
+      "loss": 1.3369,
+      "step": 3722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263400571076546e-05,
+      "loss": 1.3301,
+      "step": 3723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261325077220675e-05,
+      "loss": 1.3154,
+      "step": 3724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259249139625012e-05,
+      "loss": 1.2314,
+      "step": 3725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6257172758436683e-05,
+      "loss": 1.2637,
+      "step": 3726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625509593380284e-05,
+      "loss": 1.2656,
+      "step": 3727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625301866587066e-05,
+      "loss": 1.2852,
+      "step": 3728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250940954787363e-05,
+      "loss": 1.2588,
+      "step": 3729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248862800700203e-05,
+      "loss": 1.2695,
+      "step": 3730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624678420375645e-05,
+      "loss": 1.2842,
+      "step": 3731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6244705164103415e-05,
+      "loss": 1.3193,
+      "step": 3732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242625681888446e-05,
+      "loss": 1.2637,
+      "step": 3733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6240545757258908e-05,
+      "loss": 1.2529,
+      "step": 3734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623846539036221e-05,
+      "loss": 1.2832,
+      "step": 3735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236384581345785e-05,
+      "loss": 1.2783,
+      "step": 3736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6234303330357103e-05,
+      "loss": 1.2705,
+      "step": 3737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623222163754366e-05,
+      "loss": 1.3301,
+      "step": 3738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6230139503052985e-05,
+      "loss": 1.2549,
+      "step": 3739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622805692703264e-05,
+      "loss": 1.3086,
+      "step": 3740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622597390963021e-05,
+      "loss": 1.2422,
+      "step": 3741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223890450993333e-05,
+      "loss": 1.2939,
+      "step": 3742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622180655126965e-05,
+      "loss": 1.2852,
+      "step": 3743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219722210606855e-05,
+      "loss": 1.3037,
+      "step": 3744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6217637429152658e-05,
+      "loss": 1.2373,
+      "step": 3745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621555220705482e-05,
+      "loss": 1.3232,
+      "step": 3746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6213466544461103e-05,
+      "loss": 1.3174,
+      "step": 3747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211380441519333e-05,
+      "loss": 1.251,
+      "step": 3748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6209293898377345e-05,
+      "loss": 1.2812,
+      "step": 3749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6207206915183012e-05,
+      "loss": 1.2715,
+      "step": 3750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620511949208424e-05,
+      "loss": 1.3164,
+      "step": 3751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203031629228964e-05,
+      "loss": 1.25,
+      "step": 3752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620094332676515e-05,
+      "loss": 1.2773,
+      "step": 3753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198854584840796e-05,
+      "loss": 1.332,
+      "step": 3754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6196765403603937e-05,
+      "loss": 1.2949,
+      "step": 3755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194675783202622e-05,
+      "loss": 1.2539,
+      "step": 3756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619258572378495e-05,
+      "loss": 1.2646,
+      "step": 3757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6190495225499042e-05,
+      "loss": 1.2441,
+      "step": 3758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6188404288493052e-05,
+      "loss": 1.251,
+      "step": 3759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6186312912915164e-05,
+      "loss": 1.2852,
+      "step": 3760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184221098913595e-05,
+      "loss": 1.3076,
+      "step": 3761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182128846636585e-05,
+      "loss": 1.2842,
+      "step": 3762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618003615623242e-05,
+      "loss": 1.2969,
+      "step": 3763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177943027849407e-05,
+      "loss": 1.3291,
+      "step": 3764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175849461635883e-05,
+      "loss": 1.2891,
+      "step": 3765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173755457740224e-05,
+      "loss": 1.2979,
+      "step": 3766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171661016310825e-05,
+      "loss": 1.3184,
+      "step": 3767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6169566137496126e-05,
+      "loss": 1.2988,
+      "step": 3768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6167470821444585e-05,
+      "loss": 1.291,
+      "step": 3769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165375068304704e-05,
+      "loss": 1.2822,
+      "step": 3770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6163278878224998e-05,
+      "loss": 1.2764,
+      "step": 3771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161182251354033e-05,
+      "loss": 1.2764,
+      "step": 3772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6159085187840393e-05,
+      "loss": 1.2734,
+      "step": 3773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61569876878327e-05,
+      "loss": 1.2812,
+      "step": 3774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154889751479598e-05,
+      "loss": 1.3047,
+      "step": 3775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615279137892978e-05,
+      "loss": 1.2412,
+      "step": 3776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615069257033194e-05,
+      "loss": 1.3701,
+      "step": 3777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148593325834832e-05,
+      "loss": 1.2686,
+      "step": 3778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146493645587228e-05,
+      "loss": 1.2852,
+      "step": 3779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614439352973793e-05,
+      "loss": 1.2959,
+      "step": 3780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142292978435775e-05,
+      "loss": 1.2646,
+      "step": 3781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140191991829625e-05,
+      "loss": 1.2324,
+      "step": 3782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138090570068383e-05,
+      "loss": 1.3477,
+      "step": 3783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6135988713300973e-05,
+      "loss": 1.3037,
+      "step": 3784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6133886421676352e-05,
+      "loss": 1.2334,
+      "step": 3785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131783695343516e-05,
+      "loss": 1.2881,
+      "step": 3786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6129680534451475e-05,
+      "loss": 1.333,
+      "step": 3787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612757693914929e-05,
+      "loss": 1.2861,
+      "step": 3788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6125472909586032e-05,
+      "loss": 1.2715,
+      "step": 3789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6123368445910827e-05,
+      "loss": 1.2881,
+      "step": 3790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6121263548272802e-05,
+      "loss": 1.3271,
+      "step": 3791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6119158216821147e-05,
+      "loss": 1.3008,
+      "step": 3792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117052451705056e-05,
+      "loss": 1.3271,
+      "step": 3793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6114946253073766e-05,
+      "loss": 1.3115,
+      "step": 3794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112839621076547e-05,
+      "loss": 1.2529,
+      "step": 3795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110732555862697e-05,
+      "loss": 1.2861,
+      "step": 3796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610862505758154e-05,
+      "loss": 1.292,
+      "step": 3797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610651712638243e-05,
+      "loss": 1.3047,
+      "step": 3798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6104408762414762e-05,
+      "loss": 1.3242,
+      "step": 3799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610229996582796e-05,
+      "loss": 1.2725,
+      "step": 3800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610019073677146e-05,
+      "loss": 1.2568,
+      "step": 3801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098081075394757e-05,
+      "loss": 1.291,
+      "step": 3802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609597098184736e-05,
+      "loss": 1.2773,
+      "step": 3803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093860456278806e-05,
+      "loss": 1.2979,
+      "step": 3804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6091749498838666e-05,
+      "loss": 1.2852,
+      "step": 3805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6089638109676557e-05,
+      "loss": 1.3232,
+      "step": 3806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6087526288942094e-05,
+      "loss": 1.2803,
+      "step": 3807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608541403678496e-05,
+      "loss": 1.3047,
+      "step": 3808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6083301353354836e-05,
+      "loss": 1.3105,
+      "step": 3809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081188238801453e-05,
+      "loss": 1.3047,
+      "step": 3810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607907469327457e-05,
+      "loss": 1.2803,
+      "step": 3811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076960716923973e-05,
+      "loss": 1.2627,
+      "step": 3812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6074846309899475e-05,
+      "loss": 1.2861,
+      "step": 3813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607273147235093e-05,
+      "loss": 1.3369,
+      "step": 3814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070616204428212e-05,
+      "loss": 1.2842,
+      "step": 3815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6068500506281232e-05,
+      "loss": 1.2891,
+      "step": 3816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066384378059927e-05,
+      "loss": 1.3008,
+      "step": 3817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606426781991427e-05,
+      "loss": 1.3086,
+      "step": 3818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606215083199426e-05,
+      "loss": 1.3037,
+      "step": 3819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6060033414449927e-05,
+      "loss": 1.2773,
+      "step": 3820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057915567431334e-05,
+      "loss": 1.2656,
+      "step": 3821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055797291088573e-05,
+      "loss": 1.3047,
+      "step": 3822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053678585571764e-05,
+      "loss": 1.2441,
+      "step": 3823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605155945103106e-05,
+      "loss": 1.2861,
+      "step": 3824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049439887616646e-05,
+      "loss": 1.252,
+      "step": 3825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6047319895478732e-05,
+      "loss": 1.2705,
+      "step": 3826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045199474767565e-05,
+      "loss": 1.2471,
+      "step": 3827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043078625633418e-05,
+      "loss": 1.3125,
+      "step": 3828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040957348226595e-05,
+      "loss": 1.3115,
+      "step": 3829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038835642697428e-05,
+      "loss": 1.3057,
+      "step": 3830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603671350919629e-05,
+      "loss": 1.2568,
+      "step": 3831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6034590947873568e-05,
+      "loss": 1.3145,
+      "step": 3832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603246795887969e-05,
+      "loss": 1.2773,
+      "step": 3833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6030344542365118e-05,
+      "loss": 1.2783,
+      "step": 3834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602822069848033e-05,
+      "loss": 1.3115,
+      "step": 3835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602609642737585e-05,
+      "loss": 1.249,
+      "step": 3836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023971729202223e-05,
+      "loss": 1.3291,
+      "step": 3837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602184660411002e-05,
+      "loss": 1.2529,
+      "step": 3838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019721052249853e-05,
+      "loss": 1.2695,
+      "step": 3839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6017595073772362e-05,
+      "loss": 1.2441,
+      "step": 3840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015468668828215e-05,
+      "loss": 1.293,
+      "step": 3841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601334183756811e-05,
+      "loss": 1.3174,
+      "step": 3842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601121458014277e-05,
+      "loss": 1.2715,
+      "step": 3843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600908689670296e-05,
+      "loss": 1.2764,
+      "step": 3844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6006958787399466e-05,
+      "loss": 1.3096,
+      "step": 3845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004830252383104e-05,
+      "loss": 1.3115,
+      "step": 3846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600270129180473e-05,
+      "loss": 1.3066,
+      "step": 3847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600057190581522e-05,
+      "loss": 1.3125,
+      "step": 3848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599844209456548e-05,
+      "loss": 1.2852,
+      "step": 3849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5996311858206455e-05,
+      "loss": 1.3584,
+      "step": 3850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599418119688911e-05,
+      "loss": 1.2559,
+      "step": 3851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599205011076445e-05,
+      "loss": 1.2607,
+      "step": 3852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989918599983503e-05,
+      "loss": 1.3086,
+      "step": 3853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987786664697326e-05,
+      "loss": 1.2871,
+      "step": 3854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985654305057018e-05,
+      "loss": 1.2646,
+      "step": 3855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5983521521213682e-05,
+      "loss": 1.2812,
+      "step": 3856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598138831331849e-05,
+      "loss": 1.2949,
+      "step": 3857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5979254681522603e-05,
+      "loss": 1.2725,
+      "step": 3858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5977120625977244e-05,
+      "loss": 1.3154,
+      "step": 3859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597498614683365e-05,
+      "loss": 1.2617,
+      "step": 3860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597285124424309e-05,
+      "loss": 1.292,
+      "step": 3861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5970715918356866e-05,
+      "loss": 1.25,
+      "step": 3862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5968580169326304e-05,
+      "loss": 1.2393,
+      "step": 3863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596644399730277e-05,
+      "loss": 1.2852,
+      "step": 3864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964307402437654e-05,
+      "loss": 1.292,
+      "step": 3865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596217038488237e-05,
+      "loss": 1.2637,
+      "step": 3866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596003294478838e-05,
+      "loss": 1.293,
+      "step": 3867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5957895082307152e-05,
+      "loss": 1.293,
+      "step": 3868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5955756797590205e-05,
+      "loss": 1.2705,
+      "step": 3869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5953618090789072e-05,
+      "loss": 1.3184,
+      "step": 3870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595147896205533e-05,
+      "loss": 1.2764,
+      "step": 3871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5949339411540573e-05,
+      "loss": 1.2168,
+      "step": 3872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947199439396433e-05,
+      "loss": 1.2822,
+      "step": 3873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945059045774574e-05,
+      "loss": 1.2754,
+      "step": 3874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942918230826682e-05,
+      "loss": 1.3193,
+      "step": 3875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594077699470447e-05,
+      "loss": 1.2793,
+      "step": 3876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59386353375597e-05,
+      "loss": 1.2568,
+      "step": 3877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5936493259544142e-05,
+      "loss": 1.2979,
+      "step": 3878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5934350760809605e-05,
+      "loss": 1.2725,
+      "step": 3879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932207841507933e-05,
+      "loss": 1.3145,
+      "step": 3880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930064501790995e-05,
+      "loss": 1.2891,
+      "step": 3881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927920741810684e-05,
+      "loss": 1.2432,
+      "step": 3882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5925776561718932e-05,
+      "loss": 1.2217,
+      "step": 3883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923631961667695e-05,
+      "loss": 1.2217,
+      "step": 3884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592148694180896e-05,
+      "loss": 1.2578,
+      "step": 3885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919341502294743e-05,
+      "loss": 1.2432,
+      "step": 3886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917195643277095e-05,
+      "loss": 1.2627,
+      "step": 3887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915049364908096e-05,
+      "loss": 1.2891,
+      "step": 3888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5912902667339844e-05,
+      "loss": 1.2764,
+      "step": 3889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5910755550724482e-05,
+      "loss": 1.2656,
+      "step": 3890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908608015214173e-05,
+      "loss": 1.3008,
+      "step": 3891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906460060961112e-05,
+      "loss": 1.2725,
+      "step": 3892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904311688117528e-05,
+      "loss": 1.3291,
+      "step": 3893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5902162896835673e-05,
+      "loss": 1.3125,
+      "step": 3894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5900013687267835e-05,
+      "loss": 1.2383,
+      "step": 3895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897864059566322e-05,
+      "loss": 1.2979,
+      "step": 3896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895714013883486e-05,
+      "loss": 1.3164,
+      "step": 3897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893563550371694e-05,
+      "loss": 1.2695,
+      "step": 3898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5891412669183355e-05,
+      "loss": 1.2891,
+      "step": 3899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889261370470897e-05,
+      "loss": 1.3213,
+      "step": 3900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5887109654386783e-05,
+      "loss": 1.2617,
+      "step": 3901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588495752108351e-05,
+      "loss": 1.3096,
+      "step": 3902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588280497071359e-05,
+      "loss": 1.2725,
+      "step": 3903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880652003429583e-05,
+      "loss": 1.293,
+      "step": 3904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587849861938407e-05,
+      "loss": 1.2812,
+      "step": 3905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5876344818729652e-05,
+      "loss": 1.2773,
+      "step": 3906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587419060161898e-05,
+      "loss": 1.3096,
+      "step": 3907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872035968204712e-05,
+      "loss": 1.29,
+      "step": 3908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5869880918639558e-05,
+      "loss": 1.2588,
+      "step": 3909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586772545307624e-05,
+      "loss": 1.2607,
+      "step": 3910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586556957166751e-05,
+      "loss": 1.2598,
+      "step": 3911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863413274566167e-05,
+      "loss": 1.2637,
+      "step": 3912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5861256561925025e-05,
+      "loss": 1.2979,
+      "step": 3913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859099433896923e-05,
+      "loss": 1.2383,
+      "step": 3914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856941890634743e-05,
+      "loss": 1.2803,
+      "step": 3915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585478393229139e-05,
+      "loss": 1.333,
+      "step": 3916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852625559019793e-05,
+      "loss": 1.2715,
+      "step": 3917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5850466770972923e-05,
+      "loss": 1.3232,
+      "step": 3918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584830756830377e-05,
+      "loss": 1.3096,
+      "step": 3919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584614795116535e-05,
+      "loss": 1.2793,
+      "step": 3920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584398791971073e-05,
+      "loss": 1.2744,
+      "step": 3921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841827474092975e-05,
+      "loss": 1.3174,
+      "step": 3922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5839666614465205e-05,
+      "loss": 1.3057,
+      "step": 3923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583750534098056e-05,
+      "loss": 1.3154,
+      "step": 3924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835343653792204e-05,
+      "loss": 1.2949,
+      "step": 3925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583318155305334e-05,
+      "loss": 1.2676,
+      "step": 3926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58310190389172e-05,
+      "loss": 1.3428,
+      "step": 3927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582885611153703e-05,
+      "loss": 1.2002,
+      "step": 3928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826692771066127e-05,
+      "loss": 1.292,
+      "step": 3929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5824529017657798e-05,
+      "loss": 1.3213,
+      "step": 3930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822364851465397e-05,
+      "loss": 1.2832,
+      "step": 3931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820200272642292e-05,
+      "loss": 1.3047,
+      "step": 3932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581803528134189e-05,
+      "loss": 1.3066,
+      "step": 3933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581586987771762e-05,
+      "loss": 1.249,
+      "step": 3934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5813704061922952e-05,
+      "loss": 1.29,
+      "step": 3935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581153783411137e-05,
+      "loss": 1.3105,
+      "step": 3936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809371194436395e-05,
+      "loss": 1.3242,
+      "step": 3937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5807204143051577e-05,
+      "loss": 1.3018,
+      "step": 3938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5805036680110496e-05,
+      "loss": 1.2725,
+      "step": 3939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580286880576676e-05,
+      "loss": 1.2764,
+      "step": 3940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800700520174008e-05,
+      "loss": 1.3018,
+      "step": 3941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57985318234859e-05,
+      "loss": 1.2207,
+      "step": 3942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579636271585614e-05,
+      "loss": 1.2607,
+      "step": 3943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5794193197438446e-05,
+      "loss": 1.3145,
+      "step": 3944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5792023268386577e-05,
+      "loss": 1.2461,
+      "step": 3945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789852928854305e-05,
+      "loss": 1.2891,
+      "step": 3946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787682178995455e-05,
+      "loss": 1.2236,
+      "step": 3947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578551101896386e-05,
+      "loss": 1.3066,
+      "step": 3948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783339448913396e-05,
+      "loss": 1.2441,
+      "step": 3949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578116746899796e-05,
+      "loss": 1.2686,
+      "step": 3950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5778995079371472e-05,
+      "loss": 1.2695,
+      "step": 3951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57768222801879e-05,
+      "loss": 1.293,
+      "step": 3952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5774649071601224e-05,
+      "loss": 1.2539,
+      "step": 3953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577247545376546e-05,
+      "loss": 1.2773,
+      "step": 3954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770301426834654e-05,
+      "loss": 1.29,
+      "step": 3955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5768126990962883e-05,
+      "loss": 1.2598,
+      "step": 3956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576595214630424e-05,
+      "loss": 1.2744,
+      "step": 3957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763776893012864e-05,
+      "loss": 1.2734,
+      "step": 3958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761601231242906e-05,
+      "loss": 1.2988,
+      "step": 3959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759425161148566e-05,
+      "loss": 1.25,
+      "step": 3960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757248682884054e-05,
+      "loss": 1.2793,
+      "step": 3961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5755071796603623e-05,
+      "loss": 1.2588,
+      "step": 3962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575289450246154e-05,
+      "loss": 1.2305,
+      "step": 3963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575071680061212e-05,
+      "loss": 1.2568,
+      "step": 3964
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574853869120969e-05,
+      "loss": 1.2598,
+      "step": 3965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5746360174408617e-05,
+      "loss": 1.3076,
+      "step": 3966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5744181250363287e-05,
+      "loss": 1.2588,
+      "step": 3967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5742001919228122e-05,
+      "loss": 1.3223,
+      "step": 3968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739822181157574e-05,
+      "loss": 1.3027,
+      "step": 3969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5737642036306117e-05,
+      "loss": 1.2598,
+      "step": 3970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5735461484828264e-05,
+      "loss": 1.2871,
+      "step": 3971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573328052687854e-05,
+      "loss": 1.3066,
+      "step": 3972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573109916261152e-05,
+      "loss": 1.2461,
+      "step": 3973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5728917392181795e-05,
+      "loss": 1.2832,
+      "step": 3974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572673521574398e-05,
+      "loss": 1.2734,
+      "step": 3975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724552633452733e-05,
+      "loss": 1.3154,
+      "step": 3976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572236964546273e-05,
+      "loss": 1.2627,
+      "step": 3977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572018625192868e-05,
+      "loss": 1.2656,
+      "step": 3978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718002453005326e-05,
+      "loss": 1.3037,
+      "step": 3979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571581824884742e-05,
+      "loss": 1.3438,
+      "step": 3980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571363363960977e-05,
+      "loss": 1.2979,
+      "step": 3981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571144862544719e-05,
+      "loss": 1.3008,
+      "step": 3982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570926320651454e-05,
+      "loss": 1.2676,
+      "step": 3983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707077382966694e-05,
+      "loss": 1.252,
+      "step": 3984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570489115495856e-05,
+      "loss": 1.2363,
+      "step": 3985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570270452264508e-05,
+      "loss": 1.2715,
+      "step": 3986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570051748618122e-05,
+      "loss": 1.2852,
+      "step": 3987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569833004572198e-05,
+      "loss": 1.2471,
+      "step": 3988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569614220142237e-05,
+      "loss": 1.2871,
+      "step": 3989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693953953437457e-05,
+      "loss": 1.2764,
+      "step": 3990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569176530192231e-05,
+      "loss": 1.2803,
+      "step": 3991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689576247032054e-05,
+      "loss": 1.2646,
+      "step": 3992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568738678892181e-05,
+      "loss": 1.2412,
+      "step": 3993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5685196927746757e-05,
+      "loss": 1.2783,
+      "step": 3994
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683006663662083e-05,
+      "loss": 1.2842,
+      "step": 3995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5680815996823014e-05,
+      "loss": 1.2881,
+      "step": 3996
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567862492738481e-05,
+      "loss": 1.3115,
+      "step": 3997
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676433455502736e-05,
+      "loss": 1.2881,
+      "step": 3998
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567424158133212e-05,
+      "loss": 1.3223,
+      "step": 3999
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5672049305028286e-05,
+      "loss": 1.2783,
+      "step": 4000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5669856626746603e-05,
+      "loss": 1.2744,
+      "step": 4001
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566766354664247e-05,
+      "loss": 1.3555,
+      "step": 4002
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665470064871308e-05,
+      "loss": 1.2598,
+      "step": 4003
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5663276181588574e-05,
+      "loss": 1.2686,
+      "step": 4004
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566108189694974e-05,
+      "loss": 1.2266,
+      "step": 4005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5658887211110322e-05,
+      "loss": 1.2949,
+      "step": 4006
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5656692124225852e-05,
+      "loss": 1.2959,
+      "step": 4007
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5654496636451896e-05,
+      "loss": 1.2754,
+      "step": 4008
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652300747944054e-05,
+      "loss": 1.3271,
+      "step": 4009
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5650104458857943e-05,
+      "loss": 1.3105,
+      "step": 4010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647907769349217e-05,
+      "loss": 1.2617,
+      "step": 4011
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564571067957355e-05,
+      "loss": 1.3184,
+      "step": 4012
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5643513189686658e-05,
+      "loss": 1.3086,
+      "step": 4013
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564131529984427e-05,
+      "loss": 1.2607,
+      "step": 4014
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563911701020215e-05,
+      "loss": 1.3193,
+      "step": 4015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5636918320916096e-05,
+      "loss": 1.3018,
+      "step": 4016
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634719232141925e-05,
+      "loss": 1.2471,
+      "step": 4017
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563251974403549e-05,
+      "loss": 1.2529,
+      "step": 4018
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5630319856752664e-05,
+      "loss": 1.2871,
+      "step": 4019
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5628119570449357e-05,
+      "loss": 1.2441,
+      "step": 4020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56259188852815e-05,
+      "loss": 1.3047,
+      "step": 4021
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562371780140506e-05,
+      "loss": 1.2773,
+      "step": 4022
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562151631897602e-05,
+      "loss": 1.2656,
+      "step": 4023
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5619314438150407e-05,
+      "loss": 1.2998,
+      "step": 4024
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5617112159084263e-05,
+      "loss": 1.3076,
+      "step": 4025
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5614909481933666e-05,
+      "loss": 1.2695,
+      "step": 4026
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612706406854716e-05,
+      "loss": 1.2842,
+      "step": 4027
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561050293400355e-05,
+      "loss": 1.2822,
+      "step": 4028
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5608299063536326e-05,
+      "loss": 1.2617,
+      "step": 4029
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560609479560923e-05,
+      "loss": 1.2959,
+      "step": 4030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560389013037848e-05,
+      "loss": 1.2686,
+      "step": 4031
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5601685068000322e-05,
+      "loss": 1.2686,
+      "step": 4032
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599479608631026e-05,
+      "loss": 1.3232,
+      "step": 4033
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5597273752426896e-05,
+      "loss": 1.2578,
+      "step": 4034
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5595067499544256e-05,
+      "loss": 1.2373,
+      "step": 4035
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559286085013947e-05,
+      "loss": 1.3057,
+      "step": 4036
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590653804368915e-05,
+      "loss": 1.3076,
+      "step": 4037
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558844636238901e-05,
+      "loss": 1.291,
+      "step": 4038
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558623852435619e-05,
+      "loss": 1.3037,
+      "step": 4039
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584030290426934e-05,
+      "loss": 1.2969,
+      "step": 4040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5581821660757735e-05,
+      "loss": 1.2666,
+      "step": 4041
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5579612635505115e-05,
+      "loss": 1.3213,
+      "step": 4042
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577403214825634e-05,
+      "loss": 1.2812,
+      "step": 4043
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557519339887587e-05,
+      "loss": 1.2773,
+      "step": 4044
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5572983187812427e-05,
+      "loss": 1.2832,
+      "step": 4045
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557077258179195e-05,
+      "loss": 1.293,
+      "step": 4046
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568561580971105e-05,
+      "loss": 1.2646,
+      "step": 4047
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566350185506586e-05,
+      "loss": 1.2598,
+      "step": 4048
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5564138395555102e-05,
+      "loss": 1.293,
+      "step": 4049
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556192621127342e-05,
+      "loss": 1.3701,
+      "step": 4050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559713632818306e-05,
+      "loss": 1.3047,
+      "step": 4051
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555750066034657e-05,
+      "loss": 1.2861,
+      "step": 4052
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555287294015044e-05,
+      "loss": 1.2871,
+      "step": 4053
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5553073533980586e-05,
+      "loss": 1.2314,
+      "step": 4054
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550859380400093e-05,
+      "loss": 1.2793,
+      "step": 4055
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554864483343047e-05,
+      "loss": 1.2715,
+      "step": 4056
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5546429893228678e-05,
+      "loss": 1.2236,
+      "step": 4057
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544214559951676e-05,
+      "loss": 1.29,
+      "step": 4058
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5541998833756467e-05,
+      "loss": 1.2627,
+      "step": 4059
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539782714800087e-05,
+      "loss": 1.2891,
+      "step": 4060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553756620323958e-05,
+      "loss": 1.2559,
+      "step": 4061
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535349299232045e-05,
+      "loss": 1.2588,
+      "step": 4062
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533132002934583e-05,
+      "loss": 1.2725,
+      "step": 4063
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5530914314504335e-05,
+      "loss": 1.3086,
+      "step": 4064
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5528696234098468e-05,
+      "loss": 1.3213,
+      "step": 4065
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552647776187418e-05,
+      "loss": 1.2549,
+      "step": 4066
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524258897988696e-05,
+      "loss": 1.2998,
+      "step": 4067
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5522039642599263e-05,
+      "loss": 1.2578,
+      "step": 4068
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551981999586316e-05,
+      "loss": 1.3047,
+      "step": 4069
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5517599957937694e-05,
+      "loss": 1.2646,
+      "step": 4070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515379528980197e-05,
+      "loss": 1.3301,
+      "step": 4071
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513158709148033e-05,
+      "loss": 1.2939,
+      "step": 4072
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510937498598594e-05,
+      "loss": 1.2666,
+      "step": 4073
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550871589748929e-05,
+      "loss": 1.2646,
+      "step": 4074
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5506493905977572e-05,
+      "loss": 1.2842,
+      "step": 4075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550427152422091e-05,
+      "loss": 1.3193,
+      "step": 4076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.55020487523768e-05,
+      "loss": 1.2559,
+      "step": 4077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5499825590602778e-05,
+      "loss": 1.2256,
+      "step": 4078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497602039056392e-05,
+      "loss": 1.2793,
+      "step": 4079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549537809789523e-05,
+      "loss": 1.3359,
+      "step": 4080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5493153767276902e-05,
+      "loss": 1.2832,
+      "step": 4081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490929047359046e-05,
+      "loss": 1.2578,
+      "step": 4082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5488703938299322e-05,
+      "loss": 1.2695,
+      "step": 4083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548647844025543e-05,
+      "loss": 1.2959,
+      "step": 4084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484252553385092e-05,
+      "loss": 1.2939,
+      "step": 4085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548202627784605e-05,
+      "loss": 1.3037,
+      "step": 4086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479799613796087e-05,
+      "loss": 1.2607,
+      "step": 4087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477572561393003e-05,
+      "loss": 1.291,
+      "step": 4088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475345120794627e-05,
+      "loss": 1.2969,
+      "step": 4089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547311729215882e-05,
+      "loss": 1.2871,
+      "step": 4090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547088907564347e-05,
+      "loss": 1.292,
+      "step": 4091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5468660471406484e-05,
+      "loss": 1.3369,
+      "step": 4092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5466431479605816e-05,
+      "loss": 1.2861,
+      "step": 4093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546420210039942e-05,
+      "loss": 1.2676,
+      "step": 4094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54619723339453e-05,
+      "loss": 1.2842,
+      "step": 4095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459742180401472e-05,
+      "loss": 1.2871,
+      "step": 4096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457511639925998e-05,
+      "loss": 1.3242,
+      "step": 4097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545528071267695e-05,
+      "loss": 1.3301,
+      "step": 4098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5453049398812436e-05,
+      "loss": 1.2607,
+      "step": 4099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5450817698490586e-05,
+      "loss": 1.2676,
+      "step": 4100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5448585611869563e-05,
+      "loss": 1.2969,
+      "step": 4101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446353139107553e-05,
+      "loss": 1.3242,
+      "step": 4102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544412028036277e-05,
+      "loss": 1.2578,
+      "step": 4103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544188703579346e-05,
+      "loss": 1.1973,
+      "step": 4104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439653405557897e-05,
+      "loss": 1.3193,
+      "step": 4105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437419389814365e-05,
+      "loss": 1.2627,
+      "step": 4106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54351849887212e-05,
+      "loss": 1.3076,
+      "step": 4107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543295020243675e-05,
+      "loss": 1.2656,
+      "step": 4108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430715031119396e-05,
+      "loss": 1.3213,
+      "step": 4109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542847947492754e-05,
+      "loss": 1.2441,
+      "step": 4110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5426243534019622e-05,
+      "loss": 1.3076,
+      "step": 4111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424007208554097e-05,
+      "loss": 1.3135,
+      "step": 4112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542177049868946e-05,
+      "loss": 1.3076,
+      "step": 4113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541953340458422e-05,
+      "loss": 1.2842,
+      "step": 4114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541729592639692e-05,
+      "loss": 1.2695,
+      "step": 4115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415058064286137e-05,
+      "loss": 1.2744,
+      "step": 4116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412819818410462e-05,
+      "loss": 1.252,
+      "step": 4117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5410581188928516e-05,
+      "loss": 1.2617,
+      "step": 4118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408342175998963e-05,
+      "loss": 1.3135,
+      "step": 4119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540610277978047e-05,
+      "loss": 1.2285,
+      "step": 4120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540386300043175e-05,
+      "loss": 1.291,
+      "step": 4121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540162283811153e-05,
+      "loss": 1.2402,
+      "step": 4122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399382292978574e-05,
+      "loss": 1.3115,
+      "step": 4123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539714136519167e-05,
+      "loss": 1.2471,
+      "step": 4124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394900054909626e-05,
+      "loss": 1.2773,
+      "step": 4125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5392658362291297e-05,
+      "loss": 1.2861,
+      "step": 4126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390416287495537e-05,
+      "loss": 1.3291,
+      "step": 4127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538817383068125e-05,
+      "loss": 1.3066,
+      "step": 4128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385930992007356e-05,
+      "loss": 1.3174,
+      "step": 4129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5383687771632806e-05,
+      "loss": 1.2715,
+      "step": 4130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381444169716578e-05,
+      "loss": 1.3047,
+      "step": 4131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5379200186417674e-05,
+      "loss": 1.3008,
+      "step": 4132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376955821895122e-05,
+      "loss": 1.291,
+      "step": 4133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5374711076307986e-05,
+      "loss": 1.2988,
+      "step": 4134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537246594981535e-05,
+      "loss": 1.2744,
+      "step": 4135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5370220442576323e-05,
+      "loss": 1.2549,
+      "step": 4136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5367974554750048e-05,
+      "loss": 1.2686,
+      "step": 4137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365728286495685e-05,
+      "loss": 1.2959,
+      "step": 4138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5363481637972433e-05,
+      "loss": 1.2939,
+      "step": 4139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361234609339504e-05,
+      "loss": 1.2842,
+      "step": 4140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358987200756157e-05,
+      "loss": 1.252,
+      "step": 4141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5356739412381655e-05,
+      "loss": 1.2324,
+      "step": 4142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5354491244375303e-05,
+      "loss": 1.2188,
+      "step": 4143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5352242696896425e-05,
+      "loss": 1.2607,
+      "step": 4144
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349993770104386e-05,
+      "loss": 1.252,
+      "step": 4145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347744464158554e-05,
+      "loss": 1.2773,
+      "step": 4146
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345494779218345e-05,
+      "loss": 1.2246,
+      "step": 4147
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534324471544319e-05,
+      "loss": 1.2871,
+      "step": 4148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5340994272992554e-05,
+      "loss": 1.2744,
+      "step": 4149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5338743452025922e-05,
+      "loss": 1.2764,
+      "step": 4150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5336492252702816e-05,
+      "loss": 1.2549,
+      "step": 4151
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334240675182775e-05,
+      "loss": 1.2891,
+      "step": 4152
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5331988719625367e-05,
+      "loss": 1.2842,
+      "step": 4153
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329736386190183e-05,
+      "loss": 1.2627,
+      "step": 4154
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532748367503686e-05,
+      "loss": 1.2314,
+      "step": 4155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5325230586325035e-05,
+      "loss": 1.2285,
+      "step": 4156
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532297712021439e-05,
+      "loss": 1.252,
+      "step": 4157
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5320723276864625e-05,
+      "loss": 1.3154,
+      "step": 4158
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5318469056435472e-05,
+      "loss": 1.2793,
+      "step": 4159
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316214459086683e-05,
+      "loss": 1.2461,
+      "step": 4160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313959484978045e-05,
+      "loss": 1.3154,
+      "step": 4161
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531170413426937e-05,
+      "loss": 1.2686,
+      "step": 4162
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53094484071205e-05,
+      "loss": 1.2754,
+      "step": 4163
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530719230369128e-05,
+      "loss": 1.2617,
+      "step": 4164
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304935824141613e-05,
+      "loss": 1.29,
+      "step": 4165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302678968631413e-05,
+      "loss": 1.2969,
+      "step": 4166
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300421737320626e-05,
+      "loss": 1.251,
+      "step": 4167
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298164130369216e-05,
+      "loss": 1.2656,
+      "step": 4168
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295906147937184e-05,
+      "loss": 1.2607,
+      "step": 4169
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529364779018455e-05,
+      "loss": 1.2695,
+      "step": 4170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529138905727137e-05,
+      "loss": 1.2969,
+      "step": 4171
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528912994935771e-05,
+      "loss": 1.25,
+      "step": 4172
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5286870466603683e-05,
+      "loss": 1.2461,
+      "step": 4173
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284610609169413e-05,
+      "loss": 1.2881,
+      "step": 4174
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282350377215055e-05,
+      "loss": 1.2939,
+      "step": 4175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280089770900795e-05,
+      "loss": 1.2891,
+      "step": 4176
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277828790386844e-05,
+      "loss": 1.3047,
+      "step": 4177
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275567435833427e-05,
+      "loss": 1.2432,
+      "step": 4178
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273305707400818e-05,
+      "loss": 1.3135,
+      "step": 4179
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52710436052493e-05,
+      "loss": 1.3164,
+      "step": 4180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526878112953919e-05,
+      "loss": 1.2949,
+      "step": 4181
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526651828043083e-05,
+      "loss": 1.2217,
+      "step": 4182
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264255058084582e-05,
+      "loss": 1.2891,
+      "step": 4183
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261991462660847e-05,
+      "loss": 1.2939,
+      "step": 4184
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5259727494320045e-05,
+      "loss": 1.2637,
+      "step": 4185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525746315322262e-05,
+      "loss": 1.2861,
+      "step": 4186
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255198439529046e-05,
+      "loss": 1.2588,
+      "step": 4187
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252933353399828e-05,
+      "loss": 1.3135,
+      "step": 4188
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250667894995493e-05,
+      "loss": 1.2725,
+      "step": 4189
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248402064476582e-05,
+      "loss": 1.2529,
+      "step": 4190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246135862003687e-05,
+      "loss": 1.2549,
+      "step": 4191
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243869287737412e-05,
+      "loss": 1.2041,
+      "step": 4192
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241602341838385e-05,
+      "loss": 1.2441,
+      "step": 4193
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523933502446726e-05,
+      "loss": 1.2744,
+      "step": 4194
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237067335784736e-05,
+      "loss": 1.3252,
+      "step": 4195
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523479927595151e-05,
+      "loss": 1.2656,
+      "step": 4196
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5232530845128328e-05,
+      "loss": 1.2012,
+      "step": 4197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230262043475948e-05,
+      "loss": 1.252,
+      "step": 4198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5227992871155164e-05,
+      "loss": 1.2783,
+      "step": 4199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522572332832679e-05,
+      "loss": 1.2627,
+      "step": 4200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5223453415151665e-05,
+      "loss": 1.2715,
+      "step": 4201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221183131790667e-05,
+      "loss": 1.2773,
+      "step": 4202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218912478404683e-05,
+      "loss": 1.2295,
+      "step": 4203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216641455154633e-05,
+      "loss": 1.251,
+      "step": 4204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521437006220147e-05,
+      "loss": 1.2939,
+      "step": 4205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212098299706167e-05,
+      "loss": 1.2363,
+      "step": 4206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520982616782972e-05,
+      "loss": 1.3037,
+      "step": 4207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207553666733153e-05,
+      "loss": 1.2793,
+      "step": 4208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205280796577526e-05,
+      "loss": 1.252,
+      "step": 4209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203007557523912e-05,
+      "loss": 1.2617,
+      "step": 4210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200733949733417e-05,
+      "loss": 1.2559,
+      "step": 4211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198459973367167e-05,
+      "loss": 1.2822,
+      "step": 4212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5196185628586326e-05,
+      "loss": 1.29,
+      "step": 4213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5193910915552073e-05,
+      "loss": 1.25,
+      "step": 4214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191635834425615e-05,
+      "loss": 1.2539,
+      "step": 4215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189360385368188e-05,
+      "loss": 1.2949,
+      "step": 4216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187084568541056e-05,
+      "loss": 1.2656,
+      "step": 4217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184808384105502e-05,
+      "loss": 1.2588,
+      "step": 4218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182531832222843e-05,
+      "loss": 1.2578,
+      "step": 4219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518025491305441e-05,
+      "loss": 1.2793,
+      "step": 4220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5177977626761577e-05,
+      "loss": 1.2705,
+      "step": 4221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175699973505737e-05,
+      "loss": 1.2695,
+      "step": 4222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173421953448295e-05,
+      "loss": 1.2832,
+      "step": 4223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171143566750706e-05,
+      "loss": 1.251,
+      "step": 4224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168864813574434e-05,
+      "loss": 1.2588,
+      "step": 4225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166585694080973e-05,
+      "loss": 1.2852,
+      "step": 4226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164306208431845e-05,
+      "loss": 1.2578,
+      "step": 4227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5162026356788601e-05,
+      "loss": 1.25,
+      "step": 4228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515974613931281e-05,
+      "loss": 1.3076,
+      "step": 4229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515746555616607e-05,
+      "loss": 1.3271,
+      "step": 4230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5155184607510007e-05,
+      "loss": 1.293,
+      "step": 4231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5152903293506276e-05,
+      "loss": 1.2334,
+      "step": 4232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150621614316545e-05,
+      "loss": 1.3018,
+      "step": 4233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148339570102524e-05,
+      "loss": 1.2695,
+      "step": 4234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146057161025938e-05,
+      "loss": 1.2949,
+      "step": 4235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143774387248545e-05,
+      "loss": 1.2568,
+      "step": 4236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141491248932117e-05,
+      "loss": 1.2676,
+      "step": 4237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5139207746238467e-05,
+      "loss": 1.2754,
+      "step": 4238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136923879329424e-05,
+      "loss": 1.2979,
+      "step": 4239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134639648366847e-05,
+      "loss": 1.2979,
+      "step": 4240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5132355053512617e-05,
+      "loss": 1.3203,
+      "step": 4241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130070094928647e-05,
+      "loss": 1.3193,
+      "step": 4242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5127784772776868e-05,
+      "loss": 1.292,
+      "step": 4243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512549908721924e-05,
+      "loss": 1.2158,
+      "step": 4244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5123213038417754e-05,
+      "loss": 1.25,
+      "step": 4245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512092662653442e-05,
+      "loss": 1.2773,
+      "step": 4246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118639851731276e-05,
+      "loss": 1.3203,
+      "step": 4247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116352714170383e-05,
+      "loss": 1.2852,
+      "step": 4248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114065214013834e-05,
+      "loss": 1.2656,
+      "step": 4249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111777351423741e-05,
+      "loss": 1.2754,
+      "step": 4250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109489126562248e-05,
+      "loss": 1.3184,
+      "step": 4251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510720053959152e-05,
+      "loss": 1.2705,
+      "step": 4252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104911590673745e-05,
+      "loss": 1.2412,
+      "step": 4253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510262227997115e-05,
+      "loss": 1.25,
+      "step": 4254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510033260764597e-05,
+      "loss": 1.3135,
+      "step": 4255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098042573860476e-05,
+      "loss": 1.3418,
+      "step": 4256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095752178776963e-05,
+      "loss": 1.3066,
+      "step": 4257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093461422557753e-05,
+      "loss": 1.2773,
+      "step": 4258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091170305365188e-05,
+      "loss": 1.3027,
+      "step": 4259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5088878827361644e-05,
+      "loss": 1.2969,
+      "step": 4260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086586988709515e-05,
+      "loss": 1.3193,
+      "step": 4261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508429478957122e-05,
+      "loss": 1.2734,
+      "step": 4262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5082002230109217e-05,
+      "loss": 1.2227,
+      "step": 4263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5079709310485973e-05,
+      "loss": 1.2949,
+      "step": 4264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077416030863984e-05,
+      "loss": 1.3223,
+      "step": 4265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075122391405779e-05,
+      "loss": 1.252,
+      "step": 4266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072828392273908e-05,
+      "loss": 1.2832,
+      "step": 4267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070534033630947e-05,
+      "loss": 1.2422,
+      "step": 4268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068239315639497e-05,
+      "loss": 1.2783,
+      "step": 4269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506594423846218e-05,
+      "loss": 1.2676,
+      "step": 4270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063648802261654e-05,
+      "loss": 1.2451,
+      "step": 4271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061353007200596e-05,
+      "loss": 1.2529,
+      "step": 4272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059056853441703e-05,
+      "loss": 1.2383,
+      "step": 4273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056760341147709e-05,
+      "loss": 1.2676,
+      "step": 4274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5054463470481366e-05,
+      "loss": 1.2607,
+      "step": 4275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052166241605455e-05,
+      "loss": 1.2256,
+      "step": 4276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5049868654682773e-05,
+      "loss": 1.292,
+      "step": 4277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047570709876162e-05,
+      "loss": 1.3184,
+      "step": 4278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045272407348468e-05,
+      "loss": 1.2734,
+      "step": 4279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042973747262576e-05,
+      "loss": 1.2998,
+      "step": 4280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040674729781385e-05,
+      "loss": 1.3086,
+      "step": 4281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038375355067836e-05,
+      "loss": 1.2793,
+      "step": 4282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036075623284881e-05,
+      "loss": 1.3193,
+      "step": 4283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033775534595498e-05,
+      "loss": 1.2939,
+      "step": 4284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031475089162701e-05,
+      "loss": 1.3066,
+      "step": 4285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5029174287149521e-05,
+      "loss": 1.2314,
+      "step": 4286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026873128719012e-05,
+      "loss": 1.3115,
+      "step": 4287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502457161403426e-05,
+      "loss": 1.2764,
+      "step": 4288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502226974325837e-05,
+      "loss": 1.2588,
+      "step": 4289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019967516554483e-05,
+      "loss": 1.3193,
+      "step": 4290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017664934085749e-05,
+      "loss": 1.2715,
+      "step": 4291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015361996015357e-05,
+      "loss": 1.2275,
+      "step": 4292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013058702506511e-05,
+      "loss": 1.2656,
+      "step": 4293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010755053722452e-05,
+      "loss": 1.291,
+      "step": 4294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008451049826435e-05,
+      "loss": 1.2969,
+      "step": 4295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500614669098175e-05,
+      "loss": 1.29,
+      "step": 4296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003841977351696e-05,
+      "loss": 1.2842,
+      "step": 4297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001536909099619e-05,
+      "loss": 1.249,
+      "step": 4298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999231486388875e-05,
+      "loss": 1.2832,
+      "step": 4299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4996925709382848e-05,
+      "loss": 1.2441,
+      "step": 4300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994619578244952e-05,
+      "loss": 1.2637,
+      "step": 4301
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4992313093138616e-05,
+      "loss": 1.3271,
+      "step": 4302
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4990006254227305e-05,
+      "loss": 1.2334,
+      "step": 4303
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4987699061674506e-05,
+      "loss": 1.2646,
+      "step": 4304
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985391515643723e-05,
+      "loss": 1.2793,
+      "step": 4305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.49830836162985e-05,
+      "loss": 1.2451,
+      "step": 4306
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980775363802391e-05,
+      "loss": 1.2979,
+      "step": 4307
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978466758318988e-05,
+      "loss": 1.3271,
+      "step": 4308
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976157800011892e-05,
+      "loss": 1.2939,
+      "step": 4309
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973848489044752e-05,
+      "loss": 1.3623,
+      "step": 4310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4971538825581217e-05,
+      "loss": 1.2725,
+      "step": 4311
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496922880978498e-05,
+      "loss": 1.2588,
+      "step": 4312
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496691844181975e-05,
+      "loss": 1.2959,
+      "step": 4313
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4964607721849258e-05,
+      "loss": 1.2939,
+      "step": 4314
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4962296650037271e-05,
+      "loss": 1.249,
+      "step": 4315
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495998522654757e-05,
+      "loss": 1.3154,
+      "step": 4316
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495767345154397e-05,
+      "loss": 1.2705,
+      "step": 4317
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955361325190305e-05,
+      "loss": 1.2881,
+      "step": 4318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4953048847650431e-05,
+      "loss": 1.3076,
+      "step": 4319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950736019088237e-05,
+      "loss": 1.291,
+      "step": 4320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948422839667634e-05,
+      "loss": 1.2383,
+      "step": 4321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946109309552555e-05,
+      "loss": 1.3037,
+      "step": 4322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943795428906955e-05,
+      "loss": 1.2793,
+      "step": 4323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4941481197894829e-05,
+      "loss": 1.2705,
+      "step": 4324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4939166616680179e-05,
+      "loss": 1.2734,
+      "step": 4325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4936851685427044e-05,
+      "loss": 1.2646,
+      "step": 4326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934536404299474e-05,
+      "loss": 1.2832,
+      "step": 4327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4932220773461565e-05,
+      "loss": 1.2588,
+      "step": 4328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929904793077416e-05,
+      "loss": 1.2891,
+      "step": 4329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927588463311169e-05,
+      "loss": 1.2861,
+      "step": 4330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925271784326971e-05,
+      "loss": 1.2725,
+      "step": 4331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922954756289017e-05,
+      "loss": 1.2842,
+      "step": 4332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920637379361507e-05,
+      "loss": 1.2832,
+      "step": 4333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918319653708674e-05,
+      "loss": 1.2793,
+      "step": 4334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916001579494779e-05,
+      "loss": 1.2891,
+      "step": 4335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913683156884103e-05,
+      "loss": 1.2842,
+      "step": 4336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911364386040945e-05,
+      "loss": 1.2617,
+      "step": 4337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909045267129645e-05,
+      "loss": 1.2734,
+      "step": 4338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4906725800314555e-05,
+      "loss": 1.2793,
+      "step": 4339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904405985760057e-05,
+      "loss": 1.3281,
+      "step": 4340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4902085823630557e-05,
+      "loss": 1.3418,
+      "step": 4341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489976531409048e-05,
+      "loss": 1.2812,
+      "step": 4342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897444457304284e-05,
+      "loss": 1.3057,
+      "step": 4343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489512325343645e-05,
+      "loss": 1.2666,
+      "step": 4344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892801702651475e-05,
+      "loss": 1.2539,
+      "step": 4345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4890479805113895e-05,
+      "loss": 1.3447,
+      "step": 4346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488815756098826e-05,
+      "loss": 1.2773,
+      "step": 4347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885834970439143e-05,
+      "loss": 1.248,
+      "step": 4348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488351203363115e-05,
+      "loss": 1.2969,
+      "step": 4349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881188750728909e-05,
+      "loss": 1.3027,
+      "step": 4350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487886512189707e-05,
+      "loss": 1.2861,
+      "step": 4351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4876541147300305e-05,
+      "loss": 1.293,
+      "step": 4352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874216827103313e-05,
+      "loss": 1.2891,
+      "step": 4353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4871892161470825e-05,
+      "loss": 1.292,
+      "step": 4354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869567150567588e-05,
+      "loss": 1.2754,
+      "step": 4355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867241794558371e-05,
+      "loss": 1.2373,
+      "step": 4356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864916093607975e-05,
+      "loss": 1.2979,
+      "step": 4357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862590047881222e-05,
+      "loss": 1.2715,
+      "step": 4358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860263657542962e-05,
+      "loss": 1.3027,
+      "step": 4359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857936922758058e-05,
+      "loss": 1.2725,
+      "step": 4360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855609843691414e-05,
+      "loss": 1.3145,
+      "step": 4361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4853282420507946e-05,
+      "loss": 1.249,
+      "step": 4362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4850954653372597e-05,
+      "loss": 1.2627,
+      "step": 4363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484862654245034e-05,
+      "loss": 1.291,
+      "step": 4364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4846298087906165e-05,
+      "loss": 1.3301,
+      "step": 4365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4843969289905089e-05,
+      "loss": 1.293,
+      "step": 4366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841640148612157e-05,
+      "loss": 1.2764,
+      "step": 4367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839310664192432e-05,
+      "loss": 1.2686,
+      "step": 4368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836980836811009e-05,
+      "loss": 1.2979,
+      "step": 4369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834650666632992e-05,
+      "loss": 1.2666,
+      "step": 4370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832320153823531e-05,
+      "loss": 1.2822,
+      "step": 4371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829989298547787e-05,
+      "loss": 1.2988,
+      "step": 4372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827658100970947e-05,
+      "loss": 1.2158,
+      "step": 4373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482532656125822e-05,
+      "loss": 1.2842,
+      "step": 4374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822994679574844e-05,
+      "loss": 1.2695,
+      "step": 4375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482066245608608e-05,
+      "loss": 1.2988,
+      "step": 4376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818329890957212e-05,
+      "loss": 1.2627,
+      "step": 4377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4815996984353548e-05,
+      "loss": 1.2812,
+      "step": 4378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813663736440425e-05,
+      "loss": 1.2715,
+      "step": 4379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811330147383196e-05,
+      "loss": 1.293,
+      "step": 4380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808996217347242e-05,
+      "loss": 1.2451,
+      "step": 4381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480666194649797e-05,
+      "loss": 1.2236,
+      "step": 4382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804327335000813e-05,
+      "loss": 1.2383,
+      "step": 4383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801992383021216e-05,
+      "loss": 1.2812,
+      "step": 4384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799657090724665e-05,
+      "loss": 1.249,
+      "step": 4385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797321458276657e-05,
+      "loss": 1.2734,
+      "step": 4386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794985485842724e-05,
+      "loss": 1.2832,
+      "step": 4387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792649173588412e-05,
+      "loss": 1.29,
+      "step": 4388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4790312521679294e-05,
+      "loss": 1.2266,
+      "step": 4389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478797553028097e-05,
+      "loss": 1.251,
+      "step": 4390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785638199559067e-05,
+      "loss": 1.2529,
+      "step": 4391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783300529679224e-05,
+      "loss": 1.2578,
+      "step": 4392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780962520807114e-05,
+      "loss": 1.2139,
+      "step": 4393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778624173108434e-05,
+      "loss": 1.3291,
+      "step": 4394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776285486748903e-05,
+      "loss": 1.2598,
+      "step": 4395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773946461894258e-05,
+      "loss": 1.2764,
+      "step": 4396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4771607098710272e-05,
+      "loss": 1.2686,
+      "step": 4397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4769267397362731e-05,
+      "loss": 1.2861,
+      "step": 4398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766927358017453e-05,
+      "loss": 1.3174,
+      "step": 4399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764586980840274e-05,
+      "loss": 1.2646,
+      "step": 4400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476224626599706e-05,
+      "loss": 1.2803,
+      "step": 4401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4759905213653691e-05,
+      "loss": 1.1982,
+      "step": 4402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475756382397608e-05,
+      "loss": 1.2236,
+      "step": 4403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755222097130163e-05,
+      "loss": 1.2656,
+      "step": 4404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752880033281902e-05,
+      "loss": 1.2949,
+      "step": 4405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750537632597268e-05,
+      "loss": 1.2402,
+      "step": 4406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748194895242275e-05,
+      "loss": 1.2852,
+      "step": 4407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474585182138295e-05,
+      "loss": 1.2871,
+      "step": 4408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743508411185348e-05,
+      "loss": 1.2637,
+      "step": 4409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741164664815544e-05,
+      "loss": 1.2764,
+      "step": 4410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738820582439641e-05,
+      "loss": 1.2637,
+      "step": 4411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736476164223764e-05,
+      "loss": 1.2959,
+      "step": 4412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734131410334062e-05,
+      "loss": 1.2744,
+      "step": 4413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4731786320936703e-05,
+      "loss": 1.2324,
+      "step": 4414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4729440896197892e-05,
+      "loss": 1.2705,
+      "step": 4415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4727095136283842e-05,
+      "loss": 1.2998,
+      "step": 4416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724749041360799e-05,
+      "loss": 1.2285,
+      "step": 4417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722402611595032e-05,
+      "loss": 1.2393,
+      "step": 4418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720055847152831e-05,
+      "loss": 1.2275,
+      "step": 4419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717708748200514e-05,
+      "loss": 1.25,
+      "step": 4420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4715361314904415e-05,
+      "loss": 1.2988,
+      "step": 4421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713013547430897e-05,
+      "loss": 1.332,
+      "step": 4422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4710665445946352e-05,
+      "loss": 1.252,
+      "step": 4423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4708317010617181e-05,
+      "loss": 1.2852,
+      "step": 4424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705968241609825e-05,
+      "loss": 1.2578,
+      "step": 4425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703619139090736e-05,
+      "loss": 1.2783,
+      "step": 4426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701269703226402e-05,
+      "loss": 1.2539,
+      "step": 4427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698919934183318e-05,
+      "loss": 1.2949,
+      "step": 4428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696569832128016e-05,
+      "loss": 1.208,
+      "step": 4429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469421939722705e-05,
+      "loss": 1.3115,
+      "step": 4430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691868629646995e-05,
+      "loss": 1.2734,
+      "step": 4431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689517529554442e-05,
+      "loss": 1.2881,
+      "step": 4432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687166097116026e-05,
+      "loss": 1.291,
+      "step": 4433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684814332498383e-05,
+      "loss": 1.2715,
+      "step": 4434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682462235868188e-05,
+      "loss": 1.2256,
+      "step": 4435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680109807392127e-05,
+      "loss": 1.2842,
+      "step": 4436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4677757047236926e-05,
+      "loss": 1.2998,
+      "step": 4437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675403955569318e-05,
+      "loss": 1.2627,
+      "step": 4438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467305053255607e-05,
+      "loss": 1.2686,
+      "step": 4439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670696778363967e-05,
+      "loss": 1.2861,
+      "step": 4440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466834269315982e-05,
+      "loss": 1.2773,
+      "step": 4441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4665988277110463e-05,
+      "loss": 1.2754,
+      "step": 4442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663633530382755e-05,
+      "loss": 1.3223,
+      "step": 4443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661278453143573e-05,
+      "loss": 1.3037,
+      "step": 4444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4658923045559824e-05,
+      "loss": 1.3242,
+      "step": 4445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656567307798435e-05,
+      "loss": 1.2812,
+      "step": 4446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4654211240026355e-05,
+      "loss": 1.2646,
+      "step": 4447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651854842410562e-05,
+      "loss": 1.2588,
+      "step": 4448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649498115118055e-05,
+      "loss": 1.2842,
+      "step": 4449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647141058315848e-05,
+      "loss": 1.3291,
+      "step": 4450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4644783672170992e-05,
+      "loss": 1.2969,
+      "step": 4451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642425956850553e-05,
+      "loss": 1.2861,
+      "step": 4452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640067912521618e-05,
+      "loss": 1.2646,
+      "step": 4453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637709539351307e-05,
+      "loss": 1.2988,
+      "step": 4454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4635350837506761e-05,
+      "loss": 1.2314,
+      "step": 4455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463299180715513e-05,
+      "loss": 1.2656,
+      "step": 4456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630632448463606e-05,
+      "loss": 1.249,
+      "step": 4457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628272761599399e-05,
+      "loss": 1.2744,
+      "step": 4458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4625912746729731e-05,
+      "loss": 1.2715,
+      "step": 4459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623552404021863e-05,
+      "loss": 1.2734,
+      "step": 4460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462119173364307e-05,
+      "loss": 1.2568,
+      "step": 4461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4618830735760653e-05,
+      "loss": 1.2754,
+      "step": 4462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461646941054194e-05,
+      "loss": 1.3311,
+      "step": 4463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461410775815427e-05,
+      "loss": 1.3018,
+      "step": 4464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611745778765016e-05,
+      "loss": 1.2695,
+      "step": 4465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609383472541575e-05,
+      "loss": 1.2363,
+      "step": 4466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.460702083965136e-05,
+      "loss": 1.2471,
+      "step": 4467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.460465788026181e-05,
+      "loss": 1.2686,
+      "step": 4468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4602294594540394e-05,
+      "loss": 1.2314,
+      "step": 4469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599930982654588e-05,
+      "loss": 1.3311,
+      "step": 4470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459756704477191e-05,
+      "loss": 1.2588,
+      "step": 4471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595202781059888e-05,
+      "loss": 1.2656,
+      "step": 4472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592838191686078e-05,
+      "loss": 1.2246,
+      "step": 4473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4590473276818058e-05,
+      "loss": 1.2734,
+      "step": 4474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588108036623428e-05,
+      "loss": 1.2881,
+      "step": 4475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4585742471269814e-05,
+      "loss": 1.2295,
+      "step": 4476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583376580924866e-05,
+      "loss": 1.2676,
+      "step": 4477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4581010365756252e-05,
+      "loss": 1.2549,
+      "step": 4478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578643825931661e-05,
+      "loss": 1.2705,
+      "step": 4479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457627696161882e-05,
+      "loss": 1.2393,
+      "step": 4480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.457390977298546e-05,
+      "loss": 1.2803,
+      "step": 4481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571542260199347e-05,
+      "loss": 1.3125,
+      "step": 4482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569174423428265e-05,
+      "loss": 1.2705,
+      "step": 4483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566806262840025e-05,
+      "loss": 1.291,
+      "step": 4484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4564437778602453e-05,
+      "loss": 1.2822,
+      "step": 4485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456206897088341e-05,
+      "loss": 1.2998,
+      "step": 4486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455969983985077e-05,
+      "loss": 1.2881,
+      "step": 4487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557330385672435e-05,
+      "loss": 1.2676,
+      "step": 4488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4554960608516324e-05,
+      "loss": 1.291,
+      "step": 4489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552590508550387e-05,
+      "loss": 1.2598,
+      "step": 4490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455022008594259e-05,
+      "loss": 1.2832,
+      "step": 4491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547849340860929e-05,
+      "loss": 1.248,
+      "step": 4492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545478273473413e-05,
+      "loss": 1.2539,
+      "step": 4493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543106883948084e-05,
+      "loss": 1.2695,
+      "step": 4494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540735172452998e-05,
+      "loss": 1.2637,
+      "step": 4495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538363139156244e-05,
+      "loss": 1.2432,
+      "step": 4496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4535990784225919e-05,
+      "loss": 1.3018,
+      "step": 4497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533618107830162e-05,
+      "loss": 1.2236,
+      "step": 4498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4531245110137118e-05,
+      "loss": 1.2412,
+      "step": 4499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528871791314962e-05,
+      "loss": 1.2529,
+      "step": 4500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526498151531892e-05,
+      "loss": 1.3086,
+      "step": 4501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452412419095613e-05,
+      "loss": 1.2383,
+      "step": 4502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4521749909755914e-05,
+      "loss": 1.2637,
+      "step": 4503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451937530809951e-05,
+      "loss": 1.2842,
+      "step": 4504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517000386155209e-05,
+      "loss": 1.2842,
+      "step": 4505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514625144091319e-05,
+      "loss": 1.2549,
+      "step": 4506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4512249582076171e-05,
+      "loss": 1.2852,
+      "step": 4507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450987370027813e-05,
+      "loss": 1.2871,
+      "step": 4508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4507497498865563e-05,
+      "loss": 1.3008,
+      "step": 4509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505120978006881e-05,
+      "loss": 1.3223,
+      "step": 4510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4502744137870502e-05,
+      "loss": 1.2588,
+      "step": 4511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4500366978624872e-05,
+      "loss": 1.2881,
+      "step": 4512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497989500438464e-05,
+      "loss": 1.2871,
+      "step": 4513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495611703479767e-05,
+      "loss": 1.3037,
+      "step": 4514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493233587917297e-05,
+      "loss": 1.2871,
+      "step": 4515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4490855153919592e-05,
+      "loss": 1.2998,
+      "step": 4516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488476401655208e-05,
+      "loss": 1.2617,
+      "step": 4517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486097331292733e-05,
+      "loss": 1.2236,
+      "step": 4518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483717943000763e-05,
+      "loss": 1.3105,
+      "step": 4519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4481338236947933e-05,
+      "loss": 1.29,
+      "step": 4520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478958213302888e-05,
+      "loss": 1.2041,
+      "step": 4521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476577872234298e-05,
+      "loss": 1.2334,
+      "step": 4522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4474197213910866e-05,
+      "loss": 1.2588,
+      "step": 4523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4471816238501301e-05,
+      "loss": 1.2744,
+      "step": 4524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469434946174348e-05,
+      "loss": 1.3018,
+      "step": 4525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4467053337098766e-05,
+      "loss": 1.3105,
+      "step": 4526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446467141144334e-05,
+      "loss": 1.3242,
+      "step": 4527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446228916937688e-05,
+      "loss": 1.2617,
+      "step": 4528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4459906611068211e-05,
+      "loss": 1.2559,
+      "step": 4529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457523736686185e-05,
+      "loss": 1.29,
+      "step": 4530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445514054639968e-05,
+      "loss": 1.2822,
+      "step": 4531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445275704037759e-05,
+      "loss": 1.2334,
+      "step": 4532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450373218788831e-05,
+      "loss": 1.2529,
+      "step": 4533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447989081802354e-05,
+      "loss": 1.3076,
+      "step": 4534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445604629587112e-05,
+      "loss": 1.2646,
+      "step": 4535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443219862312095e-05,
+      "loss": 1.2734,
+      "step": 4536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440834780146312e-05,
+      "loss": 1.249,
+      "step": 4537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438449383258796e-05,
+      "loss": 1.2607,
+      "step": 4538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4436063671818594e-05,
+      "loss": 1.3047,
+      "step": 4539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433677645994786e-05,
+      "loss": 1.2588,
+      "step": 4540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431291305956468e-05,
+      "loss": 1.2979,
+      "step": 4541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4428904651872759e-05,
+      "loss": 1.2402,
+      "step": 4542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4426517683912802e-05,
+      "loss": 1.2617,
+      "step": 4543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424130402245762e-05,
+      "loss": 1.2393,
+      "step": 4544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421742807040823e-05,
+      "loss": 1.2803,
+      "step": 4545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419354898467197e-05,
+      "loss": 1.2988,
+      "step": 4546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416966676694113e-05,
+      "loss": 1.3115,
+      "step": 4547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414578141890823e-05,
+      "loss": 1.2041,
+      "step": 4548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4412189294226607e-05,
+      "loss": 1.3711,
+      "step": 4549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409800133870758e-05,
+      "loss": 1.2637,
+      "step": 4550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4407410660992595e-05,
+      "loss": 1.3057,
+      "step": 4551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405020875761464e-05,
+      "loss": 1.2695,
+      "step": 4552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402630778346727e-05,
+      "loss": 1.2715,
+      "step": 4553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400240368917768e-05,
+      "loss": 1.2646,
+      "step": 4554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397849647644e-05,
+      "loss": 1.3027,
+      "step": 4555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439545861469485e-05,
+      "loss": 1.2822,
+      "step": 4556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393067270239769e-05,
+      "loss": 1.2451,
+      "step": 4557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390675614448233e-05,
+      "loss": 1.2529,
+      "step": 4558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438828364748974e-05,
+      "loss": 1.291,
+      "step": 4559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385891369533808e-05,
+      "loss": 1.249,
+      "step": 4560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4383498780749976e-05,
+      "loss": 1.2432,
+      "step": 4561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381105881307806e-05,
+      "loss": 1.2832,
+      "step": 4562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378712671376887e-05,
+      "loss": 1.2627,
+      "step": 4563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4376319151126822e-05,
+      "loss": 1.2705,
+      "step": 4564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373925320727241e-05,
+      "loss": 1.2354,
+      "step": 4565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371531180347795e-05,
+      "loss": 1.3457,
+      "step": 4566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369136730158153e-05,
+      "loss": 1.2764,
+      "step": 4567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366741970328018e-05,
+      "loss": 1.2881,
+      "step": 4568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364346901027093e-05,
+      "loss": 1.2891,
+      "step": 4569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361951522425131e-05,
+      "loss": 1.2373,
+      "step": 4570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359555834691883e-05,
+      "loss": 1.2627,
+      "step": 4571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4357159837997135e-05,
+      "loss": 1.2715,
+      "step": 4572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354763532510687e-05,
+      "loss": 1.2949,
+      "step": 4573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352366918402373e-05,
+      "loss": 1.25,
+      "step": 4574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349969995842033e-05,
+      "loss": 1.2676,
+      "step": 4575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434757276499954e-05,
+      "loss": 1.2529,
+      "step": 4576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345175226044785e-05,
+      "loss": 1.2412,
+      "step": 4577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342777379147685e-05,
+      "loss": 1.2559,
+      "step": 4578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340379224478171e-05,
+      "loss": 1.2705,
+      "step": 4579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.43379807622062e-05,
+      "loss": 1.2637,
+      "step": 4580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335581992501752e-05,
+      "loss": 1.2959,
+      "step": 4581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333182915534834e-05,
+      "loss": 1.2783,
+      "step": 4582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330783531475456e-05,
+      "loss": 1.2793,
+      "step": 4583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432838384049367e-05,
+      "loss": 1.2764,
+      "step": 4584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432598384275954e-05,
+      "loss": 1.3057,
+      "step": 4585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4323583538443157e-05,
+      "loss": 1.2588,
+      "step": 4586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321182927714625e-05,
+      "loss": 1.2656,
+      "step": 4587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318782010744081e-05,
+      "loss": 1.3115,
+      "step": 4588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4316380787701672e-05,
+      "loss": 1.2725,
+      "step": 4589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4313979258757576e-05,
+      "loss": 1.2988,
+      "step": 4590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4311577424081987e-05,
+      "loss": 1.2666,
+      "step": 4591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4309175283845132e-05,
+      "loss": 1.2822,
+      "step": 4592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306772838217235e-05,
+      "loss": 1.3086,
+      "step": 4593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304370087368568e-05,
+      "loss": 1.2266,
+      "step": 4594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430196703146941e-05,
+      "loss": 1.251,
+      "step": 4595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4299563670690071e-05,
+      "loss": 1.2627,
+      "step": 4596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297160005200867e-05,
+      "loss": 1.2998,
+      "step": 4597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294756035172154e-05,
+      "loss": 1.2256,
+      "step": 4598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292351760774298e-05,
+      "loss": 1.2988,
+      "step": 4599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4289947182177692e-05,
+      "loss": 1.2656,
+      "step": 4600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287542299552746e-05,
+      "loss": 1.2861,
+      "step": 4601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428513711306989e-05,
+      "loss": 1.3252,
+      "step": 4602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282731622899589e-05,
+      "loss": 1.2646,
+      "step": 4603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280325829212315e-05,
+      "loss": 1.2666,
+      "step": 4604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277919732178562e-05,
+      "loss": 1.2725,
+      "step": 4605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427551333196886e-05,
+      "loss": 1.3262,
+      "step": 4606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4273106628753743e-05,
+      "loss": 1.2686,
+      "step": 4607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4270699622703774e-05,
+      "loss": 1.2539,
+      "step": 4608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426829231398954e-05,
+      "loss": 1.2979,
+      "step": 4609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265884702781647e-05,
+      "loss": 1.2959,
+      "step": 4610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263476789250724e-05,
+      "loss": 1.3066,
+      "step": 4611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261068573567414e-05,
+      "loss": 1.2314,
+      "step": 4612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258660055902389e-05,
+      "loss": 1.2754,
+      "step": 4613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4256251236426344e-05,
+      "loss": 1.293,
+      "step": 4614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253842115309989e-05,
+      "loss": 1.2646,
+      "step": 4615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251432692724056e-05,
+      "loss": 1.291,
+      "step": 4616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249022968839308e-05,
+      "loss": 1.2646,
+      "step": 4617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4246612943826517e-05,
+      "loss": 1.2832,
+      "step": 4618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244202617856481e-05,
+      "loss": 1.3115,
+      "step": 4619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424179199110002e-05,
+      "loss": 1.2744,
+      "step": 4620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4239381063727977e-05,
+      "loss": 1.2832,
+      "step": 4621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236969835911211e-05,
+      "loss": 1.2549,
+      "step": 4622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423455830782061e-05,
+      "loss": 1.2822,
+      "step": 4623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232146479627073e-05,
+      "loss": 1.2695,
+      "step": 4624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4229734351501534e-05,
+      "loss": 1.2373,
+      "step": 4625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227321923614933e-05,
+      "loss": 1.2002,
+      "step": 4626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422490919613824e-05,
+      "loss": 1.2773,
+      "step": 4627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422249616924245e-05,
+      "loss": 1.3037,
+      "step": 4628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4220082843098571e-05,
+      "loss": 1.2119,
+      "step": 4629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217669217877633e-05,
+      "loss": 1.2598,
+      "step": 4630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421525529375069e-05,
+      "loss": 1.2822,
+      "step": 4631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212841070888819e-05,
+      "loss": 1.2266,
+      "step": 4632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4210426549463117e-05,
+      "loss": 1.2617,
+      "step": 4633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208011729644694e-05,
+      "loss": 1.2549,
+      "step": 4634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4205596611604699e-05,
+      "loss": 1.2969,
+      "step": 4635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4203181195514282e-05,
+      "loss": 1.2363,
+      "step": 4636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200765481544628e-05,
+      "loss": 1.2676,
+      "step": 4637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198349469866935e-05,
+      "loss": 1.2783,
+      "step": 4638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419593316065243e-05,
+      "loss": 1.2695,
+      "step": 4639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193516554072355e-05,
+      "loss": 1.2686,
+      "step": 4640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191099650297972e-05,
+      "loss": 1.2129,
+      "step": 4641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418868244950057e-05,
+      "loss": 1.3164,
+      "step": 4642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4186264951851458e-05,
+      "loss": 1.3047,
+      "step": 4643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418384715752196e-05,
+      "loss": 1.2373,
+      "step": 4644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181429066683423e-05,
+      "loss": 1.2598,
+      "step": 4645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4179010679507222e-05,
+      "loss": 1.2803,
+      "step": 4646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176591996164747e-05,
+      "loss": 1.2119,
+      "step": 4647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174173016827409e-05,
+      "loss": 1.3096,
+      "step": 4648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171753741666638e-05,
+      "loss": 1.2783,
+      "step": 4649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169334170853892e-05,
+      "loss": 1.2744,
+      "step": 4650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166914304560649e-05,
+      "loss": 1.2139,
+      "step": 4651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164494142958394e-05,
+      "loss": 1.3291,
+      "step": 4652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4162073686218654e-05,
+      "loss": 1.3076,
+      "step": 4653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159652934512961e-05,
+      "loss": 1.249,
+      "step": 4654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157231888012876e-05,
+      "loss": 1.2412,
+      "step": 4655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415481054688998e-05,
+      "loss": 1.2578,
+      "step": 4656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4152388911315869e-05,
+      "loss": 1.2676,
+      "step": 4657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149966981462167e-05,
+      "loss": 1.3086,
+      "step": 4658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147544757500513e-05,
+      "loss": 1.2979,
+      "step": 4659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145122239602576e-05,
+      "loss": 1.2637,
+      "step": 4660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142699427940037e-05,
+      "loss": 1.333,
+      "step": 4661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4140276322684594e-05,
+      "loss": 1.2393,
+      "step": 4662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4137852924007982e-05,
+      "loss": 1.2266,
+      "step": 4663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135429232081942e-05,
+      "loss": 1.2686,
+      "step": 4664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4133005247078244e-05,
+      "loss": 1.2852,
+      "step": 4665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413058096916867e-05,
+      "loss": 1.2715,
+      "step": 4666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4128156398525032e-05,
+      "loss": 1.2412,
+      "step": 4667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4125731535319162e-05,
+      "loss": 1.2871,
+      "step": 4668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123306379722907e-05,
+      "loss": 1.2812,
+      "step": 4669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4120880931908136e-05,
+      "loss": 1.2393,
+      "step": 4670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118455192046746e-05,
+      "loss": 1.2852,
+      "step": 4671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4116029160310642e-05,
+      "loss": 1.2344,
+      "step": 4672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411360283687176e-05,
+      "loss": 1.2686,
+      "step": 4673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4111176221902055e-05,
+      "loss": 1.3291,
+      "step": 4674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41087493155735e-05,
+      "loss": 1.2715,
+      "step": 4675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4106322118058086e-05,
+      "loss": 1.2832,
+      "step": 4676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4103894629527835e-05,
+      "loss": 1.2803,
+      "step": 4677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410146685015478e-05,
+      "loss": 1.2607,
+      "step": 4678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4099038780110977e-05,
+      "loss": 1.2666,
+      "step": 4679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.40966104195685e-05,
+      "loss": 1.2588,
+      "step": 4680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4094181768699453e-05,
+      "loss": 1.2412,
+      "step": 4681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409175282767595e-05,
+      "loss": 1.3193,
+      "step": 4682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089323596670134e-05,
+      "loss": 1.2539,
+      "step": 4683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086894075854161e-05,
+      "loss": 1.2598,
+      "step": 4684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408446426540021e-05,
+      "loss": 1.2734,
+      "step": 4685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4082034165480485e-05,
+      "loss": 1.3027,
+      "step": 4686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079603776267208e-05,
+      "loss": 1.3164,
+      "step": 4687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4077173097932615e-05,
+      "loss": 1.2861,
+      "step": 4688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074742130648973e-05,
+      "loss": 1.2295,
+      "step": 4689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072310874588562e-05,
+      "loss": 1.2686,
+      "step": 4690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4069879329923687e-05,
+      "loss": 1.2549,
+      "step": 4691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4067447496826667e-05,
+      "loss": 1.2256,
+      "step": 4692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065015375469852e-05,
+      "loss": 1.3008,
+      "step": 4693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062582966025603e-05,
+      "loss": 1.2539,
+      "step": 4694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4060150268666304e-05,
+      "loss": 1.292,
+      "step": 4695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4057717283564363e-05,
+      "loss": 1.3096,
+      "step": 4696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4055284010892205e-05,
+      "loss": 1.2539,
+      "step": 4697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405285045082227e-05,
+      "loss": 1.2529,
+      "step": 4698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4050416603527034e-05,
+      "loss": 1.2646,
+      "step": 4699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047982469178974e-05,
+      "loss": 1.2793,
+      "step": 4700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045548047950606e-05,
+      "loss": 1.3311,
+      "step": 4701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4043113340014451e-05,
+      "loss": 1.2549,
+      "step": 4702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040678345543058e-05,
+      "loss": 1.2666,
+      "step": 4703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038243064708995e-05,
+      "loss": 1.3105,
+      "step": 4704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4035807497684855e-05,
+      "loss": 1.2422,
+      "step": 4705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033371644643235e-05,
+      "loss": 1.29,
+      "step": 4706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030935505756778e-05,
+      "loss": 1.2588,
+      "step": 4707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4028499081198122e-05,
+      "loss": 1.2207,
+      "step": 4708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402606237113994e-05,
+      "loss": 1.3086,
+      "step": 4709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023625375754924e-05,
+      "loss": 1.2568,
+      "step": 4710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4021188095215781e-05,
+      "loss": 1.2334,
+      "step": 4711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401875052969524e-05,
+      "loss": 1.3242,
+      "step": 4712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016312679366054e-05,
+      "loss": 1.2471,
+      "step": 4713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013874544400991e-05,
+      "loss": 1.251,
+      "step": 4714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4011436124972847e-05,
+      "loss": 1.2656,
+      "step": 4715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008997421254424e-05,
+      "loss": 1.2432,
+      "step": 4716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006558433418557e-05,
+      "loss": 1.2607,
+      "step": 4717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004119161638097e-05,
+      "loss": 1.2744,
+      "step": 4718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001679606085919e-05,
+      "loss": 1.335,
+      "step": 4719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3999239766934904e-05,
+      "loss": 1.2324,
+      "step": 4720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996799644357972e-05,
+      "loss": 1.2314,
+      "step": 4721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3994359238528053e-05,
+      "loss": 1.2656,
+      "step": 4722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991918549618097e-05,
+      "loss": 1.2402,
+      "step": 4723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989477577801075e-05,
+      "loss": 1.2812,
+      "step": 4724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398703632324998e-05,
+      "loss": 1.2354,
+      "step": 4725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984594786137821e-05,
+      "loss": 1.2793,
+      "step": 4726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3982152966637635e-05,
+      "loss": 1.2842,
+      "step": 4727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979710864922466e-05,
+      "loss": 1.2686,
+      "step": 4728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977268481165394e-05,
+      "loss": 1.3223,
+      "step": 4729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974825815539504e-05,
+      "loss": 1.3145,
+      "step": 4730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972382868217908e-05,
+      "loss": 1.2549,
+      "step": 4731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969939639373742e-05,
+      "loss": 1.2988,
+      "step": 4732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3967496129180157e-05,
+      "loss": 1.2529,
+      "step": 4733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965052337810319e-05,
+      "loss": 1.2451,
+      "step": 4734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962608265437423e-05,
+      "loss": 1.249,
+      "step": 4735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960163912234682e-05,
+      "loss": 1.3047,
+      "step": 4736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3957719278375325e-05,
+      "loss": 1.2852,
+      "step": 4737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955274364032603e-05,
+      "loss": 1.2646,
+      "step": 4738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952829169379787e-05,
+      "loss": 1.2793,
+      "step": 4739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950383694590167e-05,
+      "loss": 1.2725,
+      "step": 4740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947937939837055e-05,
+      "loss": 1.2324,
+      "step": 4741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945491905293782e-05,
+      "loss": 1.2441,
+      "step": 4742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943045591133699e-05,
+      "loss": 1.209,
+      "step": 4743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3940598997530174e-05,
+      "loss": 1.3027,
+      "step": 4744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938152124656599e-05,
+      "loss": 1.2598,
+      "step": 4745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935704972686379e-05,
+      "loss": 1.252,
+      "step": 4746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933257541792951e-05,
+      "loss": 1.2969,
+      "step": 4747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393080983214976e-05,
+      "loss": 1.2793,
+      "step": 4748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3928361843930274e-05,
+      "loss": 1.2529,
+      "step": 4749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925913577307984e-05,
+      "loss": 1.2627,
+      "step": 4750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3923465032456396e-05,
+      "loss": 1.292,
+      "step": 4751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921016209549043e-05,
+      "loss": 1.2715,
+      "step": 4752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918567108759466e-05,
+      "loss": 1.2383,
+      "step": 4753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3916117730261241e-05,
+      "loss": 1.2617,
+      "step": 4754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913668074227948e-05,
+      "loss": 1.2295,
+      "step": 4755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911218140833201e-05,
+      "loss": 1.2627,
+      "step": 4756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908767930250617e-05,
+      "loss": 1.2676,
+      "step": 4757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3906317442653852e-05,
+      "loss": 1.208,
+      "step": 4758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903866678216566e-05,
+      "loss": 1.2568,
+      "step": 4759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3901415637112448e-05,
+      "loss": 1.2578,
+      "step": 4760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898964319515199e-05,
+      "loss": 1.2529,
+      "step": 4761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389651272559855e-05,
+      "loss": 1.2705,
+      "step": 4762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389406085553624e-05,
+      "loss": 1.2607,
+      "step": 4763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3891608709502032e-05,
+      "loss": 1.2568,
+      "step": 4764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3889156287669715e-05,
+      "loss": 1.252,
+      "step": 4765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3886703590213088e-05,
+      "loss": 1.2881,
+      "step": 4766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3884250617305974e-05,
+      "loss": 1.2939,
+      "step": 4767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881797369122218e-05,
+      "loss": 1.2891,
+      "step": 4768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879343845835678e-05,
+      "loss": 1.2461,
+      "step": 4769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876890047620237e-05,
+      "loss": 1.2871,
+      "step": 4770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874435974649792e-05,
+      "loss": 1.2686,
+      "step": 4771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3871981627098268e-05,
+      "loss": 1.3027,
+      "step": 4772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869527005139603e-05,
+      "loss": 1.2939,
+      "step": 4773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867072108947754e-05,
+      "loss": 1.2852,
+      "step": 4774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38646169386967e-05,
+      "loss": 1.2363,
+      "step": 4775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3862161494560441e-05,
+      "loss": 1.2432,
+      "step": 4776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859705776712993e-05,
+      "loss": 1.2188,
+      "step": 4777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857249785328394e-05,
+      "loss": 1.3076,
+      "step": 4778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3854793520580694e-05,
+      "loss": 1.3027,
+      "step": 4779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852336982643975e-05,
+      "loss": 1.2627,
+      "step": 4780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384988017169233e-05,
+      "loss": 1.2803,
+      "step": 4781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384742308789987e-05,
+      "loss": 1.25,
+      "step": 4782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844965731440732e-05,
+      "loss": 1.2793,
+      "step": 4783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384250810248907e-05,
+      "loss": 1.248,
+      "step": 4784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840050201219048e-05,
+      "loss": 1.2764,
+      "step": 4785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837592027804867e-05,
+      "loss": 1.3047,
+      "step": 4786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3835133582420733e-05,
+      "loss": 1.2773,
+      "step": 4787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832674865240879e-05,
+      "loss": 1.2334,
+      "step": 4788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830215876439544e-05,
+      "loss": 1.2959,
+      "step": 4789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382775661619101e-05,
+      "loss": 1.3047,
+      "step": 4790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825297084669557e-05,
+      "loss": 1.2832,
+      "step": 4791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3822837282049497e-05,
+      "loss": 1.2422,
+      "step": 4792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3820377208505147e-05,
+      "loss": 1.2695,
+      "step": 4793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3817916864210861e-05,
+      "loss": 1.2588,
+      "step": 4794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815456249340999e-05,
+      "loss": 1.3174,
+      "step": 4795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3812995364069947e-05,
+      "loss": 1.2422,
+      "step": 4796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810534208572106e-05,
+      "loss": 1.2773,
+      "step": 4797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3808072783021903e-05,
+      "loss": 1.2441,
+      "step": 4798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3805611087593768e-05,
+      "loss": 1.2471,
+      "step": 4799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803149122462171e-05,
+      "loss": 1.2822,
+      "step": 4800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3800686887801592e-05,
+      "loss": 1.2783,
+      "step": 4801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3798224383786522e-05,
+      "loss": 1.2988,
+      "step": 4802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795761610591485e-05,
+      "loss": 1.3232,
+      "step": 4803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793298568391014e-05,
+      "loss": 1.2725,
+      "step": 4804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790835257359663e-05,
+      "loss": 1.2461,
+      "step": 4805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788371677672015e-05,
+      "loss": 1.2285,
+      "step": 4806
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785907829502652e-05,
+      "loss": 1.2734,
+      "step": 4807
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783443713026196e-05,
+      "loss": 1.2822,
+      "step": 4808
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780979328417276e-05,
+      "loss": 1.2891,
+      "step": 4809
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778514675850545e-05,
+      "loss": 1.248,
+      "step": 4810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776049755500667e-05,
+      "loss": 1.2725,
+      "step": 4811
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773584567542335e-05,
+      "loss": 1.2676,
+      "step": 4812
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3771119112150257e-05,
+      "loss": 1.2969,
+      "step": 4813
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768653389499158e-05,
+      "loss": 1.2256,
+      "step": 4814
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766187399763784e-05,
+      "loss": 1.2246,
+      "step": 4815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3763721143118902e-05,
+      "loss": 1.3008,
+      "step": 4816
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376125461973929e-05,
+      "loss": 1.2861,
+      "step": 4817
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758787829799756e-05,
+      "loss": 1.2793,
+      "step": 4818
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756320773475117e-05,
+      "loss": 1.2871,
+      "step": 4819
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375385345094022e-05,
+      "loss": 1.2646,
+      "step": 4820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3751385862369914e-05,
+      "loss": 1.2148,
+      "step": 4821
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748918007939085e-05,
+      "loss": 1.2637,
+      "step": 4822
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746449887822625e-05,
+      "loss": 1.29,
+      "step": 4823
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3743981502195455e-05,
+      "loss": 1.3057,
+      "step": 4824
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.37415128512325e-05,
+      "loss": 1.2773,
+      "step": 4825
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739043935108725e-05,
+      "loss": 1.291,
+      "step": 4826
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373657475399909e-05,
+      "loss": 1.2695,
+      "step": 4827
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734105308078595e-05,
+      "loss": 1.2959,
+      "step": 4828
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3731635597522245e-05,
+      "loss": 1.2217,
+      "step": 4829
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729165622505072e-05,
+      "loss": 1.2588,
+      "step": 4830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3726695383202118e-05,
+      "loss": 1.2822,
+      "step": 4831
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724224879788452e-05,
+      "loss": 1.2246,
+      "step": 4832
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721754112439157e-05,
+      "loss": 1.2461,
+      "step": 4833
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3719283081329338e-05,
+      "loss": 1.2012,
+      "step": 4834
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716811786634115e-05,
+      "loss": 1.2559,
+      "step": 4835
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714340228528626e-05,
+      "loss": 1.2822,
+      "step": 4836
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3711868407188036e-05,
+      "loss": 1.2285,
+      "step": 4837
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370939632278752e-05,
+      "loss": 1.2637,
+      "step": 4838
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706923975502276e-05,
+      "loss": 1.3281,
+      "step": 4839
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704451365507515e-05,
+      "loss": 1.2539,
+      "step": 4840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701978492978476e-05,
+      "loss": 1.2197,
+      "step": 4841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699505358090407e-05,
+      "loss": 1.2441,
+      "step": 4842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369703196101858e-05,
+      "loss": 1.2617,
+      "step": 4843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369455830193829e-05,
+      "loss": 1.2441,
+      "step": 4844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3692084381024837e-05,
+      "loss": 1.2959,
+      "step": 4845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368961019845355e-05,
+      "loss": 1.2637,
+      "step": 4846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3687135754399776e-05,
+      "loss": 1.2656,
+      "step": 4847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368466104903888e-05,
+      "loss": 1.2705,
+      "step": 4848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682186082546243e-05,
+      "loss": 1.29,
+      "step": 4849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679710855097261e-05,
+      "loss": 1.2695,
+      "step": 4850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677235366867358e-05,
+      "loss": 1.2246,
+      "step": 4851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674759618031974e-05,
+      "loss": 1.2734,
+      "step": 4852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672283608766562e-05,
+      "loss": 1.2725,
+      "step": 4853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3669807339246595e-05,
+      "loss": 1.2656,
+      "step": 4854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667330809647568e-05,
+      "loss": 1.2617,
+      "step": 4855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3664854020144992e-05,
+      "loss": 1.2754,
+      "step": 4856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36623769709144e-05,
+      "loss": 1.3096,
+      "step": 4857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3659899662131335e-05,
+      "loss": 1.252,
+      "step": 4858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657422093971366e-05,
+      "loss": 1.2422,
+      "step": 4859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654944266610082e-05,
+      "loss": 1.2334,
+      "step": 4860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652466180223079e-05,
+      "loss": 1.3037,
+      "step": 4861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3649987834985985e-05,
+      "loss": 1.2617,
+      "step": 4862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647509231074438e-05,
+      "loss": 1.2891,
+      "step": 4863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3645030368664097e-05,
+      "loss": 1.2666,
+      "step": 4864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642551247930638e-05,
+      "loss": 1.2373,
+      "step": 4865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640071869049759e-05,
+      "loss": 1.29,
+      "step": 4866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3637592232197168e-05,
+      "loss": 1.2871,
+      "step": 4867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36351123375486e-05,
+      "loss": 1.2124,
+      "step": 4868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632632185279805e-05,
+      "loss": 1.2461,
+      "step": 4869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630151775566556e-05,
+      "loss": 1.2754,
+      "step": 4870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362767110858463e-05,
+      "loss": 1.1943,
+      "step": 4871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625190184509838e-05,
+      "loss": 1.2725,
+      "step": 4872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622709003518e-05,
+      "loss": 1.2275,
+      "step": 4873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3620227565784959e-05,
+      "loss": 1.2725,
+      "step": 4874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3617745871486575e-05,
+      "loss": 1.3027,
+      "step": 4875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615263920798723e-05,
+      "loss": 1.2627,
+      "step": 4876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36127817138973e-05,
+      "loss": 1.252,
+      "step": 4877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610299250958222e-05,
+      "loss": 1.2607,
+      "step": 4878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607816532157418e-05,
+      "loss": 1.2764,
+      "step": 4879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360533355767084e-05,
+      "loss": 1.3027,
+      "step": 4880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602850327674455e-05,
+      "loss": 1.2334,
+      "step": 4881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360036684234425e-05,
+      "loss": 1.2568,
+      "step": 4882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597883101856228e-05,
+      "loss": 1.2539,
+      "step": 4883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3595399106386416e-05,
+      "loss": 1.2686,
+      "step": 4884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359291485611085e-05,
+      "loss": 1.2217,
+      "step": 4885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590430351205591e-05,
+      "loss": 1.2783,
+      "step": 4886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587945591846712e-05,
+      "loss": 1.2695,
+      "step": 4887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3585460578210314e-05,
+      "loss": 1.3027,
+      "step": 4888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3582975310472505e-05,
+      "loss": 1.249,
+      "step": 4889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3580489788809417e-05,
+      "loss": 1.2725,
+      "step": 4890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578004013397204e-05,
+      "loss": 1.2686,
+      "step": 4891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357551798441202e-05,
+      "loss": 1.2871,
+      "step": 4892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573031702030067e-05,
+      "loss": 1.2676,
+      "step": 4893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570545166427529e-05,
+      "loss": 1.2246,
+      "step": 4894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3568058377780642e-05,
+      "loss": 1.2031,
+      "step": 4895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565571336265638e-05,
+      "loss": 1.249,
+      "step": 4896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3563084042058772e-05,
+      "loss": 1.2979,
+      "step": 4897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356059649533632e-05,
+      "loss": 1.2744,
+      "step": 4898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355810869627458e-05,
+      "loss": 1.2471,
+      "step": 4899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555620645049854e-05,
+      "loss": 1.2822,
+      "step": 4900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355313234183847e-05,
+      "loss": 1.2471,
+      "step": 4901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550643786816782e-05,
+      "loss": 1.29,
+      "step": 4902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3548154980161147e-05,
+      "loss": 1.2754,
+      "step": 4903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3545665922047948e-05,
+      "loss": 1.3164,
+      "step": 4904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543176612653586e-05,
+      "loss": 1.3105,
+      "step": 4905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540687052154473e-05,
+      "loss": 1.2383,
+      "step": 4906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538197240727054e-05,
+      "loss": 1.2539,
+      "step": 4907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353570717854777e-05,
+      "loss": 1.2168,
+      "step": 4908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3533216865793102e-05,
+      "loss": 1.2822,
+      "step": 4909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3530726302639533e-05,
+      "loss": 1.2686,
+      "step": 4910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528235489263572e-05,
+      "loss": 1.2559,
+      "step": 4911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3525744425841735e-05,
+      "loss": 1.2539,
+      "step": 4912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523253112550575e-05,
+      "loss": 1.2256,
+      "step": 4913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520761549566641e-05,
+      "loss": 1.2686,
+      "step": 4914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3518269737066516e-05,
+      "loss": 1.2842,
+      "step": 4915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3515777675226794e-05,
+      "loss": 1.2637,
+      "step": 4916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513285364224087e-05,
+      "loss": 1.2832,
+      "step": 4917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510792804235023e-05,
+      "loss": 1.2773,
+      "step": 4918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508299995436252e-05,
+      "loss": 1.2725,
+      "step": 4919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3505806938004435e-05,
+      "loss": 1.2529,
+      "step": 4920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503313632116264e-05,
+      "loss": 1.2734,
+      "step": 4921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3500820077948428e-05,
+      "loss": 1.292,
+      "step": 4922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349832627567765e-05,
+      "loss": 1.2451,
+      "step": 4923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495832225480672e-05,
+      "loss": 1.251,
+      "step": 4924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3493337927534238e-05,
+      "loss": 1.2471,
+      "step": 4925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349084338201512e-05,
+      "loss": 1.2656,
+      "step": 4926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488348589100113e-05,
+      "loss": 1.333,
+      "step": 4927
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485853548966014e-05,
+      "loss": 1.2822,
+      "step": 4928
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483358261789656e-05,
+      "loss": 1.2119,
+      "step": 4929
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3480862727747868e-05,
+      "loss": 1.2578,
+      "step": 4930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478366947017521e-05,
+      "loss": 1.2158,
+      "step": 4931
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475870919775482e-05,
+      "loss": 1.2676,
+      "step": 4932
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473374646198645e-05,
+      "loss": 1.2939,
+      "step": 4933
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470878126463926e-05,
+      "loss": 1.2285,
+      "step": 4934
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346838136074825e-05,
+      "loss": 1.2598,
+      "step": 4935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465884349228563e-05,
+      "loss": 1.2803,
+      "step": 4936
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463387092081826e-05,
+      "loss": 1.2549,
+      "step": 4937
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460889589485022e-05,
+      "loss": 1.2666,
+      "step": 4938
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345839184161515e-05,
+      "loss": 1.3174,
+      "step": 4939
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3455893848649222e-05,
+      "loss": 1.2656,
+      "step": 4940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453395610764273e-05,
+      "loss": 1.2939,
+      "step": 4941
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345089712813735e-05,
+      "loss": 1.2734,
+      "step": 4942
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3448398400945527e-05,
+      "loss": 1.2314,
+      "step": 4943
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3445899429365877e-05,
+      "loss": 1.248,
+      "step": 4944
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443400213575515e-05,
+      "loss": 1.251,
+      "step": 4945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3440900753751552e-05,
+      "loss": 1.3066,
+      "step": 4946
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438401050071129e-05,
+      "loss": 1.2402,
+      "step": 4947
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435901102711392e-05,
+      "loss": 1.2422,
+      "step": 4948
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3433400911849525e-05,
+      "loss": 1.2998,
+      "step": 4949
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430900477662708e-05,
+      "loss": 1.2588,
+      "step": 4950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428399800328148e-05,
+      "loss": 1.3428,
+      "step": 4951
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3425898880023065e-05,
+      "loss": 1.249,
+      "step": 4952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423397716924705e-05,
+      "loss": 1.2295,
+      "step": 4953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3420896311210321e-05,
+      "loss": 1.2461,
+      "step": 4954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418394663057188e-05,
+      "loss": 1.2842,
+      "step": 4955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34158927726426e-05,
+      "loss": 1.2207,
+      "step": 4956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413390640143866e-05,
+      "loss": 1.2344,
+      "step": 4957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341088826573831e-05,
+      "loss": 1.3037,
+      "step": 4958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340838564960327e-05,
+      "loss": 1.2646,
+      "step": 4959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3405882791916116e-05,
+      "loss": 1.2607,
+      "step": 4960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403379692854223e-05,
+      "loss": 1.3145,
+      "step": 4961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400876352594981e-05,
+      "loss": 1.292,
+      "step": 4962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3398372771315807e-05,
+      "loss": 1.2646,
+      "step": 4963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395868949194122e-05,
+      "loss": 1.2832,
+      "step": 4964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393364886407383e-05,
+      "loss": 1.2676,
+      "step": 4965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339086058313304e-05,
+      "loss": 1.252,
+      "step": 4966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388356039548585e-05,
+      "loss": 1.2246,
+      "step": 4967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3385851255831505e-05,
+      "loss": 1.2422,
+      "step": 4968
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338334623215932e-05,
+      "loss": 1.1934,
+      "step": 4969
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380840968709559e-05,
+      "loss": 1.2803,
+      "step": 4970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378335465659772e-05,
+      "loss": 1.2354,
+      "step": 4971
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375829723187519e-05,
+      "loss": 1.2441,
+      "step": 4972
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3373323741470386e-05,
+      "loss": 1.21,
+      "step": 4973
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3370817520685968e-05,
+      "loss": 1.2871,
+      "step": 4974
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3368311061011887e-05,
+      "loss": 1.2783,
+      "step": 4975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336580436262577e-05,
+      "loss": 1.3389,
+      "step": 4976
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336329742570527e-05,
+      "loss": 1.2607,
+      "step": 4977
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360790250428049e-05,
+      "loss": 1.292,
+      "step": 4978
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3358282836971795e-05,
+      "loss": 1.2168,
+      "step": 4979
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355775185514203e-05,
+      "loss": 1.2568,
+      "step": 4980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3353267296233e-05,
+      "loss": 1.25,
+      "step": 4981
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335075916930591e-05,
+      "loss": 1.2812,
+      "step": 4982
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3348250804910689e-05,
+      "loss": 1.2627,
+      "step": 4983
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33457422032251e-05,
+      "loss": 1.3066,
+      "step": 4984
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343233364426937e-05,
+      "loss": 1.2988,
+      "step": 4985
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3340724288693989e-05,
+      "loss": 1.2627,
+      "step": 4986
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3338214976204083e-05,
+      "loss": 1.252,
+      "step": 4987
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3335705427135048e-05,
+      "loss": 1.2949,
+      "step": 4988
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3333195641664745e-05,
+      "loss": 1.2578,
+      "step": 4989
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330685619971029e-05,
+      "loss": 1.2227,
+      "step": 4990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328175362231794e-05,
+      "loss": 1.2588,
+      "step": 4991
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3325664868624939e-05,
+      "loss": 1.2881,
+      "step": 4992
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323154139328387e-05,
+      "loss": 1.2031,
+      "step": 4993
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320643174520066e-05,
+      "loss": 1.2803,
+      "step": 4994
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3318131974377932e-05,
+      "loss": 1.2598,
+      "step": 4995
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3315620539079955e-05,
+      "loss": 1.2578,
+      "step": 4996
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3313108868804118e-05,
+      "loss": 1.2695,
+      "step": 4997
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3310596963728422e-05,
+      "loss": 1.2266,
+      "step": 4998
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3308084824030888e-05,
+      "loss": 1.2598,
+      "step": 4999
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330557244988955e-05,
+      "loss": 1.2393,
+      "step": 5000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330305984148246e-05,
+      "loss": 1.2715,
+      "step": 5001
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300546998987685e-05,
+      "loss": 1.2842,
+      "step": 5002
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298033922583315e-05,
+      "loss": 1.2539,
+      "step": 5003
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295520612447445e-05,
+      "loss": 1.2695,
+      "step": 5004
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3293007068758196e-05,
+      "loss": 1.2637,
+      "step": 5005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3290493291693707e-05,
+      "loss": 1.2812,
+      "step": 5006
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287979281432123e-05,
+      "loss": 1.2852,
+      "step": 5007
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3285465038151615e-05,
+      "loss": 1.2773,
+      "step": 5008
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3282950562030362e-05,
+      "loss": 1.2529,
+      "step": 5009
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280435853246574e-05,
+      "loss": 1.2539,
+      "step": 5010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327792091197846e-05,
+      "loss": 1.2705,
+      "step": 5011
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3275405738404259e-05,
+      "loss": 1.2578,
+      "step": 5012
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272890332702218e-05,
+      "loss": 1.2988,
+      "step": 5013
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3270374695050608e-05,
+      "loss": 1.2451,
+      "step": 5014
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3267858825627707e-05,
+      "loss": 1.2656,
+      "step": 5015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265342724611814e-05,
+      "loss": 1.2852,
+      "step": 5016
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3262826392181251e-05,
+      "loss": 1.2227,
+      "step": 5017
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3260309828514348e-05,
+      "loss": 1.2217,
+      "step": 5018
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325779303378945e-05,
+      "loss": 1.335,
+      "step": 5019
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255276008184928e-05,
+      "loss": 1.2949,
+      "step": 5020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3252758751879156e-05,
+      "loss": 1.2383,
+      "step": 5021
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325024126505054e-05,
+      "loss": 1.2354,
+      "step": 5022
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247723547877488e-05,
+      "loss": 1.2324,
+      "step": 5023
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3245205600538435e-05,
+      "loss": 1.2021,
+      "step": 5024
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242687423211824e-05,
+      "loss": 1.2568,
+      "step": 5025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324016901607612e-05,
+      "loss": 1.3037,
+      "step": 5026
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32376503793098e-05,
+      "loss": 1.3037,
+      "step": 5027
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235131513091366e-05,
+      "loss": 1.2793,
+      "step": 5028
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232612417599324e-05,
+      "loss": 1.2168,
+      "step": 5029
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230093093012203e-05,
+      "loss": 1.249,
+      "step": 5030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227573539508547e-05,
+      "loss": 1.2568,
+      "step": 5031
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3225053757266921e-05,
+      "loss": 1.2139,
+      "step": 5032
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222533746465897e-05,
+      "loss": 1.3008,
+      "step": 5033
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322001350728407e-05,
+      "loss": 1.2695,
+      "step": 5034
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3217493039900049e-05,
+      "loss": 1.2344,
+      "step": 5035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321497234449246e-05,
+      "loss": 1.2373,
+      "step": 5036
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3212451421239942e-05,
+      "loss": 1.2471,
+      "step": 5037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209930270321157e-05,
+      "loss": 1.2773,
+      "step": 5038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3207408891914775e-05,
+      "loss": 1.2744,
+      "step": 5039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3204887286199491e-05,
+      "loss": 1.2441,
+      "step": 5040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3202365453354002e-05,
+      "loss": 1.2461,
+      "step": 5041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199843393557036e-05,
+      "loss": 1.2598,
+      "step": 5042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197321106987335e-05,
+      "loss": 1.3291,
+      "step": 5043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194798593823648e-05,
+      "loss": 1.2871,
+      "step": 5044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3192275854244742e-05,
+      "loss": 1.2295,
+      "step": 5045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3189752888429415e-05,
+      "loss": 1.3174,
+      "step": 5046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187229696556459e-05,
+      "loss": 1.2432,
+      "step": 5047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3184706278804697e-05,
+      "loss": 1.2285,
+      "step": 5048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.318218263535296e-05,
+      "loss": 1.2881,
+      "step": 5049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179658766380103e-05,
+      "loss": 1.2598,
+      "step": 5050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177134672064991e-05,
+      "loss": 1.293,
+      "step": 5051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3174610352586505e-05,
+      "loss": 1.2715,
+      "step": 5052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172085808123543e-05,
+      "loss": 1.2588,
+      "step": 5053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169561038855023e-05,
+      "loss": 1.2734,
+      "step": 5054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167036044959871e-05,
+      "loss": 1.2363,
+      "step": 5055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164510826617036e-05,
+      "loss": 1.2646,
+      "step": 5056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161985384005478e-05,
+      "loss": 1.2285,
+      "step": 5057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159459717304179e-05,
+      "loss": 1.249,
+      "step": 5058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3156933826692128e-05,
+      "loss": 1.2402,
+      "step": 5059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154407712348335e-05,
+      "loss": 1.2598,
+      "step": 5060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315188137445183e-05,
+      "loss": 1.2539,
+      "step": 5061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149354813181652e-05,
+      "loss": 1.2422,
+      "step": 5062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146828028716856e-05,
+      "loss": 1.2461,
+      "step": 5063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144301021236519e-05,
+      "loss": 1.2119,
+      "step": 5064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141773790919725e-05,
+      "loss": 1.2197,
+      "step": 5065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139246337945585e-05,
+      "loss": 1.249,
+      "step": 5066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136718662493215e-05,
+      "loss": 1.2373,
+      "step": 5067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134190764741751e-05,
+      "loss": 1.3115,
+      "step": 5068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313166264487035e-05,
+      "loss": 1.2109,
+      "step": 5069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3129134303058172e-05,
+      "loss": 1.248,
+      "step": 5070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126605739484406e-05,
+      "loss": 1.2451,
+      "step": 5071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3124076954328254e-05,
+      "loss": 1.29,
+      "step": 5072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312154794776892e-05,
+      "loss": 1.2725,
+      "step": 5073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119018719985646e-05,
+      "loss": 1.2744,
+      "step": 5074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116489271157668e-05,
+      "loss": 1.2012,
+      "step": 5075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113959601464262e-05,
+      "loss": 1.2461,
+      "step": 5076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111429711084688e-05,
+      "loss": 1.2715,
+      "step": 5077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108899600198253e-05,
+      "loss": 1.2969,
+      "step": 5078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106369268984258e-05,
+      "loss": 1.2959,
+      "step": 5079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3103838717622034e-05,
+      "loss": 1.2354,
+      "step": 5080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101307946290916e-05,
+      "loss": 1.2568,
+      "step": 5081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309877695517026e-05,
+      "loss": 1.2744,
+      "step": 5082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096245744439439e-05,
+      "loss": 1.2578,
+      "step": 5083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093714314277841e-05,
+      "loss": 1.2549,
+      "step": 5084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091182664864861e-05,
+      "loss": 1.2891,
+      "step": 5085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088650796379927e-05,
+      "loss": 1.2715,
+      "step": 5086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3086118709002467e-05,
+      "loss": 1.251,
+      "step": 5087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308358640291193e-05,
+      "loss": 1.1963,
+      "step": 5088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308105387828778e-05,
+      "loss": 1.2744,
+      "step": 5089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078521135309501e-05,
+      "loss": 1.2646,
+      "step": 5090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3075988174156581e-05,
+      "loss": 1.2695,
+      "step": 5091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073454995008536e-05,
+      "loss": 1.2617,
+      "step": 5092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070921598044893e-05,
+      "loss": 1.3174,
+      "step": 5093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3068387983445192e-05,
+      "loss": 1.2764,
+      "step": 5094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306585415138899e-05,
+      "loss": 1.3018,
+      "step": 5095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306332010205586e-05,
+      "loss": 1.248,
+      "step": 5096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3060785835625386e-05,
+      "loss": 1.2578,
+      "step": 5097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305825135227718e-05,
+      "loss": 1.2812,
+      "step": 5098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055716652190851e-05,
+      "loss": 1.2852,
+      "step": 5099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305318173554604e-05,
+      "loss": 1.2832,
+      "step": 5100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3050646602522395e-05,
+      "loss": 1.2402,
+      "step": 5101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048111253299582e-05,
+      "loss": 1.2861,
+      "step": 5102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045575688057275e-05,
+      "loss": 1.248,
+      "step": 5103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043039906975174e-05,
+      "loss": 1.2881,
+      "step": 5104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304050391023299e-05,
+      "loss": 1.2656,
+      "step": 5105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037967698010447e-05,
+      "loss": 1.2812,
+      "step": 5106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035431270487289e-05,
+      "loss": 1.291,
+      "step": 5107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303289462784327e-05,
+      "loss": 1.2373,
+      "step": 5108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030357770258163e-05,
+      "loss": 1.2783,
+      "step": 5109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3027820697911752e-05,
+      "loss": 1.291,
+      "step": 5110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3025283410983844e-05,
+      "loss": 1.2715,
+      "step": 5111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3022745909654255e-05,
+      "loss": 1.208,
+      "step": 5112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020208194102814e-05,
+      "loss": 1.2783,
+      "step": 5113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017670264509372e-05,
+      "loss": 1.2432,
+      "step": 5114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015132121053793e-05,
+      "loss": 1.2988,
+      "step": 5115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012593763915953e-05,
+      "loss": 1.2656,
+      "step": 5116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3010055193275744e-05,
+      "loss": 1.2705,
+      "step": 5117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3007516409313077e-05,
+      "loss": 1.2939,
+      "step": 5118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004977412207876e-05,
+      "loss": 1.2529,
+      "step": 5119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3002438202140076e-05,
+      "loss": 1.2725,
+      "step": 5120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299989877928963e-05,
+      "loss": 1.2324,
+      "step": 5121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997359143836516e-05,
+      "loss": 1.2637,
+      "step": 5122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994819295960708e-05,
+      "loss": 1.2539,
+      "step": 5123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299227923584221e-05,
+      "loss": 1.25,
+      "step": 5124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989738963661032e-05,
+      "loss": 1.2939,
+      "step": 5125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987198479597207e-05,
+      "loss": 1.3154,
+      "step": 5126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984657783830776e-05,
+      "loss": 1.2402,
+      "step": 5127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2982116876541799e-05,
+      "loss": 1.2559,
+      "step": 5128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297957575791035e-05,
+      "loss": 1.2715,
+      "step": 5129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977034428116522e-05,
+      "loss": 1.2344,
+      "step": 5130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297449288734041e-05,
+      "loss": 1.2402,
+      "step": 5131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971951135762139e-05,
+      "loss": 1.2383,
+      "step": 5132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969409173561843e-05,
+      "loss": 1.2617,
+      "step": 5133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296686700091967e-05,
+      "loss": 1.2471,
+      "step": 5134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296432461801578e-05,
+      "loss": 1.2285,
+      "step": 5135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2961782025030356e-05,
+      "loss": 1.2822,
+      "step": 5136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959239222143592e-05,
+      "loss": 1.2725,
+      "step": 5137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956696209535693e-05,
+      "loss": 1.251,
+      "step": 5138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954152987386882e-05,
+      "loss": 1.2705,
+      "step": 5139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2951609555877397e-05,
+      "loss": 1.2266,
+      "step": 5140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29490659151875e-05,
+      "loss": 1.2275,
+      "step": 5141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2946522065497445e-05,
+      "loss": 1.2158,
+      "step": 5142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294397800698752e-05,
+      "loss": 1.2764,
+      "step": 5143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941433739838027e-05,
+      "loss": 1.2695,
+      "step": 5144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938889264229272e-05,
+      "loss": 1.2471,
+      "step": 5145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936344580341582e-05,
+      "loss": 1.2275,
+      "step": 5146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2933799688355304e-05,
+      "loss": 1.1924,
+      "step": 5147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931254588450789e-05,
+      "loss": 1.25,
+      "step": 5148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2928709280808412e-05,
+      "loss": 1.2422,
+      "step": 5149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292616376560855e-05,
+      "loss": 1.2617,
+      "step": 5150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923618043031615e-05,
+      "loss": 1.2617,
+      "step": 5151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292107211325802e-05,
+      "loss": 1.252,
+      "step": 5152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2918525976468188e-05,
+      "loss": 1.2393,
+      "step": 5153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291597963284257e-05,
+      "loss": 1.1738,
+      "step": 5154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913433082561624e-05,
+      "loss": 1.2852,
+      "step": 5155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2910886325805823e-05,
+      "loss": 1.2158,
+      "step": 5156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908339362755652e-05,
+      "loss": 1.293,
+      "step": 5157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905792193591622e-05,
+      "loss": 1.2715,
+      "step": 5158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903244818494246e-05,
+      "loss": 1.2109,
+      "step": 5159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900697237644057e-05,
+      "loss": 1.2754,
+      "step": 5160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.28981494512216e-05,
+      "loss": 1.3203,
+      "step": 5161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895601459407438e-05,
+      "loss": 1.3164,
+      "step": 5162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893053262382152e-05,
+      "loss": 1.2197,
+      "step": 5163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289050486032632e-05,
+      "loss": 1.2275,
+      "step": 5164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887956253420564e-05,
+      "loss": 1.2373,
+      "step": 5165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288540744184549e-05,
+      "loss": 1.2256,
+      "step": 5166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2882858425781736e-05,
+      "loss": 1.29,
+      "step": 5167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288030920540995e-05,
+      "loss": 1.2559,
+      "step": 5168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2877759780910803e-05,
+      "loss": 1.3115,
+      "step": 5169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875210152464964e-05,
+      "loss": 1.2598,
+      "step": 5170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872660320253124e-05,
+      "loss": 1.291,
+      "step": 5171
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2870110284455995e-05,
+      "loss": 1.2373,
+      "step": 5172
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867560045254295e-05,
+      "loss": 1.249,
+      "step": 5173
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865009602828762e-05,
+      "loss": 1.2979,
+      "step": 5174
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286245895736014e-05,
+      "loss": 1.2168,
+      "step": 5175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.28599081090292e-05,
+      "loss": 1.29,
+      "step": 5176
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2857357058016718e-05,
+      "loss": 1.292,
+      "step": 5177
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285480580450348e-05,
+      "loss": 1.2461,
+      "step": 5178
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852254348670306e-05,
+      "loss": 1.2461,
+      "step": 5179
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2849702690698008e-05,
+      "loss": 1.2607,
+      "step": 5180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2847150830767426e-05,
+      "loss": 1.2656,
+      "step": 5181
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2844598769059406e-05,
+      "loss": 1.2891,
+      "step": 5182
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2842046505754819e-05,
+      "loss": 1.251,
+      "step": 5183
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283949404103454e-05,
+      "loss": 1.2598,
+      "step": 5184
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836941375079462e-05,
+      "loss": 1.2939,
+      "step": 5185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2834388508070488e-05,
+      "loss": 1.2363,
+      "step": 5186
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2831835440188549e-05,
+      "loss": 1.3008,
+      "step": 5187
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829282171614574e-05,
+      "loss": 1.2842,
+      "step": 5188
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826728702529516e-05,
+      "loss": 1.3047,
+      "step": 5189
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2824175033114336e-05,
+      "loss": 1.3008,
+      "step": 5190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821621163550019e-05,
+      "loss": 1.2266,
+      "step": 5191
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281906709401755e-05,
+      "loss": 1.2832,
+      "step": 5192
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281651282469794e-05,
+      "loss": 1.2578,
+      "step": 5193
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813958355772206e-05,
+      "loss": 1.2891,
+      "step": 5194
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811403687421393e-05,
+      "loss": 1.2568,
+      "step": 5195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808848819826537e-05,
+      "loss": 1.2783,
+      "step": 5196
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2806293753168709e-05,
+      "loss": 1.2432,
+      "step": 5197
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2803738487628985e-05,
+      "loss": 1.3242,
+      "step": 5198
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801183023388459e-05,
+      "loss": 1.2646,
+      "step": 5199
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798627360628228e-05,
+      "loss": 1.2725,
+      "step": 5200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796071499529424e-05,
+      "loss": 1.2764,
+      "step": 5201
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2793515440273172e-05,
+      "loss": 1.2324,
+      "step": 5202
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2790959183040623e-05,
+      "loss": 1.2754,
+      "step": 5203
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2788402728012936e-05,
+      "loss": 1.2471,
+      "step": 5204
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2785846075371292e-05,
+      "loss": 1.2275,
+      "step": 5205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783289225296876e-05,
+      "loss": 1.2676,
+      "step": 5206
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2780732177970894e-05,
+      "loss": 1.209,
+      "step": 5207
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778174933574562e-05,
+      "loss": 1.2656,
+      "step": 5208
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775617492289118e-05,
+      "loss": 1.2998,
+      "step": 5209
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27730598542958e-05,
+      "loss": 1.25,
+      "step": 5210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770502019775873e-05,
+      "loss": 1.2891,
+      "step": 5211
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767943988910605e-05,
+      "loss": 1.3271,
+      "step": 5212
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765385761881288e-05,
+      "loss": 1.2852,
+      "step": 5213
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762827338869222e-05,
+      "loss": 1.252,
+      "step": 5214
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760268720055722e-05,
+      "loss": 1.2695,
+      "step": 5215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2757709905622119e-05,
+      "loss": 1.2969,
+      "step": 5216
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755150895749755e-05,
+      "loss": 1.2432,
+      "step": 5217
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752591690619985e-05,
+      "loss": 1.2734,
+      "step": 5218
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750032290414183e-05,
+      "loss": 1.2607,
+      "step": 5219
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747472695313732e-05,
+      "loss": 1.2588,
+      "step": 5220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274491290550003e-05,
+      "loss": 1.292,
+      "step": 5221
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742352921154489e-05,
+      "loss": 1.248,
+      "step": 5222
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739792742458537e-05,
+      "loss": 1.2998,
+      "step": 5223
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273723236959361e-05,
+      "loss": 1.2705,
+      "step": 5224
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2734671802741165e-05,
+      "loss": 1.2549,
+      "step": 5225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732111042082667e-05,
+      "loss": 1.2744,
+      "step": 5226
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27295500877996e-05,
+      "loss": 1.3008,
+      "step": 5227
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726988940073458e-05,
+      "loss": 1.207,
+      "step": 5228
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2724427599085743e-05,
+      "loss": 1.2607,
+      "step": 5229
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721866065017984e-05,
+      "loss": 1.2158,
+      "step": 5230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2719304338051716e-05,
+      "loss": 1.2891,
+      "step": 5231
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2716742418368488e-05,
+      "loss": 1.2363,
+      "step": 5232
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714180306149862e-05,
+      "loss": 1.2773,
+      "step": 5233
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711618001577412e-05,
+      "loss": 1.3311,
+      "step": 5234
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709055504832736e-05,
+      "loss": 1.2754,
+      "step": 5235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2706492816097433e-05,
+      "loss": 1.2393,
+      "step": 5236
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2703929935553121e-05,
+      "loss": 1.249,
+      "step": 5237
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2701366863381432e-05,
+      "loss": 1.2402,
+      "step": 5238
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698803599764009e-05,
+      "loss": 1.2607,
+      "step": 5239
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269624014488251e-05,
+      "loss": 1.2471,
+      "step": 5240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693676498918613e-05,
+      "loss": 1.3203,
+      "step": 5241
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2691112662053997e-05,
+      "loss": 1.2539,
+      "step": 5242
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688548634470363e-05,
+      "loss": 1.2598,
+      "step": 5243
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685984416349424e-05,
+      "loss": 1.251,
+      "step": 5244
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2683420007872906e-05,
+      "loss": 1.2627,
+      "step": 5245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2680855409222548e-05,
+      "loss": 1.2266,
+      "step": 5246
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678290620580101e-05,
+      "loss": 1.2676,
+      "step": 5247
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675725642127334e-05,
+      "loss": 1.2715,
+      "step": 5248
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673160474046029e-05,
+      "loss": 1.2383,
+      "step": 5249
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2670595116517974e-05,
+      "loss": 1.1904,
+      "step": 5250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266802956972498e-05,
+      "loss": 1.2568,
+      "step": 5251
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2665463833848864e-05,
+      "loss": 1.2002,
+      "step": 5252
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662897909071464e-05,
+      "loss": 1.1973,
+      "step": 5253
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660331795574621e-05,
+      "loss": 1.2314,
+      "step": 5254
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.26577654935402e-05,
+      "loss": 1.2695,
+      "step": 5255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2655199003150074e-05,
+      "loss": 1.2197,
+      "step": 5256
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265263232458613e-05,
+      "loss": 1.2568,
+      "step": 5257
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2650065458030263e-05,
+      "loss": 1.2734,
+      "step": 5258
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2647498403664395e-05,
+      "loss": 1.2666,
+      "step": 5259
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2644931161670449e-05,
+      "loss": 1.2539,
+      "step": 5260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2642363732230365e-05,
+      "loss": 1.2803,
+      "step": 5261
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639796115526096e-05,
+      "loss": 1.2344,
+      "step": 5262
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637228311739614e-05,
+      "loss": 1.2441,
+      "step": 5263
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634660321052894e-05,
+      "loss": 1.25,
+      "step": 5264
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263209214364793e-05,
+      "loss": 1.3271,
+      "step": 5265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2629523779706728e-05,
+      "loss": 1.2754,
+      "step": 5266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2626955229411311e-05,
+      "loss": 1.2627,
+      "step": 5267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2624386492943708e-05,
+      "loss": 1.2773,
+      "step": 5268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621817570485968e-05,
+      "loss": 1.2773,
+      "step": 5269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619248462220149e-05,
+      "loss": 1.2529,
+      "step": 5270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616679168328327e-05,
+      "loss": 1.2129,
+      "step": 5271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2614109688992579e-05,
+      "loss": 1.2559,
+      "step": 5272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2611540024395015e-05,
+      "loss": 1.2559,
+      "step": 5273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608970174717742e-05,
+      "loss": 1.2217,
+      "step": 5274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606400140142885e-05,
+      "loss": 1.2842,
+      "step": 5275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603829920852579e-05,
+      "loss": 1.2666,
+      "step": 5276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260125951702898e-05,
+      "loss": 1.2334,
+      "step": 5277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2598688928854255e-05,
+      "loss": 1.2334,
+      "step": 5278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596118156510573e-05,
+      "loss": 1.2383,
+      "step": 5279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593547200180129e-05,
+      "loss": 1.3008,
+      "step": 5280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590976060045125e-05,
+      "loss": 1.3271,
+      "step": 5281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588404736287784e-05,
+      "loss": 1.2402,
+      "step": 5282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585833229090328e-05,
+      "loss": 1.2549,
+      "step": 5283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583261538635e-05,
+      "loss": 1.2646,
+      "step": 5284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258068966510406e-05,
+      "loss": 1.2754,
+      "step": 5285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2578117608679776e-05,
+      "loss": 1.2754,
+      "step": 5286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575545369544422e-05,
+      "loss": 1.251,
+      "step": 5287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572972947880304e-05,
+      "loss": 1.2627,
+      "step": 5288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2570400343869722e-05,
+      "loss": 1.208,
+      "step": 5289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2567827557694996e-05,
+      "loss": 1.2959,
+      "step": 5290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2565254589538462e-05,
+      "loss": 1.2646,
+      "step": 5291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2562681439582467e-05,
+      "loss": 1.2314,
+      "step": 5292
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560108108009368e-05,
+      "loss": 1.2168,
+      "step": 5293
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557534595001536e-05,
+      "loss": 1.2939,
+      "step": 5294
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2554960900741358e-05,
+      "loss": 1.2549,
+      "step": 5295
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255238702541123e-05,
+      "loss": 1.2705,
+      "step": 5296
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549812969193563e-05,
+      "loss": 1.2178,
+      "step": 5297
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2547238732270779e-05,
+      "loss": 1.2031,
+      "step": 5298
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544664314825314e-05,
+      "loss": 1.252,
+      "step": 5299
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542089717039622e-05,
+      "loss": 1.2539,
+      "step": 5300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539514939096158e-05,
+      "loss": 1.2568,
+      "step": 5301
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536939981177398e-05,
+      "loss": 1.2783,
+      "step": 5302
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534364843465832e-05,
+      "loss": 1.2861,
+      "step": 5303
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253178952614396e-05,
+      "loss": 1.29,
+      "step": 5304
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2529214029394288e-05,
+      "loss": 1.2344,
+      "step": 5305
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2526638353399349e-05,
+      "loss": 1.2627,
+      "step": 5306
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2524062498341679e-05,
+      "loss": 1.2822,
+      "step": 5307
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2521486464403825e-05,
+      "loss": 1.2734,
+      "step": 5308
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518910251768352e-05,
+      "loss": 1.3076,
+      "step": 5309
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2516333860617842e-05,
+      "loss": 1.3027,
+      "step": 5310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2513757291134876e-05,
+      "loss": 1.2207,
+      "step": 5311
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511180543502057e-05,
+      "loss": 1.2783,
+      "step": 5312
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2508603617902003e-05,
+      "loss": 1.292,
+      "step": 5313
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2506026514517338e-05,
+      "loss": 1.2646,
+      "step": 5314
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2503449233530697e-05,
+      "loss": 1.2461,
+      "step": 5315
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500871775124739e-05,
+      "loss": 1.2617,
+      "step": 5316
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498294139482124e-05,
+      "loss": 1.2412,
+      "step": 5317
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2495716326785531e-05,
+      "loss": 1.25,
+      "step": 5318
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493138337217645e-05,
+      "loss": 1.2285,
+      "step": 5319
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490560170961175e-05,
+      "loss": 1.2578,
+      "step": 5320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487981828198832e-05,
+      "loss": 1.2646,
+      "step": 5321
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485403309113341e-05,
+      "loss": 1.2012,
+      "step": 5322
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482824613887443e-05,
+      "loss": 1.251,
+      "step": 5323
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2480245742703893e-05,
+      "loss": 1.3389,
+      "step": 5324
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2477666695745452e-05,
+      "loss": 1.2334,
+      "step": 5325
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475087473194898e-05,
+      "loss": 1.2207,
+      "step": 5326
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247250807523502e-05,
+      "loss": 1.2939,
+      "step": 5327
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246992850204862e-05,
+      "loss": 1.2686,
+      "step": 5328
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2467348753818511e-05,
+      "loss": 1.2949,
+      "step": 5329
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464768830727523e-05,
+      "loss": 1.2588,
+      "step": 5330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2462188732958492e-05,
+      "loss": 1.2725,
+      "step": 5331
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459608460694273e-05,
+      "loss": 1.2822,
+      "step": 5332
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2457028014117723e-05,
+      "loss": 1.2705,
+      "step": 5333
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2454447393411722e-05,
+      "loss": 1.25,
+      "step": 5334
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2451866598759161e-05,
+      "loss": 1.2588,
+      "step": 5335
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449285630342938e-05,
+      "loss": 1.2686,
+      "step": 5336
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2446704488345965e-05,
+      "loss": 1.2695,
+      "step": 5337
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244412317295117e-05,
+      "loss": 1.249,
+      "step": 5338
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2441541684341491e-05,
+      "loss": 1.2695,
+      "step": 5339
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438960022699878e-05,
+      "loss": 1.2627,
+      "step": 5340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436378188209286e-05,
+      "loss": 1.2852,
+      "step": 5341
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24337961810527e-05,
+      "loss": 1.2539,
+      "step": 5342
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24312140014131e-05,
+      "loss": 1.2764,
+      "step": 5343
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428631649473488e-05,
+      "loss": 1.252,
+      "step": 5344
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426049125416872e-05,
+      "loss": 1.2559,
+      "step": 5345
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423466429426281e-05,
+      "loss": 1.3086,
+      "step": 5346
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2420883561684747e-05,
+      "loss": 1.3154,
+      "step": 5347
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2418300522375315e-05,
+      "loss": 1.2012,
+      "step": 5348
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415717311681049e-05,
+      "loss": 1.2344,
+      "step": 5349
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241313392978502e-05,
+      "loss": 1.2764,
+      "step": 5350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241055037687031e-05,
+      "loss": 1.3223,
+      "step": 5351
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2407966653120017e-05,
+      "loss": 1.251,
+      "step": 5352
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240538275871725e-05,
+      "loss": 1.2539,
+      "step": 5353
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240279869384513e-05,
+      "loss": 1.3301,
+      "step": 5354
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400214458686786e-05,
+      "loss": 1.2695,
+      "step": 5355
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397630053425367e-05,
+      "loss": 1.2627,
+      "step": 5356
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2395045478244028e-05,
+      "loss": 1.2764,
+      "step": 5357
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392460733325938e-05,
+      "loss": 1.248,
+      "step": 5358
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389875818854273e-05,
+      "loss": 1.2539,
+      "step": 5359
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387290735012234e-05,
+      "loss": 1.2324,
+      "step": 5360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238470548198302e-05,
+      "loss": 1.2695,
+      "step": 5361
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2382120059949849e-05,
+      "loss": 1.2295,
+      "step": 5362
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379534469095949e-05,
+      "loss": 1.2578,
+      "step": 5363
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376948709604565e-05,
+      "loss": 1.25,
+      "step": 5364
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2374362781658943e-05,
+      "loss": 1.252,
+      "step": 5365
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2371776685442354e-05,
+      "loss": 1.2373,
+      "step": 5366
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236919042113807e-05,
+      "loss": 1.2891,
+      "step": 5367
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366603988929385e-05,
+      "loss": 1.2666,
+      "step": 5368
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236401738899959e-05,
+      "loss": 1.2617,
+      "step": 5369
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361430621532003e-05,
+      "loss": 1.2549,
+      "step": 5370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235884368670995e-05,
+      "loss": 1.2803,
+      "step": 5371
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356256584716764e-05,
+      "loss": 1.2676,
+      "step": 5372
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235366931573579e-05,
+      "loss": 1.2607,
+      "step": 5373
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351081879950395e-05,
+      "loss": 1.2402,
+      "step": 5374
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348494277543942e-05,
+      "loss": 1.2812,
+      "step": 5375
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345906508699826e-05,
+      "loss": 1.2686,
+      "step": 5376
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343318573601427e-05,
+      "loss": 1.2061,
+      "step": 5377
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340730472432163e-05,
+      "loss": 1.2725,
+      "step": 5378
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233814220537545e-05,
+      "loss": 1.2275,
+      "step": 5379
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335553772614715e-05,
+      "loss": 1.2178,
+      "step": 5380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332965174333402e-05,
+      "loss": 1.168,
+      "step": 5381
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330376410714969e-05,
+      "loss": 1.3096,
+      "step": 5382
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2327787481942877e-05,
+      "loss": 1.2764,
+      "step": 5383
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325198388200602e-05,
+      "loss": 1.29,
+      "step": 5384
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322609129671634e-05,
+      "loss": 1.2734,
+      "step": 5385
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232001970653948e-05,
+      "loss": 1.2764,
+      "step": 5386
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317430118987643e-05,
+      "loss": 1.2588,
+      "step": 5387
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2314840367199652e-05,
+      "loss": 1.293,
+      "step": 5388
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231225045135904e-05,
+      "loss": 1.252,
+      "step": 5389
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2309660371649359e-05,
+      "loss": 1.2744,
+      "step": 5390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307070128254159e-05,
+      "loss": 1.2666,
+      "step": 5391
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2304479721357024e-05,
+      "loss": 1.3018,
+      "step": 5392
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301889151141521e-05,
+      "loss": 1.2705,
+      "step": 5393
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299298417791255e-05,
+      "loss": 1.2451,
+      "step": 5394
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2296707521489823e-05,
+      "loss": 1.2217,
+      "step": 5395
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294116462420848e-05,
+      "loss": 1.2705,
+      "step": 5396
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2291525240767954e-05,
+      "loss": 1.2539,
+      "step": 5397
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288933856714785e-05,
+      "loss": 1.2617,
+      "step": 5398
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286342310444988e-05,
+      "loss": 1.252,
+      "step": 5399
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283750602142228e-05,
+      "loss": 1.2227,
+      "step": 5400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2281158731990178e-05,
+      "loss": 1.2656,
+      "step": 5401
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278566700172523e-05,
+      "loss": 1.2861,
+      "step": 5402
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227597450687296e-05,
+      "loss": 1.2598,
+      "step": 5403
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273382152275202e-05,
+      "loss": 1.2695,
+      "step": 5404
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270789636562967e-05,
+      "loss": 1.2432,
+      "step": 5405
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268196959919982e-05,
+      "loss": 1.3027,
+      "step": 5406
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226560412253e-05,
+      "loss": 1.2998,
+      "step": 5407
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263011124576762e-05,
+      "loss": 1.2568,
+      "step": 5408
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2260417966244043e-05,
+      "loss": 1.2549,
+      "step": 5409
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257824647715617e-05,
+      "loss": 1.2881,
+      "step": 5410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2255231169175274e-05,
+      "loss": 1.2363,
+      "step": 5411
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.225263753080681e-05,
+      "loss": 1.2422,
+      "step": 5412
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.225004373279404e-05,
+      "loss": 1.2441,
+      "step": 5413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247449775320783e-05,
+      "loss": 1.3135,
+      "step": 5414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244855658570878e-05,
+      "loss": 1.2227,
+      "step": 5415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2242261382728162e-05,
+      "loss": 1.2344,
+      "step": 5416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2239666947976499e-05,
+      "loss": 1.2891,
+      "step": 5417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223707235449975e-05,
+      "loss": 1.2383,
+      "step": 5418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.22344776024818e-05,
+      "loss": 1.209,
+      "step": 5419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231882692106533e-05,
+      "loss": 1.2754,
+      "step": 5420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229287623557852e-05,
+      "loss": 1.2539,
+      "step": 5421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2226692397019673e-05,
+      "loss": 1.2256,
+      "step": 5422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2224097012675915e-05,
+      "loss": 1.2529,
+      "step": 5423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2221501470710513e-05,
+      "loss": 1.3047,
+      "step": 5424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2218905771307417e-05,
+      "loss": 1.2158,
+      "step": 5425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216309914650581e-05,
+      "loss": 1.2939,
+      "step": 5426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2213713900923972e-05,
+      "loss": 1.2275,
+      "step": 5427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211117730311572e-05,
+      "loss": 1.2422,
+      "step": 5428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208521402997372e-05,
+      "loss": 1.2764,
+      "step": 5429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220592491916537e-05,
+      "loss": 1.2061,
+      "step": 5430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2203328278999582e-05,
+      "loss": 1.2871,
+      "step": 5431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200731482684032e-05,
+      "loss": 1.2539,
+      "step": 5432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2198134530402752e-05,
+      "loss": 1.2666,
+      "step": 5433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2195537422339789e-05,
+      "loss": 1.2715,
+      "step": 5434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21929401586792e-05,
+      "loss": 1.2471,
+      "step": 5435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190342739605052e-05,
+      "loss": 1.2676,
+      "step": 5436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187745165301427e-05,
+      "loss": 1.2344,
+      "step": 5437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185147435952413e-05,
+      "loss": 1.2373,
+      "step": 5438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182549551742112e-05,
+      "loss": 1.2559,
+      "step": 5439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217995151285463e-05,
+      "loss": 1.2334,
+      "step": 5440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2177353319474102e-05,
+      "loss": 1.2236,
+      "step": 5441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174754971784647e-05,
+      "loss": 1.251,
+      "step": 5442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172156469970422e-05,
+      "loss": 1.2627,
+      "step": 5443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2169557814215577e-05,
+      "loss": 1.2715,
+      "step": 5444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2166959004704279e-05,
+      "loss": 1.2695,
+      "step": 5445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216436004162071e-05,
+      "loss": 1.2627,
+      "step": 5446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161760925149051e-05,
+      "loss": 1.2441,
+      "step": 5447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159161655473506e-05,
+      "loss": 1.2451,
+      "step": 5448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156562232778282e-05,
+      "loss": 1.3037,
+      "step": 5449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153962657247604e-05,
+      "loss": 1.2559,
+      "step": 5450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151362929065703e-05,
+      "loss": 1.2764,
+      "step": 5451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2148763048416817e-05,
+      "loss": 1.249,
+      "step": 5452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146163015485203e-05,
+      "loss": 1.2256,
+      "step": 5453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2143562830455126e-05,
+      "loss": 1.29,
+      "step": 5454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140962493510862e-05,
+      "loss": 1.29,
+      "step": 5455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138362004836692e-05,
+      "loss": 1.2734,
+      "step": 5456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135761364616916e-05,
+      "loss": 1.2744,
+      "step": 5457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133160573035839e-05,
+      "loss": 1.3379,
+      "step": 5458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130559630277783e-05,
+      "loss": 1.2354,
+      "step": 5459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127958536527069e-05,
+      "loss": 1.2734,
+      "step": 5460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2125357291968047e-05,
+      "loss": 1.2334,
+      "step": 5461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212275589678506e-05,
+      "loss": 1.2324,
+      "step": 5462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120154351162467e-05,
+      "loss": 1.25,
+      "step": 5463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117552655284643e-05,
+      "loss": 1.1855,
+      "step": 5464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114950809335975e-05,
+      "loss": 1.2734,
+      "step": 5465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112348813500846e-05,
+      "loss": 1.2412,
+      "step": 5466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109746667963665e-05,
+      "loss": 1.2393,
+      "step": 5467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2107144372908844e-05,
+      "loss": 1.2285,
+      "step": 5468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210454192852081e-05,
+      "loss": 1.2822,
+      "step": 5469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101939334983994e-05,
+      "loss": 1.2344,
+      "step": 5470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2099336592482844e-05,
+      "loss": 1.2432,
+      "step": 5471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096733701201815e-05,
+      "loss": 1.2598,
+      "step": 5472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094130661325379e-05,
+      "loss": 1.2139,
+      "step": 5473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2091527473038007e-05,
+      "loss": 1.2617,
+      "step": 5474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2088924136524187e-05,
+      "loss": 1.2744,
+      "step": 5475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086320651968422e-05,
+      "loss": 1.3008,
+      "step": 5476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2083717019555218e-05,
+      "loss": 1.2812,
+      "step": 5477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2081113239469092e-05,
+      "loss": 1.251,
+      "step": 5478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2078509311894578e-05,
+      "loss": 1.2402,
+      "step": 5479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2075905237016215e-05,
+      "loss": 1.2363,
+      "step": 5480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073301015018551e-05,
+      "loss": 1.2061,
+      "step": 5481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070696646086151e-05,
+      "loss": 1.2256,
+      "step": 5482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068092130403586e-05,
+      "loss": 1.2061,
+      "step": 5483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065487468155435e-05,
+      "loss": 1.2686,
+      "step": 5484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062882659526289e-05,
+      "loss": 1.2559,
+      "step": 5485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2060277704700755e-05,
+      "loss": 1.2549,
+      "step": 5486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057672603863449e-05,
+      "loss": 1.2637,
+      "step": 5487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055067357198987e-05,
+      "loss": 1.2725,
+      "step": 5488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052461964892004e-05,
+      "loss": 1.2412,
+      "step": 5489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049856427127147e-05,
+      "loss": 1.2666,
+      "step": 5490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204725074408907e-05,
+      "loss": 1.2139,
+      "step": 5491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204464491596243e-05,
+      "loss": 1.25,
+      "step": 5492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042038942931917e-05,
+      "loss": 1.2422,
+      "step": 5493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039432825182204e-05,
+      "loss": 1.2344,
+      "step": 5494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036826562897992e-05,
+      "loss": 1.2676,
+      "step": 5495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034220156263981e-05,
+      "loss": 1.2861,
+      "step": 5496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031613605464894e-05,
+      "loss": 1.2373,
+      "step": 5497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2029006910685453e-05,
+      "loss": 1.2471,
+      "step": 5498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2026400072110394e-05,
+      "loss": 1.3115,
+      "step": 5499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2023793089924464e-05,
+      "loss": 1.2666,
+      "step": 5500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2021185964312423e-05,
+      "loss": 1.2871,
+      "step": 5501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018578695459034e-05,
+      "loss": 1.2715,
+      "step": 5502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2015971283549075e-05,
+      "loss": 1.1982,
+      "step": 5503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201336372876733e-05,
+      "loss": 1.2266,
+      "step": 5504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010756031298608e-05,
+      "loss": 1.2832,
+      "step": 5505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008148191327701e-05,
+      "loss": 1.2412,
+      "step": 5506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005540209039435e-05,
+      "loss": 1.2324,
+      "step": 5507
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002932084618636e-05,
+      "loss": 1.2402,
+      "step": 5508
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000323818250143e-05,
+      "loss": 1.2012,
+      "step": 5509
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997715410118799e-05,
+      "loss": 1.209,
+      "step": 5510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199510686040947e-05,
+      "loss": 1.2588,
+      "step": 5511
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992498169307016e-05,
+      "loss": 1.2178,
+      "step": 5512
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198988933699632e-05,
+      "loss": 1.2549,
+      "step": 5513
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987280363662264e-05,
+      "loss": 1.2373,
+      "step": 5514
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1984671249489753e-05,
+      "loss": 1.2021,
+      "step": 5515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198206199466369e-05,
+      "loss": 1.2451,
+      "step": 5516
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979452599368992e-05,
+      "loss": 1.2627,
+      "step": 5517
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976843063790588e-05,
+      "loss": 1.2754,
+      "step": 5518
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197423338811342e-05,
+      "loss": 1.2637,
+      "step": 5519
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197162357252243e-05,
+      "loss": 1.248,
+      "step": 5520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969013617202579e-05,
+      "loss": 1.2559,
+      "step": 5521
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196640352233883e-05,
+      "loss": 1.2256,
+      "step": 5522
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963793288116166e-05,
+      "loss": 1.2471,
+      "step": 5523
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196118291471957e-05,
+      "loss": 1.2266,
+      "step": 5524
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958572402334039e-05,
+      "loss": 1.2842,
+      "step": 5525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955961751144583e-05,
+      "loss": 1.3047,
+      "step": 5526
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953350961336216e-05,
+      "loss": 1.2441,
+      "step": 5527
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950740033093964e-05,
+      "loss": 1.2236,
+      "step": 5528
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948128966602867e-05,
+      "loss": 1.2598,
+      "step": 5529
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945517762047968e-05,
+      "loss": 1.2676,
+      "step": 5530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942906419614323e-05,
+      "loss": 1.2588,
+      "step": 5531
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940294939486998e-05,
+      "loss": 1.2275,
+      "step": 5532
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937683321851072e-05,
+      "loss": 1.2334,
+      "step": 5533
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935071566891626e-05,
+      "loss": 1.29,
+      "step": 5534
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932459674793755e-05,
+      "loss": 1.2334,
+      "step": 5535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929847645742565e-05,
+      "loss": 1.2197,
+      "step": 5536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192723547992317e-05,
+      "loss": 1.2305,
+      "step": 5537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924623177520698e-05,
+      "loss": 1.2197,
+      "step": 5538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922010738720275e-05,
+      "loss": 1.2725,
+      "step": 5539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919398163707048e-05,
+      "loss": 1.2588,
+      "step": 5540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191678545266617e-05,
+      "loss": 1.2686,
+      "step": 5541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914172605782807e-05,
+      "loss": 1.2158,
+      "step": 5542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911559623242125e-05,
+      "loss": 1.2354,
+      "step": 5543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908946505229315e-05,
+      "loss": 1.2607,
+      "step": 5544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906333251929559e-05,
+      "loss": 1.291,
+      "step": 5545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903719863528063e-05,
+      "loss": 1.1846,
+      "step": 5546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1901106340210036e-05,
+      "loss": 1.3115,
+      "step": 5547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18984926821607e-05,
+      "loss": 1.2764,
+      "step": 5548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895878889565284e-05,
+      "loss": 1.2822,
+      "step": 5549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1893264962609026e-05,
+      "loss": 1.2529,
+      "step": 5550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1890650901477177e-05,
+      "loss": 1.2314,
+      "step": 5551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1888036706354995e-05,
+      "loss": 1.2842,
+      "step": 5552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1885422377427747e-05,
+      "loss": 1.252,
+      "step": 5553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188280791488071e-05,
+      "loss": 1.2734,
+      "step": 5554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1880193318899171e-05,
+      "loss": 1.2432,
+      "step": 5555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877578589668428e-05,
+      "loss": 1.2119,
+      "step": 5556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1874963727373785e-05,
+      "loss": 1.2158,
+      "step": 5557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872348732200558e-05,
+      "loss": 1.2207,
+      "step": 5558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186973360433407e-05,
+      "loss": 1.2734,
+      "step": 5559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186711834395966e-05,
+      "loss": 1.2422,
+      "step": 5560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1864502951262662e-05,
+      "loss": 1.2637,
+      "step": 5561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1861887426428438e-05,
+      "loss": 1.2549,
+      "step": 5562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1859271769642346e-05,
+      "loss": 1.2402,
+      "step": 5563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1856655981089758e-05,
+      "loss": 1.2275,
+      "step": 5564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854040060956053e-05,
+      "loss": 1.2773,
+      "step": 5565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851424009426627e-05,
+      "loss": 1.2197,
+      "step": 5566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848807826686872e-05,
+      "loss": 1.252,
+      "step": 5567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1846191512922201e-05,
+      "loss": 1.2871,
+      "step": 5568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184357506831803e-05,
+      "loss": 1.2812,
+      "step": 5569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1840958493059789e-05,
+      "loss": 1.2734,
+      "step": 5570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838341787332912e-05,
+      "loss": 1.2568,
+      "step": 5571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1835724951322844e-05,
+      "loss": 1.2832,
+      "step": 5572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833107985215044e-05,
+      "loss": 1.2461,
+      "step": 5573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830490889194977e-05,
+      "loss": 1.2637,
+      "step": 5574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827873663448109e-05,
+      "loss": 1.2266,
+      "step": 5575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825256308159927e-05,
+      "loss": 1.2793,
+      "step": 5576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1822638823515924e-05,
+      "loss": 1.207,
+      "step": 5577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820021209701601e-05,
+      "loss": 1.2393,
+      "step": 5578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1817403466902465e-05,
+      "loss": 1.292,
+      "step": 5579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814785595304039e-05,
+      "loss": 1.2617,
+      "step": 5580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181216759509185e-05,
+      "loss": 1.2715,
+      "step": 5581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1809549466451435e-05,
+      "loss": 1.2158,
+      "step": 5582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180693120956834e-05,
+      "loss": 1.2607,
+      "step": 5583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804312824628124e-05,
+      "loss": 1.2832,
+      "step": 5584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1801694311816348e-05,
+      "loss": 1.2148,
+      "step": 5585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799075671318587e-05,
+      "loss": 1.2754,
+      "step": 5586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796456903320425e-05,
+      "loss": 1.252,
+      "step": 5587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793838008007456e-05,
+      "loss": 1.2676,
+      "step": 5588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791218985565276e-05,
+      "loss": 1.2842,
+      "step": 5589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1788599836179496e-05,
+      "loss": 1.2344,
+      "step": 5590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785980560035737e-05,
+      "loss": 1.3096,
+      "step": 5591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783361157319629e-05,
+      "loss": 1.252,
+      "step": 5592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1780741628216803e-05,
+      "loss": 1.2754,
+      "step": 5593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1778121972912908e-05,
+      "loss": 1.2646,
+      "step": 5594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17755021915936e-05,
+      "loss": 1.2305,
+      "step": 5595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1772882284444542e-05,
+      "loss": 1.2568,
+      "step": 5596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770262251651406e-05,
+      "loss": 1.2236,
+      "step": 5597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767642093399874e-05,
+      "loss": 1.2676,
+      "step": 5598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1765021809875637e-05,
+      "loss": 1.2705,
+      "step": 5599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1762401401264392e-05,
+      "loss": 1.2676,
+      "step": 5600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759780867751852e-05,
+      "loss": 1.2119,
+      "step": 5601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1757160209523729e-05,
+      "loss": 1.2139,
+      "step": 5602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754539426765752e-05,
+      "loss": 1.2695,
+      "step": 5603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751918519663652e-05,
+      "loss": 1.2695,
+      "step": 5604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749297488403178e-05,
+      "loss": 1.2676,
+      "step": 5605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746676333170081e-05,
+      "loss": 1.2344,
+      "step": 5606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744055054150119e-05,
+      "loss": 1.2588,
+      "step": 5607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741433651529062e-05,
+      "loss": 1.2568,
+      "step": 5608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738812125492691e-05,
+      "loss": 1.2656,
+      "step": 5609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736190476226796e-05,
+      "loss": 1.2178,
+      "step": 5610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173356870391717e-05,
+      "loss": 1.2676,
+      "step": 5611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730946808749613e-05,
+      "loss": 1.2334,
+      "step": 5612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1728324790909947e-05,
+      "loss": 1.2881,
+      "step": 5613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172570265058399e-05,
+      "loss": 1.248,
+      "step": 5614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723080387957571e-05,
+      "loss": 1.2451,
+      "step": 5615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1720458003216537e-05,
+      "loss": 1.2354,
+      "step": 5616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171783549654673e-05,
+      "loss": 1.2627,
+      "step": 5617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715212868134007e-05,
+      "loss": 1.2686,
+      "step": 5618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1712590118164235e-05,
+      "loss": 1.332,
+      "step": 5619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709967246823291e-05,
+      "loss": 1.252,
+      "step": 5620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170734425429705e-05,
+      "loss": 1.2793,
+      "step": 5621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704721140771411e-05,
+      "loss": 1.2451,
+      "step": 5622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702097906432272e-05,
+      "loss": 1.2852,
+      "step": 5623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1699474551465541e-05,
+      "loss": 1.2285,
+      "step": 5624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1696851076057134e-05,
+      "loss": 1.2764,
+      "step": 5625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1694227480392977e-05,
+      "loss": 1.2705,
+      "step": 5626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691603764659003e-05,
+      "loss": 1.2959,
+      "step": 5627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1688979929041158e-05,
+      "loss": 1.2129,
+      "step": 5628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1686355973725389e-05,
+      "loss": 1.2715,
+      "step": 5629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1683731898897658e-05,
+      "loss": 1.2969,
+      "step": 5630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1681107704743934e-05,
+      "loss": 1.2363,
+      "step": 5631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678483391450195e-05,
+      "loss": 1.2188,
+      "step": 5632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1675858959202415e-05,
+      "loss": 1.3193,
+      "step": 5633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1673234408186604e-05,
+      "loss": 1.2197,
+      "step": 5634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1670609738588752e-05,
+      "loss": 1.2959,
+      "step": 5635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166798495059487e-05,
+      "loss": 1.2334,
+      "step": 5636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1665360044390982e-05,
+      "loss": 1.2744,
+      "step": 5637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1662735020163111e-05,
+      "loss": 1.2354,
+      "step": 5638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660109878097294e-05,
+      "loss": 1.2236,
+      "step": 5639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657484618379574e-05,
+      "loss": 1.2705,
+      "step": 5640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1654859241196003e-05,
+      "loss": 1.21,
+      "step": 5641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652233746732642e-05,
+      "loss": 1.2646,
+      "step": 5642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649608135175558e-05,
+      "loss": 1.2568,
+      "step": 5643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164698240671083e-05,
+      "loss": 1.252,
+      "step": 5644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164435656152454e-05,
+      "loss": 1.2773,
+      "step": 5645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164173059980279e-05,
+      "loss": 1.2568,
+      "step": 5646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163910452173167e-05,
+      "loss": 1.334,
+      "step": 5647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1636478327497297e-05,
+      "loss": 1.2461,
+      "step": 5648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1633852017285787e-05,
+      "loss": 1.2812,
+      "step": 5649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163122559128327e-05,
+      "loss": 1.2656,
+      "step": 5650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1628599049675875e-05,
+      "loss": 1.2686,
+      "step": 5651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1625972392649751e-05,
+      "loss": 1.2256,
+      "step": 5652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1623345620391045e-05,
+      "loss": 1.248,
+      "step": 5653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1620718733085914e-05,
+      "loss": 1.2539,
+      "step": 5654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618091730920531e-05,
+      "loss": 1.249,
+      "step": 5655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615464614081071e-05,
+      "loss": 1.2861,
+      "step": 5656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1612837382753714e-05,
+      "loss": 1.249,
+      "step": 5657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610210037124651e-05,
+      "loss": 1.2217,
+      "step": 5658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607582577380088e-05,
+      "loss": 1.2314,
+      "step": 5659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1604955003706228e-05,
+      "loss": 1.2607,
+      "step": 5660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1602327316289288e-05,
+      "loss": 1.2725,
+      "step": 5661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1599699515315493e-05,
+      "loss": 1.2715,
+      "step": 5662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1597071600971072e-05,
+      "loss": 1.2412,
+      "step": 5663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159444357344227e-05,
+      "loss": 1.3066,
+      "step": 5664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1591815432915335e-05,
+      "loss": 1.29,
+      "step": 5665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589187179576516e-05,
+      "loss": 1.3018,
+      "step": 5666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586558813612085e-05,
+      "loss": 1.248,
+      "step": 5667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1583930335208311e-05,
+      "loss": 1.2461,
+      "step": 5668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581301744551473e-05,
+      "loss": 1.2598,
+      "step": 5669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1578673041827862e-05,
+      "loss": 1.2705,
+      "step": 5670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576044227223772e-05,
+      "loss": 1.2393,
+      "step": 5671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1573415300925508e-05,
+      "loss": 1.2285,
+      "step": 5672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.157078626311938e-05,
+      "loss": 1.2344,
+      "step": 5673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156815711399171e-05,
+      "loss": 1.2793,
+      "step": 5674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565527853728826e-05,
+      "loss": 1.2256,
+      "step": 5675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562898482517062e-05,
+      "loss": 1.2549,
+      "step": 5676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156026900054276e-05,
+      "loss": 1.2568,
+      "step": 5677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1557639407992273e-05,
+      "loss": 1.2969,
+      "step": 5678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555009705051965e-05,
+      "loss": 1.2295,
+      "step": 5679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552379891908193e-05,
+      "loss": 1.252,
+      "step": 5680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549749968747338e-05,
+      "loss": 1.2178,
+      "step": 5681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547119935755781e-05,
+      "loss": 1.2471,
+      "step": 5682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544489793119913e-05,
+      "loss": 1.2441,
+      "step": 5683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154185954102613e-05,
+      "loss": 1.2227,
+      "step": 5684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539229179660842e-05,
+      "loss": 1.2852,
+      "step": 5685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153659870921046e-05,
+      "loss": 1.2393,
+      "step": 5686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1533968129861404e-05,
+      "loss": 1.2383,
+      "step": 5687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1531337441800105e-05,
+      "loss": 1.2227,
+      "step": 5688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528706645213002e-05,
+      "loss": 1.2773,
+      "step": 5689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1526075740286533e-05,
+      "loss": 1.2168,
+      "step": 5690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1523444727207154e-05,
+      "loss": 1.249,
+      "step": 5691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1520813606161323e-05,
+      "loss": 1.2607,
+      "step": 5692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1518182377335512e-05,
+      "loss": 1.2422,
+      "step": 5693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151555104091619e-05,
+      "loss": 1.2344,
+      "step": 5694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512919597089845e-05,
+      "loss": 1.2559,
+      "step": 5695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510288046042964e-05,
+      "loss": 1.2363,
+      "step": 5696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1507656387962047e-05,
+      "loss": 1.2217,
+      "step": 5697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1505024623033593e-05,
+      "loss": 1.2617,
+      "step": 5698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502392751444127e-05,
+      "loss": 1.2529,
+      "step": 5699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149976077338016e-05,
+      "loss": 1.2549,
+      "step": 5700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1497128689028223e-05,
+      "loss": 1.2256,
+      "step": 5701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149449649857485e-05,
+      "loss": 1.2773,
+      "step": 5702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1491864202206591e-05,
+      "loss": 1.2461,
+      "step": 5703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489231800109985e-05,
+      "loss": 1.2812,
+      "step": 5704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486599292471602e-05,
+      "loss": 1.2383,
+      "step": 5705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1483966679478e-05,
+      "loss": 1.2764,
+      "step": 5706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481333961315757e-05,
+      "loss": 1.2822,
+      "step": 5707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478701138171448e-05,
+      "loss": 1.2314,
+      "step": 5708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476068210231666e-05,
+      "loss": 1.2881,
+      "step": 5709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473435177683003e-05,
+      "loss": 1.2812,
+      "step": 5710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470802040712069e-05,
+      "loss": 1.2598,
+      "step": 5711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1468168799505463e-05,
+      "loss": 1.2129,
+      "step": 5712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465535454249812e-05,
+      "loss": 1.3145,
+      "step": 5713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462902005131736e-05,
+      "loss": 1.2393,
+      "step": 5714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460268452337871e-05,
+      "loss": 1.2061,
+      "step": 5715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145763479605485e-05,
+      "loss": 1.252,
+      "step": 5716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455001036469331e-05,
+      "loss": 1.3076,
+      "step": 5717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452367173767957e-05,
+      "loss": 1.2207,
+      "step": 5718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14497332081374e-05,
+      "loss": 1.2148,
+      "step": 5719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447099139764319e-05,
+      "loss": 1.1982,
+      "step": 5720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444464968835401e-05,
+      "loss": 1.2822,
+      "step": 5721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144183069553732e-05,
+      "loss": 1.2432,
+      "step": 5722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1439196320056769e-05,
+      "loss": 1.2549,
+      "step": 5723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143656184258045e-05,
+      "loss": 1.2744,
+      "step": 5724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1433927263295068e-05,
+      "loss": 1.248,
+      "step": 5725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431292582387332e-05,
+      "loss": 1.2559,
+      "step": 5726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428657800043962e-05,
+      "loss": 1.2363,
+      "step": 5727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426022916451687e-05,
+      "loss": 1.2441,
+      "step": 5728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423387931797243e-05,
+      "loss": 1.2939,
+      "step": 5729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1420752846267365e-05,
+      "loss": 1.2373,
+      "step": 5730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1418117660048806e-05,
+      "loss": 1.2578,
+      "step": 5731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141548237332832e-05,
+      "loss": 1.2539,
+      "step": 5732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412846986292671e-05,
+      "loss": 1.2295,
+      "step": 5733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410211499128623e-05,
+      "loss": 1.2783,
+      "step": 5734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407575912022964e-05,
+      "loss": 1.2754,
+      "step": 5735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404940225162466e-05,
+      "loss": 1.249,
+      "step": 5736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402304438733929e-05,
+      "loss": 1.2334,
+      "step": 5737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399668552924146e-05,
+      "loss": 1.2793,
+      "step": 5738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397032567919926e-05,
+      "loss": 1.2666,
+      "step": 5739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394396483908078e-05,
+      "loss": 1.2539,
+      "step": 5740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1391760301075419e-05,
+      "loss": 1.2773,
+      "step": 5741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1389124019608779e-05,
+      "loss": 1.2686,
+      "step": 5742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1386487639694994e-05,
+      "loss": 1.1846,
+      "step": 5743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1383851161520896e-05,
+      "loss": 1.2549,
+      "step": 5744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1381214585273336e-05,
+      "loss": 1.2256,
+      "step": 5745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137857791113917e-05,
+      "loss": 1.2627,
+      "step": 5746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137594113930526e-05,
+      "loss": 1.2266,
+      "step": 5747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137330426995847e-05,
+      "loss": 1.293,
+      "step": 5748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1370667303285672e-05,
+      "loss": 1.249,
+      "step": 5749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368030239473755e-05,
+      "loss": 1.2178,
+      "step": 5750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365393078709605e-05,
+      "loss": 1.2695,
+      "step": 5751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1362755821180112e-05,
+      "loss": 1.2842,
+      "step": 5752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360118467072188e-05,
+      "loss": 1.2744,
+      "step": 5753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357481016572736e-05,
+      "loss": 1.2285,
+      "step": 5754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354843469868672e-05,
+      "loss": 1.2344,
+      "step": 5755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135220582714692e-05,
+      "loss": 1.2998,
+      "step": 5756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349568088594411e-05,
+      "loss": 1.2871,
+      "step": 5757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134693025439808e-05,
+      "loss": 1.2861,
+      "step": 5758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344292324744869e-05,
+      "loss": 1.251,
+      "step": 5759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134165429982173e-05,
+      "loss": 1.25,
+      "step": 5760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339016179815618e-05,
+      "loss": 1.2324,
+      "step": 5761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1336377964913497e-05,
+      "loss": 1.2959,
+      "step": 5762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333739655302338e-05,
+      "loss": 1.2314,
+      "step": 5763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331101251169116e-05,
+      "loss": 1.2588,
+      "step": 5764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132846275270082e-05,
+      "loss": 1.2119,
+      "step": 5765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1325824160084431e-05,
+      "loss": 1.2773,
+      "step": 5766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323185473506953e-05,
+      "loss": 1.2178,
+      "step": 5767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1320546693155388e-05,
+      "loss": 1.2578,
+      "step": 5768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317907819216747e-05,
+      "loss": 1.2432,
+      "step": 5769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131526885187804e-05,
+      "loss": 1.2988,
+      "step": 5770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312629791326301e-05,
+      "loss": 1.293,
+      "step": 5771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1309990637748556e-05,
+      "loss": 1.2529,
+      "step": 5772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1307351391331838e-05,
+      "loss": 1.2715,
+      "step": 5773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1304712052263193e-05,
+      "loss": 1.21,
+      "step": 5774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302072620729676e-05,
+      "loss": 1.2041,
+      "step": 5775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1299433096918336e-05,
+      "loss": 1.2393,
+      "step": 5776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296793481016235e-05,
+      "loss": 1.2441,
+      "step": 5777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294153773210447e-05,
+      "loss": 1.29,
+      "step": 5778
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1291513973688048e-05,
+      "loss": 1.2695,
+      "step": 5779
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288874082636121e-05,
+      "loss": 1.2959,
+      "step": 5780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1286234100241748e-05,
+      "loss": 1.2617,
+      "step": 5781
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1283594026692033e-05,
+      "loss": 1.2461,
+      "step": 5782
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280953862174075e-05,
+      "loss": 1.2373,
+      "step": 5783
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127831360687498e-05,
+      "loss": 1.292,
+      "step": 5784
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275673260981865e-05,
+      "loss": 1.2324,
+      "step": 5785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273032824681851e-05,
+      "loss": 1.2266,
+      "step": 5786
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1270392298162065e-05,
+      "loss": 1.2793,
+      "step": 5787
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126775168160964e-05,
+      "loss": 1.2051,
+      "step": 5788
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126511097521172e-05,
+      "loss": 1.2461,
+      "step": 5789
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262470179155446e-05,
+      "loss": 1.2949,
+      "step": 5790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1259829293627977e-05,
+      "loss": 1.2529,
+      "step": 5791
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125718831881647e-05,
+      "loss": 1.2246,
+      "step": 5792
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125454725490809e-05,
+      "loss": 1.2422,
+      "step": 5793
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251906102090011e-05,
+      "loss": 1.2637,
+      "step": 5794
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249264860549407e-05,
+      "loss": 1.2627,
+      "step": 5795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246623530473466e-05,
+      "loss": 1.2109,
+      "step": 5796
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243982112049383e-05,
+      "loss": 1.2373,
+      "step": 5797
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1241340605464346e-05,
+      "loss": 1.2539,
+      "step": 5798
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1238699010905565e-05,
+      "loss": 1.2695,
+      "step": 5799
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236057328560246e-05,
+      "loss": 1.2129,
+      "step": 5800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.123341555861561e-05,
+      "loss": 1.209,
+      "step": 5801
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230773701258874e-05,
+      "loss": 1.291,
+      "step": 5802
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228131756677268e-05,
+      "loss": 1.2324,
+      "step": 5803
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1225489725058027e-05,
+      "loss": 1.3184,
+      "step": 5804
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222847606588392e-05,
+      "loss": 1.2607,
+      "step": 5805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1220205401455609e-05,
+      "loss": 1.249,
+      "step": 5806
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121756310984693e-05,
+      "loss": 1.2412,
+      "step": 5807
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214920731949616e-05,
+      "loss": 1.2646,
+      "step": 5808
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212278267950932e-05,
+      "loss": 1.2559,
+      "step": 5809
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209635718038148e-05,
+      "loss": 1.2383,
+      "step": 5810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120699308239854e-05,
+      "loss": 1.2354,
+      "step": 5811
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1204350361219396e-05,
+      "loss": 1.2715,
+      "step": 5812
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1201707554688002e-05,
+      "loss": 1.2461,
+      "step": 5813
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1199064662991654e-05,
+      "loss": 1.249,
+      "step": 5814
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196421686317655e-05,
+      "loss": 1.2217,
+      "step": 5815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193778624853313e-05,
+      "loss": 1.3037,
+      "step": 5816
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191135478785937e-05,
+      "loss": 1.2861,
+      "step": 5817
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188492248302853e-05,
+      "loss": 1.2754,
+      "step": 5818
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185848933591382e-05,
+      "loss": 1.21,
+      "step": 5819
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118320553483886e-05,
+      "loss": 1.3057,
+      "step": 5820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180562052232617e-05,
+      "loss": 1.2031,
+      "step": 5821
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177918485960004e-05,
+      "loss": 1.1807,
+      "step": 5822
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175274836208368e-05,
+      "loss": 1.2578,
+      "step": 5823
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117263110316506e-05,
+      "loss": 1.2207,
+      "step": 5824
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116998728701745e-05,
+      "loss": 1.2568,
+      "step": 5825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11673433879529e-05,
+      "loss": 1.2715,
+      "step": 5826
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116469940615878e-05,
+      "loss": 1.2559,
+      "step": 5827
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1162055341822475e-05,
+      "loss": 1.2197,
+      "step": 5828
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1159411195131367e-05,
+      "loss": 1.249,
+      "step": 5829
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156766966272845e-05,
+      "loss": 1.2275,
+      "step": 5830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154122655434307e-05,
+      "loss": 1.2617,
+      "step": 5831
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1151478262803154e-05,
+      "loss": 1.2559,
+      "step": 5832
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1148833788566797e-05,
+      "loss": 1.2305,
+      "step": 5833
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146189232912648e-05,
+      "loss": 1.2432,
+      "step": 5834
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143544596028121e-05,
+      "loss": 1.2266,
+      "step": 5835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140899878100653e-05,
+      "loss": 1.2949,
+      "step": 5836
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138255079317664e-05,
+      "loss": 1.2793,
+      "step": 5837
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11356101998666e-05,
+      "loss": 1.2627,
+      "step": 5838
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132965239934892e-05,
+      "loss": 1.3086,
+      "step": 5839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113032019971e-05,
+      "loss": 1.2334,
+      "step": 5840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112767507937937e-05,
+      "loss": 1.2539,
+      "step": 5841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1125029879130467e-05,
+      "loss": 1.2129,
+      "step": 5842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122384599150751e-05,
+      "loss": 1.2969,
+      "step": 5843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119739239627699e-05,
+      "loss": 1.2168,
+      "step": 5844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111709380074878e-05,
+      "loss": 1.2061,
+      "step": 5845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1114448282701479e-05,
+      "loss": 1.2139,
+      "step": 5846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111802685673287e-05,
+      "loss": 1.2461,
+      "step": 5847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109157009851697e-05,
+      "loss": 1.2217,
+      "step": 5848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106511255424204e-05,
+      "loss": 1.249,
+      "step": 5849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103865422578313e-05,
+      "loss": 1.2363,
+      "step": 5850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1101219511501535e-05,
+      "loss": 1.2793,
+      "step": 5851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109857352238139e-05,
+      "loss": 1.2812,
+      "step": 5852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1095927455405391e-05,
+      "loss": 1.251,
+      "step": 5853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1093281310761073e-05,
+      "loss": 1.2607,
+      "step": 5854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109063508863596e-05,
+      "loss": 1.252,
+      "step": 5855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087988789217594e-05,
+      "loss": 1.25,
+      "step": 5856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085342412693519e-05,
+      "loss": 1.2363,
+      "step": 5857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1082695959251283e-05,
+      "loss": 1.2832,
+      "step": 5858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080049429078439e-05,
+      "loss": 1.2236,
+      "step": 5859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077402822362546e-05,
+      "loss": 1.3125,
+      "step": 5860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074756139291169e-05,
+      "loss": 1.2002,
+      "step": 5861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107210938005188e-05,
+      "loss": 1.2441,
+      "step": 5862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1069462544832252e-05,
+      "loss": 1.249,
+      "step": 5863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106681563381987e-05,
+      "loss": 1.2422,
+      "step": 5864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1064168647202316e-05,
+      "loss": 1.2559,
+      "step": 5865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1061521585167187e-05,
+      "loss": 1.2617,
+      "step": 5866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1058874447902075e-05,
+      "loss": 1.2217,
+      "step": 5867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1056227235594584e-05,
+      "loss": 1.251,
+      "step": 5868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105357994843232e-05,
+      "loss": 1.2842,
+      "step": 5869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050932586602905e-05,
+      "loss": 1.2559,
+      "step": 5870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1048285150293944e-05,
+      "loss": 1.2539,
+      "step": 5871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1045637639693072e-05,
+      "loss": 1.2637,
+      "step": 5872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042990054987912e-05,
+      "loss": 1.2363,
+      "step": 5873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10403423963661e-05,
+      "loss": 1.2959,
+      "step": 5874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1037694664015275e-05,
+      "loss": 1.2471,
+      "step": 5875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035046858123081e-05,
+      "loss": 1.2061,
+      "step": 5876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103239897887717e-05,
+      "loss": 1.2529,
+      "step": 5877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029751026465194e-05,
+      "loss": 1.2207,
+      "step": 5878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027103001074817e-05,
+      "loss": 1.2852,
+      "step": 5879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1024454902893701e-05,
+      "loss": 1.2666,
+      "step": 5880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102180673210952e-05,
+      "loss": 1.3135,
+      "step": 5881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019158488909944e-05,
+      "loss": 1.2637,
+      "step": 5882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101651017348266e-05,
+      "loss": 1.2109,
+      "step": 5883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013861786015355e-05,
+      "loss": 1.2568,
+      "step": 5884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1011213326695711e-05,
+      "loss": 1.2686,
+      "step": 5885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100856479571143e-05,
+      "loss": 1.2578,
+      "step": 5886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005916193250214e-05,
+      "loss": 1.2197,
+      "step": 5887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003267519499772e-05,
+      "loss": 1.2217,
+      "step": 5888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000618774647807e-05,
+      "loss": 1.2139,
+      "step": 5889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997969958882043e-05,
+      "loss": 1.249,
+      "step": 5890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0995321072390197e-05,
+      "loss": 1.2744,
+      "step": 5891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992672115359996e-05,
+      "loss": 1.2627,
+      "step": 5892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990023087979174e-05,
+      "loss": 1.2393,
+      "step": 5893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987373990435468e-05,
+      "loss": 1.2354,
+      "step": 5894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0984724822916616e-05,
+      "loss": 1.2852,
+      "step": 5895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0982075585610363e-05,
+      "loss": 1.2109,
+      "step": 5896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979426278704466e-05,
+      "loss": 1.2344,
+      "step": 5897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976776902386678e-05,
+      "loss": 1.252,
+      "step": 5898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974127456844761e-05,
+      "loss": 1.2559,
+      "step": 5899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971477942266478e-05,
+      "loss": 1.2803,
+      "step": 5900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968828358839603e-05,
+      "loss": 1.2383,
+      "step": 5901
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966178706751916e-05,
+      "loss": 1.249,
+      "step": 5902
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963528986191188e-05,
+      "loss": 1.291,
+      "step": 5903
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960879197345208e-05,
+      "loss": 1.2559,
+      "step": 5904
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958229340401771e-05,
+      "loss": 1.2441,
+      "step": 5905
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955579415548672e-05,
+      "loss": 1.2441,
+      "step": 5906
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09529294229737e-05,
+      "loss": 1.2383,
+      "step": 5907
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0950279362864678e-05,
+      "loss": 1.3262,
+      "step": 5908
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09476292354094e-05,
+      "loss": 1.3193,
+      "step": 5909
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944979040795687e-05,
+      "loss": 1.251,
+      "step": 5910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0942328779211358e-05,
+      "loss": 1.2871,
+      "step": 5911
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939678450844238e-05,
+      "loss": 1.2197,
+      "step": 5912
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093702805588215e-05,
+      "loss": 1.2637,
+      "step": 5913
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934377594512932e-05,
+      "loss": 1.2549,
+      "step": 5914
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0931727066924422e-05,
+      "loss": 1.2285,
+      "step": 5915
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929076473304463e-05,
+      "loss": 1.2754,
+      "step": 5916
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09264258138409e-05,
+      "loss": 1.2822,
+      "step": 5917
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923775088721587e-05,
+      "loss": 1.2256,
+      "step": 5918
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092112429813438e-05,
+      "loss": 1.2139,
+      "step": 5919
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918473442267144e-05,
+      "loss": 1.2256,
+      "step": 5920
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915822521307737e-05,
+      "loss": 1.2393,
+      "step": 5921
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913171535444036e-05,
+      "loss": 1.2578,
+      "step": 5922
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910520484863914e-05,
+      "loss": 1.2852,
+      "step": 5923
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0907869369755253e-05,
+      "loss": 1.2695,
+      "step": 5924
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0905218190305933e-05,
+      "loss": 1.2461,
+      "step": 5925
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090256694670385e-05,
+      "loss": 1.2344,
+      "step": 5926
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899915639136889e-05,
+      "loss": 1.2949,
+      "step": 5927
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897264267792952e-05,
+      "loss": 1.1855,
+      "step": 5928
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0894612832859945e-05,
+      "loss": 1.2344,
+      "step": 5929
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0891961334525774e-05,
+      "loss": 1.208,
+      "step": 5930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889309772978343e-05,
+      "loss": 1.2812,
+      "step": 5931
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0886658148405575e-05,
+      "loss": 1.2715,
+      "step": 5932
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088400646099539e-05,
+      "loss": 1.207,
+      "step": 5933
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0881354710935715e-05,
+      "loss": 1.2451,
+      "step": 5934
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878702898414471e-05,
+      "loss": 1.1846,
+      "step": 5935
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876051023619597e-05,
+      "loss": 1.2412,
+      "step": 5936
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0873399086739034e-05,
+      "loss": 1.2256,
+      "step": 5937
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0870747087960722e-05,
+      "loss": 1.2656,
+      "step": 5938
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0868095027472609e-05,
+      "loss": 1.2432,
+      "step": 5939
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086544290546264e-05,
+      "loss": 1.2715,
+      "step": 5940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0862790722118784e-05,
+      "loss": 1.249,
+      "step": 5941
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0860138477628987e-05,
+      "loss": 1.2461,
+      "step": 5942
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0857486172181223e-05,
+      "loss": 1.2373,
+      "step": 5943
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854833805963454e-05,
+      "loss": 1.2588,
+      "step": 5944
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852181379163662e-05,
+      "loss": 1.293,
+      "step": 5945
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0849528891969814e-05,
+      "loss": 1.251,
+      "step": 5946
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846876344569898e-05,
+      "loss": 1.292,
+      "step": 5947
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08442237371519e-05,
+      "loss": 1.2334,
+      "step": 5948
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841571069903808e-05,
+      "loss": 1.2373,
+      "step": 5949
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838918343013615e-05,
+      "loss": 1.2324,
+      "step": 5950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836265556669324e-05,
+      "loss": 1.2344,
+      "step": 5951
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833612711058931e-05,
+      "loss": 1.2725,
+      "step": 5952
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830959806370454e-05,
+      "loss": 1.2344,
+      "step": 5953
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828306842791891e-05,
+      "loss": 1.2236,
+      "step": 5954
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082565382051127e-05,
+      "loss": 1.2422,
+      "step": 5955
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08230007397166e-05,
+      "loss": 1.209,
+      "step": 5956
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820347600595913e-05,
+      "loss": 1.2539,
+      "step": 5957
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817694403337228e-05,
+      "loss": 1.2363,
+      "step": 5958
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081504114812859e-05,
+      "loss": 1.2363,
+      "step": 5959
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081238783515802e-05,
+      "loss": 1.2256,
+      "step": 5960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0809734464613567e-05,
+      "loss": 1.2539,
+      "step": 5961
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807081036683274e-05,
+      "loss": 1.2432,
+      "step": 5962
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804427551555192e-05,
+      "loss": 1.2148,
+      "step": 5963
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0801774009417368e-05,
+      "loss": 1.208,
+      "step": 5964
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799120410457858e-05,
+      "loss": 1.2656,
+      "step": 5965
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796466754864727e-05,
+      "loss": 1.2559,
+      "step": 5966
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0793813042826037e-05,
+      "loss": 1.2676,
+      "step": 5967
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791159274529858e-05,
+      "loss": 1.2158,
+      "step": 5968
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0788505450164261e-05,
+      "loss": 1.25,
+      "step": 5969
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785851569917323e-05,
+      "loss": 1.21,
+      "step": 5970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783197633977124e-05,
+      "loss": 1.2715,
+      "step": 5971
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780543642531744e-05,
+      "loss": 1.1826,
+      "step": 5972
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0777889595769282e-05,
+      "loss": 1.249,
+      "step": 5973
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0775235493877822e-05,
+      "loss": 1.2793,
+      "step": 5974
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0772581337045462e-05,
+      "loss": 1.2812,
+      "step": 5975
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0769927125460298e-05,
+      "loss": 1.2295,
+      "step": 5976
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767272859310442e-05,
+      "loss": 1.2217,
+      "step": 5977
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0764618538783998e-05,
+      "loss": 1.2197,
+      "step": 5978
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0761964164069072e-05,
+      "loss": 1.2607,
+      "step": 5979
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759309735353786e-05,
+      "loss": 1.3086,
+      "step": 5980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0756655252826262e-05,
+      "loss": 1.2607,
+      "step": 5981
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754000716674615e-05,
+      "loss": 1.251,
+      "step": 5982
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0751346127086974e-05,
+      "loss": 1.2451,
+      "step": 5983
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0748691484251475e-05,
+      "loss": 1.248,
+      "step": 5984
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0746036788356246e-05,
+      "loss": 1.2539,
+      "step": 5985
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743382039589427e-05,
+      "loss": 1.2852,
+      "step": 5986
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740727238139162e-05,
+      "loss": 1.2764,
+      "step": 5987
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0738072384193594e-05,
+      "loss": 1.2568,
+      "step": 5988
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735417477940877e-05,
+      "loss": 1.248,
+      "step": 5989
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0732762519569155e-05,
+      "loss": 1.2383,
+      "step": 5990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0730107509266594e-05,
+      "loss": 1.2158,
+      "step": 5991
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0727452447221349e-05,
+      "loss": 1.2031,
+      "step": 5992
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072479733362159e-05,
+      "loss": 1.2861,
+      "step": 5993
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0722142168655476e-05,
+      "loss": 1.2568,
+      "step": 5994
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0719486952511187e-05,
+      "loss": 1.2305,
+      "step": 5995
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716831685376889e-05,
+      "loss": 1.2656,
+      "step": 5996
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714176367440769e-05,
+      "loss": 1.2354,
+      "step": 5997
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0711520998891005e-05,
+      "loss": 1.2705,
+      "step": 5998
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708865579915787e-05,
+      "loss": 1.2373,
+      "step": 5999
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.07062101107033e-05,
+      "loss": 1.2666,
+      "step": 6000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703554591441736e-05,
+      "loss": 1.2598,
+      "step": 6001
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0700899022319293e-05,
+      "loss": 1.2646,
+      "step": 6002
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0698243403524175e-05,
+      "loss": 1.2256,
+      "step": 6003
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069558773524458e-05,
+      "loss": 1.2549,
+      "step": 6004
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069293201766872e-05,
+      "loss": 1.2021,
+      "step": 6005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690276250984801e-05,
+      "loss": 1.291,
+      "step": 6006
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687620435381043e-05,
+      "loss": 1.2188,
+      "step": 6007
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0684964571045652e-05,
+      "loss": 1.2617,
+      "step": 6008
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682308658166865e-05,
+      "loss": 1.2393,
+      "step": 6009
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679652696932892e-05,
+      "loss": 1.3086,
+      "step": 6010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676996687531971e-05,
+      "loss": 1.21,
+      "step": 6011
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674340630152324e-05,
+      "loss": 1.2256,
+      "step": 6012
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671684524982196e-05,
+      "loss": 1.2715,
+      "step": 6013
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0669028372209818e-05,
+      "loss": 1.2646,
+      "step": 6014
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0666372172023432e-05,
+      "loss": 1.249,
+      "step": 6015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0663715924611282e-05,
+      "loss": 1.2666,
+      "step": 6016
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661059630161623e-05,
+      "loss": 1.2285,
+      "step": 6017
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.06584032888627e-05,
+      "loss": 1.2783,
+      "step": 6018
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0655746900902767e-05,
+      "loss": 1.2529,
+      "step": 6019
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653090466470084e-05,
+      "loss": 1.291,
+      "step": 6020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0650433985752918e-05,
+      "loss": 1.2363,
+      "step": 6021
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0647777458939524e-05,
+      "loss": 1.2832,
+      "step": 6022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0645120886218174e-05,
+      "loss": 1.2764,
+      "step": 6023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0642464267777142e-05,
+      "loss": 1.2207,
+      "step": 6024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639807603804699e-05,
+      "loss": 1.2412,
+      "step": 6025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063715089448912e-05,
+      "loss": 1.2188,
+      "step": 6026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634494140018695e-05,
+      "loss": 1.2559,
+      "step": 6027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0631837340581699e-05,
+      "loss": 1.2422,
+      "step": 6028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629180496366425e-05,
+      "loss": 1.2139,
+      "step": 6029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0626523607561158e-05,
+      "loss": 1.249,
+      "step": 6030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06238666743542e-05,
+      "loss": 1.2217,
+      "step": 6031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621209696933836e-05,
+      "loss": 1.2354,
+      "step": 6032
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618552675488376e-05,
+      "loss": 1.2646,
+      "step": 6033
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0615895610206117e-05,
+      "loss": 1.2129,
+      "step": 6034
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0613238501275371e-05,
+      "loss": 1.2227,
+      "step": 6035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061058134888444e-05,
+      "loss": 1.2549,
+      "step": 6036
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0607924153221643e-05,
+      "loss": 1.3145,
+      "step": 6037
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060526691447529e-05,
+      "loss": 1.2412,
+      "step": 6038
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0602609632833702e-05,
+      "loss": 1.2832,
+      "step": 6039
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05999523084852e-05,
+      "loss": 1.2637,
+      "step": 6040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597294941618107e-05,
+      "loss": 1.2773,
+      "step": 6041
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594637532420753e-05,
+      "loss": 1.2725,
+      "step": 6042
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0591980081081468e-05,
+      "loss": 1.2217,
+      "step": 6043
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058932258778858e-05,
+      "loss": 1.2119,
+      "step": 6044
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586665052730436e-05,
+      "loss": 1.2764,
+      "step": 6045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584007476095366e-05,
+      "loss": 1.2266,
+      "step": 6046
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581349858071717e-05,
+      "loss": 1.2402,
+      "step": 6047
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0578692198847832e-05,
+      "loss": 1.2275,
+      "step": 6048
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0576034498612061e-05,
+      "loss": 1.2422,
+      "step": 6049
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573376757552756e-05,
+      "loss": 1.2275,
+      "step": 6050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570718975858267e-05,
+      "loss": 1.1543,
+      "step": 6051
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0568061153716954e-05,
+      "loss": 1.2285,
+      "step": 6052
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565403291317176e-05,
+      "loss": 1.1885,
+      "step": 6053
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562745388847296e-05,
+      "loss": 1.2031,
+      "step": 6054
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560087446495676e-05,
+      "loss": 1.2412,
+      "step": 6055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557429464450689e-05,
+      "loss": 1.2227,
+      "step": 6056
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0554771442900706e-05,
+      "loss": 1.2373,
+      "step": 6057
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0552113382034095e-05,
+      "loss": 1.2178,
+      "step": 6058
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0549455282039239e-05,
+      "loss": 1.2217,
+      "step": 6059
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0546797143104514e-05,
+      "loss": 1.2363,
+      "step": 6060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544138965418304e-05,
+      "loss": 1.2939,
+      "step": 6061
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054148074916899e-05,
+      "loss": 1.2637,
+      "step": 6062
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538822494544969e-05,
+      "loss": 1.2432,
+      "step": 6063
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536164201734622e-05,
+      "loss": 1.2959,
+      "step": 6064
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0533505870926344e-05,
+      "loss": 1.2725,
+      "step": 6065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530847502308532e-05,
+      "loss": 1.2783,
+      "step": 6066
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0528189096069586e-05,
+      "loss": 1.2031,
+      "step": 6067
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525530652397907e-05,
+      "loss": 1.2422,
+      "step": 6068
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0522872171481892e-05,
+      "loss": 1.2305,
+      "step": 6069
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0520213653509956e-05,
+      "loss": 1.2471,
+      "step": 6070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517555098670505e-05,
+      "loss": 1.2764,
+      "step": 6071
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514896507151952e-05,
+      "loss": 1.2422,
+      "step": 6072
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0512237879142705e-05,
+      "loss": 1.2725,
+      "step": 6073
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0509579214831189e-05,
+      "loss": 1.3047,
+      "step": 6074
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0506920514405818e-05,
+      "loss": 1.2695,
+      "step": 6075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0504261778055021e-05,
+      "loss": 1.2275,
+      "step": 6076
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050160300596721e-05,
+      "loss": 1.2461,
+      "step": 6077
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498944198330822e-05,
+      "loss": 1.2588,
+      "step": 6078
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0496285355334286e-05,
+      "loss": 1.2324,
+      "step": 6079
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049362647716603e-05,
+      "loss": 1.2383,
+      "step": 6080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049096756401449e-05,
+      "loss": 1.2334,
+      "step": 6081
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488308616068107e-05,
+      "loss": 1.2031,
+      "step": 6082
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485649633515314e-05,
+      "loss": 1.2803,
+      "step": 6083
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482990616544558e-05,
+      "loss": 1.2334,
+      "step": 6084
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048033156534428e-05,
+      "loss": 1.2383,
+      "step": 6085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047767248010293e-05,
+      "loss": 1.2402,
+      "step": 6086
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475013361008955e-05,
+      "loss": 1.2354,
+      "step": 6087
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472354208250806e-05,
+      "loss": 1.2529,
+      "step": 6088
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0469695022016936e-05,
+      "loss": 1.2734,
+      "step": 6089
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467035802495808e-05,
+      "loss": 1.2822,
+      "step": 6090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464376549875875e-05,
+      "loss": 1.2568,
+      "step": 6091
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04617172643456e-05,
+      "loss": 1.2002,
+      "step": 6092
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0459057946093442e-05,
+      "loss": 1.2422,
+      "step": 6093
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456398595307877e-05,
+      "loss": 1.2812,
+      "step": 6094
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045373921217736e-05,
+      "loss": 1.2705,
+      "step": 6095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0451079796890374e-05,
+      "loss": 1.2441,
+      "step": 6096
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448420349635381e-05,
+      "loss": 1.2578,
+      "step": 6097
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0445760870600863e-05,
+      "loss": 1.2627,
+      "step": 6098
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443101359975294e-05,
+      "loss": 1.2842,
+      "step": 6099
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0440441817947159e-05,
+      "loss": 1.2705,
+      "step": 6100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0437782244704928e-05,
+      "loss": 1.2666,
+      "step": 6101
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0435122640437091e-05,
+      "loss": 1.2607,
+      "step": 6102
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0432463005332138e-05,
+      "loss": 1.2324,
+      "step": 6103
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0429803339578554e-05,
+      "loss": 1.2363,
+      "step": 6104
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427143643364827e-05,
+      "loss": 1.2852,
+      "step": 6105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0424483916879454e-05,
+      "loss": 1.2217,
+      "step": 6106
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421824160310926e-05,
+      "loss": 1.2266,
+      "step": 6107
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0419164373847744e-05,
+      "loss": 1.2227,
+      "step": 6108
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0416504557678397e-05,
+      "loss": 1.2236,
+      "step": 6109
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413844711991398e-05,
+      "loss": 1.2402,
+      "step": 6110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411184836975245e-05,
+      "loss": 1.2139,
+      "step": 6111
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0408524932818445e-05,
+      "loss": 1.2393,
+      "step": 6112
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0405864999709499e-05,
+      "loss": 1.2334,
+      "step": 6113
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0403205037836925e-05,
+      "loss": 1.2373,
+      "step": 6114
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040054504738923e-05,
+      "loss": 1.252,
+      "step": 6115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397885028554925e-05,
+      "loss": 1.2305,
+      "step": 6116
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395224981522529e-05,
+      "loss": 1.2617,
+      "step": 6117
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0392564906480563e-05,
+      "loss": 1.2295,
+      "step": 6118
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038990480361754e-05,
+      "loss": 1.2568,
+      "step": 6119
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387244673121979e-05,
+      "loss": 1.2432,
+      "step": 6120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384584515182411e-05,
+      "loss": 1.2344,
+      "step": 6121
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381924329987358e-05,
+      "loss": 1.2979,
+      "step": 6122
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0379264117725347e-05,
+      "loss": 1.2354,
+      "step": 6123
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376603878584907e-05,
+      "loss": 1.2119,
+      "step": 6124
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037394361275457e-05,
+      "loss": 1.2441,
+      "step": 6125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371283320422867e-05,
+      "loss": 1.2412,
+      "step": 6126
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368623001778333e-05,
+      "loss": 1.2471,
+      "step": 6127
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365962657009508e-05,
+      "loss": 1.2891,
+      "step": 6128
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363302286304925e-05,
+      "loss": 1.2188,
+      "step": 6129
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360641889853132e-05,
+      "loss": 1.2383,
+      "step": 6130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357981467842661e-05,
+      "loss": 1.249,
+      "step": 6131
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355321020462066e-05,
+      "loss": 1.2617,
+      "step": 6132
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352660547899886e-05,
+      "loss": 1.2412,
+      "step": 6133
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0350000050344671e-05,
+      "loss": 1.2256,
+      "step": 6134
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034733952798497e-05,
+      "loss": 1.2236,
+      "step": 6135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344678981009337e-05,
+      "loss": 1.2529,
+      "step": 6136
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034201840960632e-05,
+      "loss": 1.208,
+      "step": 6137
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339357813964475e-05,
+      "loss": 1.252,
+      "step": 6138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0336697194272359e-05,
+      "loss": 1.2646,
+      "step": 6139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334036550718532e-05,
+      "loss": 1.2021,
+      "step": 6140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033137588349155e-05,
+      "loss": 1.2891,
+      "step": 6141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328715192779975e-05,
+      "loss": 1.2803,
+      "step": 6142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326054478772372e-05,
+      "loss": 1.2734,
+      "step": 6143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0323393741657306e-05,
+      "loss": 1.2793,
+      "step": 6144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320732981623335e-05,
+      "loss": 1.2051,
+      "step": 6145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0318072198859044e-05,
+      "loss": 1.251,
+      "step": 6146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0315411393552986e-05,
+      "loss": 1.1904,
+      "step": 6147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031275056589374e-05,
+      "loss": 1.2715,
+      "step": 6148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310089716069878e-05,
+      "loss": 1.2227,
+      "step": 6149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0307428844269974e-05,
+      "loss": 1.2666,
+      "step": 6150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304767950682602e-05,
+      "loss": 1.2188,
+      "step": 6151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030210703549634e-05,
+      "loss": 1.2285,
+      "step": 6152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299446098899768e-05,
+      "loss": 1.248,
+      "step": 6153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296785141081469e-05,
+      "loss": 1.2598,
+      "step": 6154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0294124162230019e-05,
+      "loss": 1.2188,
+      "step": 6155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291463162534005e-05,
+      "loss": 1.2373,
+      "step": 6156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288802142182013e-05,
+      "loss": 1.2764,
+      "step": 6157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286141101362627e-05,
+      "loss": 1.2148,
+      "step": 6158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283480040264437e-05,
+      "loss": 1.208,
+      "step": 6159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280818959076029e-05,
+      "loss": 1.2861,
+      "step": 6160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278157857985995e-05,
+      "loss": 1.2285,
+      "step": 6161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027549673718293e-05,
+      "loss": 1.2471,
+      "step": 6162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272835596855422e-05,
+      "loss": 1.1895,
+      "step": 6163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0270174437192074e-05,
+      "loss": 1.2734,
+      "step": 6164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267513258381475e-05,
+      "loss": 1.2344,
+      "step": 6165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0264852060612225e-05,
+      "loss": 1.2471,
+      "step": 6166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262190844072919e-05,
+      "loss": 1.2666,
+      "step": 6167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259529608952165e-05,
+      "loss": 1.2432,
+      "step": 6168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0256868355438561e-05,
+      "loss": 1.2607,
+      "step": 6169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025420708372071e-05,
+      "loss": 1.2725,
+      "step": 6170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251545793987212e-05,
+      "loss": 1.2803,
+      "step": 6171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024888448642668e-05,
+      "loss": 1.2783,
+      "step": 6172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0246223161227717e-05,
+      "loss": 1.2344,
+      "step": 6173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0243561818578928e-05,
+      "loss": 1.2422,
+      "step": 6174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0240900458668928e-05,
+      "loss": 1.2266,
+      "step": 6175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238239081686327e-05,
+      "loss": 1.2168,
+      "step": 6176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0235577687819731e-05,
+      "loss": 1.2441,
+      "step": 6177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232916277257757e-05,
+      "loss": 1.251,
+      "step": 6178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230254850189016e-05,
+      "loss": 1.2305,
+      "step": 6179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.022759340680213e-05,
+      "loss": 1.2451,
+      "step": 6180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224931947285706e-05,
+      "loss": 1.2539,
+      "step": 6181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0222270471828368e-05,
+      "loss": 1.2822,
+      "step": 6182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219608980618733e-05,
+      "loss": 1.2529,
+      "step": 6183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216947473845424e-05,
+      "loss": 1.2617,
+      "step": 6184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214285951697053e-05,
+      "loss": 1.1973,
+      "step": 6185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211624414362251e-05,
+      "loss": 1.2803,
+      "step": 6186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208962862029636e-05,
+      "loss": 1.2764,
+      "step": 6187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0206301294887834e-05,
+      "loss": 1.2061,
+      "step": 6188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020363971312547e-05,
+      "loss": 1.292,
+      "step": 6189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200978116931171e-05,
+      "loss": 1.1973,
+      "step": 6190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198316506493562e-05,
+      "loss": 1.2402,
+      "step": 6191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195654882001271e-05,
+      "loss": 1.2148,
+      "step": 6192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192993243642928e-05,
+      "loss": 1.1914,
+      "step": 6193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190331591607168e-05,
+      "loss": 1.3008,
+      "step": 6194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187669926082615e-05,
+      "loss": 1.2148,
+      "step": 6195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185008247257901e-05,
+      "loss": 1.2285,
+      "step": 6196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0182346555321664e-05,
+      "loss": 1.2578,
+      "step": 6197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017968485046254e-05,
+      "loss": 1.2314,
+      "step": 6198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0177023132869153e-05,
+      "loss": 1.2998,
+      "step": 6199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174361402730148e-05,
+      "loss": 1.2686,
+      "step": 6200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0171699660234163e-05,
+      "loss": 1.2559,
+      "step": 6201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169037905569827e-05,
+      "loss": 1.1885,
+      "step": 6202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0166376138925784e-05,
+      "loss": 1.2324,
+      "step": 6203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163714360490674e-05,
+      "loss": 1.2432,
+      "step": 6204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161052570453136e-05,
+      "loss": 1.2852,
+      "step": 6205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158390769001809e-05,
+      "loss": 1.2588,
+      "step": 6206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0155728956325337e-05,
+      "loss": 1.2227,
+      "step": 6207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015306713261236e-05,
+      "loss": 1.2715,
+      "step": 6208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150405298051526e-05,
+      "loss": 1.2363,
+      "step": 6209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147743452831471e-05,
+      "loss": 1.2422,
+      "step": 6210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0145081597140852e-05,
+      "loss": 1.2295,
+      "step": 6211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0142419731168303e-05,
+      "loss": 1.2344,
+      "step": 6212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139757855102479e-05,
+      "loss": 1.2783,
+      "step": 6213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0137095969132016e-05,
+      "loss": 1.2559,
+      "step": 6214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134434073445573e-05,
+      "loss": 1.2285,
+      "step": 6215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131772168231796e-05,
+      "loss": 1.2412,
+      "step": 6216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0129110253679329e-05,
+      "loss": 1.2207,
+      "step": 6217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0126448329976824e-05,
+      "loss": 1.2422,
+      "step": 6218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123786397312936e-05,
+      "loss": 1.2441,
+      "step": 6219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012112445587631e-05,
+      "loss": 1.2217,
+      "step": 6220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01184625058556e-05,
+      "loss": 1.1885,
+      "step": 6221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115800547439456e-05,
+      "loss": 1.2412,
+      "step": 6222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0113138580816537e-05,
+      "loss": 1.2188,
+      "step": 6223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110476606175492e-05,
+      "loss": 1.3369,
+      "step": 6224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107814623704974e-05,
+      "loss": 1.2783,
+      "step": 6225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0105152633593638e-05,
+      "loss": 1.2178,
+      "step": 6226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0102490636030145e-05,
+      "loss": 1.2422,
+      "step": 6227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009982863120314e-05,
+      "loss": 1.208,
+      "step": 6228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097166619301288e-05,
+      "loss": 1.2793,
+      "step": 6229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094504600513241e-05,
+      "loss": 1.2822,
+      "step": 6230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0091842575027661e-05,
+      "loss": 1.248,
+      "step": 6231
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089180543033197e-05,
+      "loss": 1.249,
+      "step": 6232
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086518504718519e-05,
+      "loss": 1.2021,
+      "step": 6233
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083856460272274e-05,
+      "loss": 1.2529,
+      "step": 6234
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081194409883129e-05,
+      "loss": 1.2178,
+      "step": 6235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078532353739737e-05,
+      "loss": 1.2686,
+      "step": 6236
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075870292030766e-05,
+      "loss": 1.208,
+      "step": 6237
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073208224944867e-05,
+      "loss": 1.2617,
+      "step": 6238
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070546152670705e-05,
+      "loss": 1.2314,
+      "step": 6239
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006788407539694e-05,
+      "loss": 1.2266,
+      "step": 6240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0065221993312238e-05,
+      "loss": 1.2822,
+      "step": 6241
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0062559906605254e-05,
+      "loss": 1.2656,
+      "step": 6242
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059897815464653e-05,
+      "loss": 1.2588,
+      "step": 6243
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0057235720079097e-05,
+      "loss": 1.248,
+      "step": 6244
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0054573620637249e-05,
+      "loss": 1.2051,
+      "step": 6245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0051911517327767e-05,
+      "loss": 1.2256,
+      "step": 6246
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0049249410339324e-05,
+      "loss": 1.2832,
+      "step": 6247
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046587299860576e-05,
+      "loss": 1.2422,
+      "step": 6248
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004392518608019e-05,
+      "loss": 1.2207,
+      "step": 6249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041263069186824e-05,
+      "loss": 1.2617,
+      "step": 6250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038600949369151e-05,
+      "loss": 1.2715,
+      "step": 6251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035938826815828e-05,
+      "loss": 1.2637,
+      "step": 6252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003327670171552e-05,
+      "loss": 1.2715,
+      "step": 6253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030614574256896e-05,
+      "loss": 1.2578,
+      "step": 6254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002795244462862e-05,
+      "loss": 1.2354,
+      "step": 6255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0025290313019355e-05,
+      "loss": 1.2539,
+      "step": 6256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022628179617766e-05,
+      "loss": 1.2588,
+      "step": 6257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001996604461252e-05,
+      "loss": 1.25,
+      "step": 6258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017303908192282e-05,
+      "loss": 1.2109,
+      "step": 6259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014641770545717e-05,
+      "loss": 1.2295,
+      "step": 6260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011979631861489e-05,
+      "loss": 1.2529,
+      "step": 6261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009317492328265e-05,
+      "loss": 1.209,
+      "step": 6262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0006655352134714e-05,
+      "loss": 1.207,
+      "step": 6263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003993211469492e-05,
+      "loss": 1.2393,
+      "step": 6264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0001331070521276e-05,
+      "loss": 1.2139,
+      "step": 6265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998668929478727e-06,
+      "loss": 1.2549,
+      "step": 6266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99600678853051e-06,
+      "loss": 1.2109,
+      "step": 6267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.993344647865293e-06,
+      "loss": 1.2334,
+      "step": 6268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990682507671738e-06,
+      "loss": 1.1885,
+      "step": 6269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988020368138515e-06,
+      "loss": 1.2588,
+      "step": 6270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985358229454286e-06,
+      "loss": 1.2803,
+      "step": 6271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982696091807722e-06,
+      "loss": 1.2314,
+      "step": 6272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.980033955387482e-06,
+      "loss": 1.3213,
+      "step": 6273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977371820382236e-06,
+      "loss": 1.2666,
+      "step": 6274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974709686980647e-06,
+      "loss": 1.2686,
+      "step": 6275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972047555371384e-06,
+      "loss": 1.2539,
+      "step": 6276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969385425743107e-06,
+      "loss": 1.21,
+      "step": 6277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966723298284482e-06,
+      "loss": 1.2158,
+      "step": 6278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964061173184173e-06,
+      "loss": 1.2324,
+      "step": 6279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961399050630856e-06,
+      "loss": 1.2529,
+      "step": 6280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95873693081318e-06,
+      "loss": 1.2637,
+      "step": 6281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.956074813919814e-06,
+      "loss": 1.2373,
+      "step": 6282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953412700139426e-06,
+      "loss": 1.2002,
+      "step": 6283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95075058966068e-06,
+      "loss": 1.1973,
+      "step": 6284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.948088482672234e-06,
+      "loss": 1.2412,
+      "step": 6285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.945426379362754e-06,
+      "loss": 1.2256,
+      "step": 6286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942764279920905e-06,
+      "loss": 1.2168,
+      "step": 6287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940102184535352e-06,
+      "loss": 1.2676,
+      "step": 6288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.937440093394749e-06,
+      "loss": 1.1445,
+      "step": 6289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.934778006687764e-06,
+      "loss": 1.2119,
+      "step": 6290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93211592460306e-06,
+      "loss": 1.2773,
+      "step": 6291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929453847329297e-06,
+      "loss": 1.2432,
+      "step": 6292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926791775055137e-06,
+      "loss": 1.209,
+      "step": 6293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.924129707969238e-06,
+      "loss": 1.2236,
+      "step": 6294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921467646260263e-06,
+      "loss": 1.2402,
+      "step": 6295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918805590116876e-06,
+      "loss": 1.293,
+      "step": 6296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916143539727728e-06,
+      "loss": 1.2441,
+      "step": 6297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913481495281485e-06,
+      "loss": 1.2168,
+      "step": 6298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910819456966803e-06,
+      "loss": 1.1982,
+      "step": 6299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.908157424972344e-06,
+      "loss": 1.2041,
+      "step": 6300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90549539948676e-06,
+      "loss": 1.1943,
+      "step": 6301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902833380698714e-06,
+      "loss": 1.2676,
+      "step": 6302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90017136879686e-06,
+      "loss": 1.2119,
+      "step": 6303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89750936396986e-06,
+      "loss": 1.2715,
+      "step": 6304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894847366406364e-06,
+      "loss": 1.2236,
+      "step": 6305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89218537629503e-06,
+      "loss": 1.2744,
+      "step": 6306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889523393824512e-06,
+      "loss": 1.251,
+      "step": 6307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.886861419183467e-06,
+      "loss": 1.2793,
+      "step": 6308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884199452560546e-06,
+      "loss": 1.1963,
+      "step": 6309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.881537494144405e-06,
+      "loss": 1.1934,
+      "step": 6310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878875544123694e-06,
+      "loss": 1.2764,
+      "step": 6311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876213602687069e-06,
+      "loss": 1.2822,
+      "step": 6312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873551670023177e-06,
+      "loss": 1.2393,
+      "step": 6313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870889746320674e-06,
+      "loss": 1.2529,
+      "step": 6314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868227831768206e-06,
+      "loss": 1.2529,
+      "step": 6315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86556592655443e-06,
+      "loss": 1.2256,
+      "step": 6316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862904030867987e-06,
+      "loss": 1.2285,
+      "step": 6317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860242144897524e-06,
+      "loss": 1.2314,
+      "step": 6318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857580268831697e-06,
+      "loss": 1.2588,
+      "step": 6319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.854918402859153e-06,
+      "loss": 1.2598,
+      "step": 6320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85225654716853e-06,
+      "loss": 1.2051,
+      "step": 6321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849594701948477e-06,
+      "loss": 1.1582,
+      "step": 6322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84693286738764e-06,
+      "loss": 1.248,
+      "step": 6323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84427104367467e-06,
+      "loss": 1.2266,
+      "step": 6324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841609230998194e-06,
+      "loss": 1.2441,
+      "step": 6325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838947429546867e-06,
+      "loss": 1.2334,
+      "step": 6326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.836285639509328e-06,
+      "loss": 1.2627,
+      "step": 6327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833623861074219e-06,
+      "loss": 1.2266,
+      "step": 6328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830962094430176e-06,
+      "loss": 1.2178,
+      "step": 6329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.828300339765842e-06,
+      "loss": 1.2051,
+      "step": 6330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.825638597269853e-06,
+      "loss": 1.251,
+      "step": 6331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822976867130849e-06,
+      "loss": 1.2324,
+      "step": 6332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820315149537466e-06,
+      "loss": 1.2021,
+      "step": 6333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.817653444678337e-06,
+      "loss": 1.2363,
+      "step": 6334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8149917527421e-06,
+      "loss": 1.2402,
+      "step": 6335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812330073917389e-06,
+      "loss": 1.2207,
+      "step": 6336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809668408392837e-06,
+      "loss": 1.2324,
+      "step": 6337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807006756357073e-06,
+      "loss": 1.2324,
+      "step": 6338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804345117998732e-06,
+      "loss": 1.2393,
+      "step": 6339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.801683493506442e-06,
+      "loss": 1.2637,
+      "step": 6340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.799021883068834e-06,
+      "loss": 1.2793,
+      "step": 6341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796360286874534e-06,
+      "loss": 1.25,
+      "step": 6342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.793698705112167e-06,
+      "loss": 1.2354,
+      "step": 6343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791037137970364e-06,
+      "loss": 1.2334,
+      "step": 6344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788375585637754e-06,
+      "loss": 1.251,
+      "step": 6345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78571404830295e-06,
+      "loss": 1.2373,
+      "step": 6346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78305252615458e-06,
+      "loss": 1.2695,
+      "step": 6347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.780391019381267e-06,
+      "loss": 1.2354,
+      "step": 6348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777729528171635e-06,
+      "loss": 1.2129,
+      "step": 6349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775068052714299e-06,
+      "loss": 1.2725,
+      "step": 6350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.772406593197874e-06,
+      "loss": 1.2207,
+      "step": 6351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.769745149810984e-06,
+      "loss": 1.251,
+      "step": 6352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767083722742248e-06,
+      "loss": 1.1826,
+      "step": 6353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764422312180272e-06,
+      "loss": 1.252,
+      "step": 6354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761760918313676e-06,
+      "loss": 1.2676,
+      "step": 6355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.759099541331072e-06,
+      "loss": 1.251,
+      "step": 6356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756438181421075e-06,
+      "loss": 1.2412,
+      "step": 6357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753776838772286e-06,
+      "loss": 1.2451,
+      "step": 6358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751115513573321e-06,
+      "loss": 1.3066,
+      "step": 6359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748454206012788e-06,
+      "loss": 1.2139,
+      "step": 6360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.745792916279296e-06,
+      "loss": 1.2207,
+      "step": 6361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743131644561442e-06,
+      "loss": 1.2529,
+      "step": 6362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740470391047838e-06,
+      "loss": 1.2305,
+      "step": 6363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737809155927081e-06,
+      "loss": 1.2041,
+      "step": 6364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.735147939387782e-06,
+      "loss": 1.2393,
+      "step": 6365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73248674161853e-06,
+      "loss": 1.2725,
+      "step": 6366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72982556280793e-06,
+      "loss": 1.2578,
+      "step": 6367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72716440314458e-06,
+      "loss": 1.2461,
+      "step": 6368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.724503262817074e-06,
+      "loss": 1.2441,
+      "step": 6369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721842142014007e-06,
+      "loss": 1.1973,
+      "step": 6370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719181040923975e-06,
+      "loss": 1.2363,
+      "step": 6371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716519959735566e-06,
+      "loss": 1.2256,
+      "step": 6372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.713858898637376e-06,
+      "loss": 1.207,
+      "step": 6373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71119785781799e-06,
+      "loss": 1.2686,
+      "step": 6374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.708536837465997e-06,
+      "loss": 1.2314,
+      "step": 6375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705875837769983e-06,
+      "loss": 1.2227,
+      "step": 6376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703214858918536e-06,
+      "loss": 1.2051,
+      "step": 6377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700553901100234e-06,
+      "loss": 1.2109,
+      "step": 6378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697892964503663e-06,
+      "loss": 1.2529,
+      "step": 6379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.6952320493174e-06,
+      "loss": 1.2324,
+      "step": 6380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692571155730031e-06,
+      "loss": 1.2246,
+      "step": 6381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.689910283930126e-06,
+      "loss": 1.2715,
+      "step": 6382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687249434106261e-06,
+      "loss": 1.209,
+      "step": 6383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684588606447014e-06,
+      "loss": 1.2402,
+      "step": 6384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681927801140961e-06,
+      "loss": 1.2852,
+      "step": 6385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679267018376666e-06,
+      "loss": 1.293,
+      "step": 6386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.676606258342697e-06,
+      "loss": 1.2588,
+      "step": 6387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67394552122763e-06,
+      "loss": 1.2246,
+      "step": 6388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67128480722003e-06,
+      "loss": 1.2246,
+      "step": 6389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.668624116508453e-06,
+      "loss": 1.2402,
+      "step": 6390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665963449281472e-06,
+      "loss": 1.2422,
+      "step": 6391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.663302805727643e-06,
+      "loss": 1.2334,
+      "step": 6392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660642186035529e-06,
+      "loss": 1.2451,
+      "step": 6393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.657981590393683e-06,
+      "loss": 1.2549,
+      "step": 6394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655321018990666e-06,
+      "loss": 1.2676,
+      "step": 6395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652660472015031e-06,
+      "loss": 1.2471,
+      "step": 6396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649999949655332e-06,
+      "loss": 1.2617,
+      "step": 6397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647339452100117e-06,
+      "loss": 1.2422,
+      "step": 6398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644678979537937e-06,
+      "loss": 1.1904,
+      "step": 6399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642018532157339e-06,
+      "loss": 1.2285,
+      "step": 6400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.639358110146873e-06,
+      "loss": 1.2266,
+      "step": 6401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636697713695076e-06,
+      "loss": 1.2148,
+      "step": 6402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634037342990495e-06,
+      "loss": 1.2441,
+      "step": 6403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631376998221669e-06,
+      "loss": 1.2295,
+      "step": 6404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628716679577138e-06,
+      "loss": 1.2539,
+      "step": 6405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626056387245434e-06,
+      "loss": 1.2461,
+      "step": 6406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623396121415096e-06,
+      "loss": 1.2109,
+      "step": 6407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620735882274656e-06,
+      "loss": 1.2246,
+      "step": 6408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618075670012645e-06,
+      "loss": 1.1895,
+      "step": 6409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615415484817592e-06,
+      "loss": 1.249,
+      "step": 6410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612755326878023e-06,
+      "loss": 1.2061,
+      "step": 6411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610095196382465e-06,
+      "loss": 1.2207,
+      "step": 6412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607435093519442e-06,
+      "loss": 1.2266,
+      "step": 6413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604775018477473e-06,
+      "loss": 1.207,
+      "step": 6414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602114971445078e-06,
+      "loss": 1.2119,
+      "step": 6415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.599454952610772e-06,
+      "loss": 1.2412,
+      "step": 6416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59679496216308e-06,
+      "loss": 1.2627,
+      "step": 6417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594135000290504e-06,
+      "loss": 1.2686,
+      "step": 6418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591475067181559e-06,
+      "loss": 1.249,
+      "step": 6419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588815163024757e-06,
+      "loss": 1.2471,
+      "step": 6420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586155288008607e-06,
+      "loss": 1.2002,
+      "step": 6421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.583495442321607e-06,
+      "loss": 1.2168,
+      "step": 6422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580835626152261e-06,
+      "loss": 1.2246,
+      "step": 6423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578175839689076e-06,
+      "loss": 1.2598,
+      "step": 6424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.575516083120551e-06,
+      "loss": 1.2744,
+      "step": 6425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572856356635175e-06,
+      "loss": 1.251,
+      "step": 6426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570196660421448e-06,
+      "loss": 1.2744,
+      "step": 6427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567536994667862e-06,
+      "loss": 1.2539,
+      "step": 6428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56487735956291e-06,
+      "loss": 1.2324,
+      "step": 6429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.562217755295076e-06,
+      "loss": 1.2646,
+      "step": 6430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.559558182052846e-06,
+      "loss": 1.2764,
+      "step": 6431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556898640024705e-06,
+      "loss": 1.2197,
+      "step": 6432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554239129399139e-06,
+      "loss": 1.208,
+      "step": 6433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55157965036462e-06,
+      "loss": 1.2393,
+      "step": 6434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.548920203109628e-06,
+      "loss": 1.2061,
+      "step": 6435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54626078782264e-06,
+      "loss": 1.209,
+      "step": 6436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543601404692128e-06,
+      "loss": 1.2617,
+      "step": 6437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54094205390656e-06,
+      "loss": 1.2686,
+      "step": 6438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538282735654404e-06,
+      "loss": 1.2441,
+      "step": 6439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535623450124127e-06,
+      "loss": 1.2529,
+      "step": 6440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532964197504195e-06,
+      "loss": 1.2314,
+      "step": 6441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530304977983065e-06,
+      "loss": 1.249,
+      "step": 6442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.527645791749199e-06,
+      "loss": 1.2236,
+      "step": 6443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52498663899105e-06,
+      "loss": 1.2588,
+      "step": 6444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.522327519897076e-06,
+      "loss": 1.2363,
+      "step": 6445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519668434655723e-06,
+      "loss": 1.1738,
+      "step": 6446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517009383455446e-06,
+      "loss": 1.2373,
+      "step": 6447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514350366484688e-06,
+      "loss": 1.2598,
+      "step": 6448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.511691383931898e-06,
+      "loss": 1.251,
+      "step": 6449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509032435985513e-06,
+      "loss": 1.2539,
+      "step": 6450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506373522833972e-06,
+      "loss": 1.2227,
+      "step": 6451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503714644665714e-06,
+      "loss": 1.3164,
+      "step": 6452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.501055801669181e-06,
+      "loss": 1.2168,
+      "step": 6453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498396994032794e-06,
+      "loss": 1.1621,
+      "step": 6454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495738221944984e-06,
+      "loss": 1.2275,
+      "step": 6455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493079485594182e-06,
+      "loss": 1.2637,
+      "step": 6456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490420785168816e-06,
+      "loss": 1.2266,
+      "step": 6457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487762120857297e-06,
+      "loss": 1.2188,
+      "step": 6458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485103492848052e-06,
+      "loss": 1.252,
+      "step": 6459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482444901329497e-06,
+      "loss": 1.2324,
+      "step": 6460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.479786346490049e-06,
+      "loss": 1.2568,
+      "step": 6461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47712782851811e-06,
+      "loss": 1.2129,
+      "step": 6462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474469347602099e-06,
+      "loss": 1.2666,
+      "step": 6463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471810903930416e-06,
+      "loss": 1.3037,
+      "step": 6464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.469152497691468e-06,
+      "loss": 1.2178,
+      "step": 6465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466494129073659e-06,
+      "loss": 1.2529,
+      "step": 6466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463835798265381e-06,
+      "loss": 1.2686,
+      "step": 6467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461177505455033e-06,
+      "loss": 1.2676,
+      "step": 6468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458519250831009e-06,
+      "loss": 1.2656,
+      "step": 6469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4558610345817e-06,
+      "loss": 1.2637,
+      "step": 6470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453202856895489e-06,
+      "loss": 1.2324,
+      "step": 6471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450544717960763e-06,
+      "loss": 1.2578,
+      "step": 6472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447886617965907e-06,
+      "loss": 1.2539,
+      "step": 6473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4452285570993e-06,
+      "loss": 1.2266,
+      "step": 6474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442570535549313e-06,
+      "loss": 1.2041,
+      "step": 6475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439912553504327e-06,
+      "loss": 1.2471,
+      "step": 6476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437254611152707e-06,
+      "loss": 1.2656,
+      "step": 6477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434596708682828e-06,
+      "loss": 1.2432,
+      "step": 6478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43193884628305e-06,
+      "loss": 1.2061,
+      "step": 6479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429281024141736e-06,
+      "loss": 1.2617,
+      "step": 6480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426623242447245e-06,
+      "loss": 1.1855,
+      "step": 6481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423965501387942e-06,
+      "loss": 1.2178,
+      "step": 6482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421307801152171e-06,
+      "loss": 1.2324,
+      "step": 6483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418650141928287e-06,
+      "loss": 1.2148,
+      "step": 6484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415992523904634e-06,
+      "loss": 1.3057,
+      "step": 6485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413334947269569e-06,
+      "loss": 1.2236,
+      "step": 6486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.410677412211421e-06,
+      "loss": 1.2412,
+      "step": 6487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408019918918535e-06,
+      "loss": 1.2305,
+      "step": 6488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405362467579249e-06,
+      "loss": 1.2197,
+      "step": 6489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402705058381898e-06,
+      "loss": 1.2607,
+      "step": 6490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400047691514803e-06,
+      "loss": 1.2959,
+      "step": 6491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397390367166301e-06,
+      "loss": 1.2803,
+      "step": 6492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39473308552471e-06,
+      "loss": 1.2246,
+      "step": 6493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.392075846778362e-06,
+      "loss": 1.2285,
+      "step": 6494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.389418651115561e-06,
+      "loss": 1.2246,
+      "step": 6495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386761498724632e-06,
+      "loss": 1.2354,
+      "step": 6496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384104389793883e-06,
+      "loss": 1.1782,
+      "step": 6497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.381447324511627e-06,
+      "loss": 1.2227,
+      "step": 6498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.378790303066166e-06,
+      "loss": 1.2119,
+      "step": 6499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376133325645805e-06,
+      "loss": 1.2314,
+      "step": 6500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.373476392438844e-06,
+      "loss": 1.248,
+      "step": 6501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370819503633578e-06,
+      "loss": 1.2441,
+      "step": 6502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368162659418303e-06,
+      "loss": 1.2256,
+      "step": 6503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365505859981307e-06,
+      "loss": 1.2246,
+      "step": 6504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362849105510881e-06,
+      "loss": 1.2832,
+      "step": 6505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.360192396195306e-06,
+      "loss": 1.2656,
+      "step": 6506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357535732222861e-06,
+      "loss": 1.2344,
+      "step": 6507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354879113781829e-06,
+      "loss": 1.249,
+      "step": 6508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.35222254106048e-06,
+      "loss": 1.2461,
+      "step": 6509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.349566014247087e-06,
+      "loss": 1.2373,
+      "step": 6510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346909533529917e-06,
+      "loss": 1.29,
+      "step": 6511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344253099097235e-06,
+      "loss": 1.2031,
+      "step": 6512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341596711137303e-06,
+      "loss": 1.2529,
+      "step": 6513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338940369838382e-06,
+      "loss": 1.1797,
+      "step": 6514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33628407538872e-06,
+      "loss": 1.2529,
+      "step": 6515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.333627827976573e-06,
+      "loss": 1.249,
+      "step": 6516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330971627790183e-06,
+      "loss": 1.3125,
+      "step": 6517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32831547501781e-06,
+      "loss": 1.2695,
+      "step": 6518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325659369847679e-06,
+      "loss": 1.2705,
+      "step": 6519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323003312468032e-06,
+      "loss": 1.2178,
+      "step": 6520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320347303067108e-06,
+      "loss": 1.2578,
+      "step": 6521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317691341833142e-06,
+      "loss": 1.2246,
+      "step": 6522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31503542895435e-06,
+      "loss": 1.2012,
+      "step": 6523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31237956461896e-06,
+      "loss": 1.2324,
+      "step": 6524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309723749015199e-06,
+      "loss": 1.2188,
+      "step": 6525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307067982331284e-06,
+      "loss": 1.2852,
+      "step": 6526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304412264755421e-06,
+      "loss": 1.2266,
+      "step": 6527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301756596475827e-06,
+      "loss": 1.2275,
+      "step": 6528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299100977680705e-06,
+      "loss": 1.209,
+      "step": 6529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.29644540855827e-06,
+      "loss": 1.2646,
+      "step": 6530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293789889296704e-06,
+      "loss": 1.1992,
+      "step": 6531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291134420084216e-06,
+      "loss": 1.2354,
+      "step": 6532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288479001108995e-06,
+      "loss": 1.2363,
+      "step": 6533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285823632559233e-06,
+      "loss": 1.2578,
+      "step": 6534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283168314623113e-06,
+      "loss": 1.2139,
+      "step": 6535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280513047488817e-06,
+      "loss": 1.2539,
+      "step": 6536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277857831344525e-06,
+      "loss": 1.2246,
+      "step": 6537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.275202666378415e-06,
+      "loss": 1.2715,
+      "step": 6538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272547552778653e-06,
+      "loss": 1.2383,
+      "step": 6539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269892490733407e-06,
+      "loss": 1.2422,
+      "step": 6540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267237480430846e-06,
+      "loss": 1.2676,
+      "step": 6541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26458252205913e-06,
+      "loss": 1.2617,
+      "step": 6542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26192761580641e-06,
+      "loss": 1.2578,
+      "step": 6543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25927276186084e-06,
+      "loss": 1.25,
+      "step": 6544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256617960410574e-06,
+      "loss": 1.2646,
+      "step": 6545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253963211643759e-06,
+      "loss": 1.21,
+      "step": 6546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25130851574853e-06,
+      "loss": 1.2314,
+      "step": 6547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248653872913027e-06,
+      "loss": 1.2773,
+      "step": 6548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.245999283325387e-06,
+      "loss": 1.2832,
+      "step": 6549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243344747173743e-06,
+      "loss": 1.2686,
+      "step": 6550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.240690264646216e-06,
+      "loss": 1.2412,
+      "step": 6551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23803583593093e-06,
+      "loss": 1.2314,
+      "step": 6552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.235381461216004e-06,
+      "loss": 1.2549,
+      "step": 6553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232727140689563e-06,
+      "loss": 1.1826,
+      "step": 6554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230072874539704e-06,
+      "loss": 1.2949,
+      "step": 6555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227418662954541e-06,
+      "loss": 1.2041,
+      "step": 6556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22476450612218e-06,
+      "loss": 1.1924,
+      "step": 6557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222110404230723e-06,
+      "loss": 1.2275,
+      "step": 6558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219456357468258e-06,
+      "loss": 1.2285,
+      "step": 6559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216802366022879e-06,
+      "loss": 1.2256,
+      "step": 6560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214148430082678e-06,
+      "loss": 1.2012,
+      "step": 6561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211494549835744e-06,
+      "loss": 1.2275,
+      "step": 6562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.208840725470143e-06,
+      "loss": 1.2363,
+      "step": 6563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.206186957173965e-06,
+      "loss": 1.2373,
+      "step": 6564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203533245135275e-06,
+      "loss": 1.2637,
+      "step": 6565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200879589542144e-06,
+      "loss": 1.2412,
+      "step": 6566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.198225990582636e-06,
+      "loss": 1.2695,
+      "step": 6567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195572448444811e-06,
+      "loss": 1.2285,
+      "step": 6568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.192918963316726e-06,
+      "loss": 1.2451,
+      "step": 6569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190265535386436e-06,
+      "loss": 1.2217,
+      "step": 6570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187612164841983e-06,
+      "loss": 1.2275,
+      "step": 6571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184958851871416e-06,
+      "loss": 1.2705,
+      "step": 6572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182305596662772e-06,
+      "loss": 1.251,
+      "step": 6573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179652399404092e-06,
+      "loss": 1.2334,
+      "step": 6574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176999260283402e-06,
+      "loss": 1.2344,
+      "step": 6575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174346179488733e-06,
+      "loss": 1.2754,
+      "step": 6576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171693157208109e-06,
+      "loss": 1.2568,
+      "step": 6577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16904019362955e-06,
+      "loss": 1.2637,
+      "step": 6578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16638728894107e-06,
+      "loss": 1.2598,
+      "step": 6579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16373444333068e-06,
+      "loss": 1.2949,
+      "step": 6580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.161081656986387e-06,
+      "loss": 1.2373,
+      "step": 6581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158428930096197e-06,
+      "loss": 1.1797,
+      "step": 6582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.155776262848104e-06,
+      "loss": 1.2031,
+      "step": 6583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153123655430103e-06,
+      "loss": 1.2578,
+      "step": 6584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.150471108030188e-06,
+      "loss": 1.1953,
+      "step": 6585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147818620836344e-06,
+      "loss": 1.2871,
+      "step": 6586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145166194036548e-06,
+      "loss": 1.2568,
+      "step": 6587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14251382781878e-06,
+      "loss": 1.2021,
+      "step": 6588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.139861522371013e-06,
+      "loss": 1.2139,
+      "step": 6589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137209277881223e-06,
+      "loss": 1.2266,
+      "step": 6590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134557094537362e-06,
+      "loss": 1.2529,
+      "step": 6591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131904972527394e-06,
+      "loss": 1.2695,
+      "step": 6592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12925291203928e-06,
+      "loss": 1.2061,
+      "step": 6593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12660091326097e-06,
+      "loss": 1.1943,
+      "step": 6594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123948976380407e-06,
+      "loss": 1.2705,
+      "step": 6595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121297101585532e-06,
+      "loss": 1.2881,
+      "step": 6596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11864528906429e-06,
+      "loss": 1.252,
+      "step": 6597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.115993539004611e-06,
+      "loss": 1.209,
+      "step": 6598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113341851594426e-06,
+      "loss": 1.2109,
+      "step": 6599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110690227021659e-06,
+      "loss": 1.1602,
+      "step": 6600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10803866547423e-06,
+      "loss": 1.2529,
+      "step": 6601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.105387167140055e-06,
+      "loss": 1.2646,
+      "step": 6602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10273573220705e-06,
+      "loss": 1.2402,
+      "step": 6603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.100084360863114e-06,
+      "loss": 1.2412,
+      "step": 6604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097433053296154e-06,
+      "loss": 1.2686,
+      "step": 6605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.094781809694067e-06,
+      "loss": 1.208,
+      "step": 6606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09213063024475e-06,
+      "loss": 1.2568,
+      "step": 6607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08947951513609e-06,
+      "loss": 1.2695,
+      "step": 6608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086828464555966e-06,
+      "loss": 1.2246,
+      "step": 6609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084177478692265e-06,
+      "loss": 1.1787,
+      "step": 6610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.081526557732863e-06,
+      "loss": 1.2549,
+      "step": 6611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078875701865622e-06,
+      "loss": 1.252,
+      "step": 6612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076224911278415e-06,
+      "loss": 1.2305,
+      "step": 6613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.073574186159102e-06,
+      "loss": 1.21,
+      "step": 6614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07092352669554e-06,
+      "loss": 1.1836,
+      "step": 6615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.068272933075582e-06,
+      "loss": 1.1973,
+      "step": 6616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06562240548707e-06,
+      "loss": 1.21,
+      "step": 6617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.062971944117851e-06,
+      "loss": 1.2227,
+      "step": 6618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060321549155769e-06,
+      "loss": 1.1953,
+      "step": 6619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.057671220788645e-06,
+      "loss": 1.2363,
+      "step": 6620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055020959204316e-06,
+      "loss": 1.2168,
+      "step": 6621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0523707645906e-06,
+      "loss": 1.2549,
+      "step": 6622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049720637135327e-06,
+      "loss": 1.2412,
+      "step": 6623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047070577026301e-06,
+      "loss": 1.2549,
+      "step": 6624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044420584451332e-06,
+      "loss": 1.2217,
+      "step": 6625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.041770659598229e-06,
+      "loss": 1.2207,
+      "step": 6626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039120802654795e-06,
+      "loss": 1.1826,
+      "step": 6627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036471013808815e-06,
+      "loss": 1.2461,
+      "step": 6628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033821293248089e-06,
+      "loss": 1.2031,
+      "step": 6629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031171641160397e-06,
+      "loss": 1.2061,
+      "step": 6630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028522057733527e-06,
+      "loss": 1.2402,
+      "step": 6631
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025872543155244e-06,
+      "loss": 1.2627,
+      "step": 6632
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023223097613324e-06,
+      "loss": 1.2207,
+      "step": 6633
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020573721295535e-06,
+      "loss": 1.2051,
+      "step": 6634
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017924414389639e-06,
+      "loss": 1.2529,
+      "step": 6635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015275177083388e-06,
+      "loss": 1.2578,
+      "step": 6636
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012626009564535e-06,
+      "loss": 1.2686,
+      "step": 6637
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009976912020826e-06,
+      "loss": 1.2158,
+      "step": 6638
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007327884640005e-06,
+      "loss": 1.2402,
+      "step": 6639
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.004678927609806e-06,
+      "loss": 1.3047,
+      "step": 6640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00203004111796e-06,
+      "loss": 1.2666,
+      "step": 6641
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999381225352194e-06,
+      "loss": 1.2822,
+      "step": 6642
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996732480500233e-06,
+      "loss": 1.2373,
+      "step": 6643
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994083806749788e-06,
+      "loss": 1.2529,
+      "step": 6644
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991435204288572e-06,
+      "loss": 1.1982,
+      "step": 6645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988786673304292e-06,
+      "loss": 1.2627,
+      "step": 6646
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986138213984652e-06,
+      "loss": 1.2402,
+      "step": 6647
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983489826517341e-06,
+      "loss": 1.1855,
+      "step": 6648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980841511090057e-06,
+      "loss": 1.2627,
+      "step": 6649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.978193267890484e-06,
+      "loss": 1.2695,
+      "step": 6650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975545097106304e-06,
+      "loss": 1.2793,
+      "step": 6651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972896998925187e-06,
+      "loss": 1.1826,
+      "step": 6652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970248973534808e-06,
+      "loss": 1.2842,
+      "step": 6653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96760102112283e-06,
+      "loss": 1.2061,
+      "step": 6654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964953141876924e-06,
+      "loss": 1.2861,
+      "step": 6655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962305335984728e-06,
+      "loss": 1.2314,
+      "step": 6656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959657603633904e-06,
+      "loss": 1.2324,
+      "step": 6657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957009945012088e-06,
+      "loss": 1.2041,
+      "step": 6658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.954362360306933e-06,
+      "loss": 1.2285,
+      "step": 6659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951714849706059e-06,
+      "loss": 1.2627,
+      "step": 6660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.949067413397098e-06,
+      "loss": 1.2344,
+      "step": 6661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946420051567679e-06,
+      "loss": 1.2393,
+      "step": 6662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943772764405421e-06,
+      "loss": 1.2822,
+      "step": 6663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941125552097928e-06,
+      "loss": 1.2373,
+      "step": 6664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938478414832816e-06,
+      "loss": 1.2754,
+      "step": 6665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935831352797684e-06,
+      "loss": 1.1826,
+      "step": 6666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933184366180135e-06,
+      "loss": 1.2158,
+      "step": 6667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.93053745516775e-06,
+      "loss": 1.2607,
+      "step": 6668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927890619948123e-06,
+      "loss": 1.2051,
+      "step": 6669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925243860708833e-06,
+      "loss": 1.2266,
+      "step": 6670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922597177637458e-06,
+      "loss": 1.1904,
+      "step": 6671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919950570921564e-06,
+      "loss": 1.2354,
+      "step": 6672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917304040748718e-06,
+      "loss": 1.249,
+      "step": 6673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914657587306481e-06,
+      "loss": 1.2373,
+      "step": 6674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912011210782408e-06,
+      "loss": 1.2227,
+      "step": 6675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.909364911364042e-06,
+      "loss": 1.2373,
+      "step": 6676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906718689238932e-06,
+      "loss": 1.2568,
+      "step": 6677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90407254459461e-06,
+      "loss": 1.208,
+      "step": 6678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901426477618615e-06,
+      "loss": 1.2461,
+      "step": 6679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898780488498467e-06,
+      "loss": 1.2207,
+      "step": 6680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89613457742169e-06,
+      "loss": 1.2441,
+      "step": 6681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8934887445758e-06,
+      "loss": 1.2891,
+      "step": 6682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890842990148308e-06,
+      "loss": 1.2637,
+      "step": 6683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888197314326716e-06,
+      "loss": 1.2188,
+      "step": 6684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.885551717298524e-06,
+      "loss": 1.2725,
+      "step": 6685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.882906199251225e-06,
+      "loss": 1.2197,
+      "step": 6686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880260760372308e-06,
+      "loss": 1.2354,
+      "step": 6687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877615400849252e-06,
+      "loss": 1.2549,
+      "step": 6688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874970120869537e-06,
+      "loss": 1.1934,
+      "step": 6689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872324920620629e-06,
+      "loss": 1.2305,
+      "step": 6690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869679800290005e-06,
+      "loss": 1.2344,
+      "step": 6691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867034760065111e-06,
+      "loss": 1.2578,
+      "step": 6692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864389800133403e-06,
+      "loss": 1.2412,
+      "step": 6693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861744920682336e-06,
+      "loss": 1.2207,
+      "step": 6694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859100121899352e-06,
+      "loss": 1.2314,
+      "step": 6695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85645540397188e-06,
+      "loss": 1.2441,
+      "step": 6696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853810767087356e-06,
+      "loss": 1.2881,
+      "step": 6697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851166211433204e-06,
+      "loss": 1.2568,
+      "step": 6698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.848521737196851e-06,
+      "loss": 1.2236,
+      "step": 6699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845877344565695e-06,
+      "loss": 1.2939,
+      "step": 6700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.843233033727157e-06,
+      "loss": 1.1895,
+      "step": 6701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.840588804868635e-06,
+      "loss": 1.2334,
+      "step": 6702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83794465817753e-06,
+      "loss": 1.248,
+      "step": 6703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835300593841221e-06,
+      "loss": 1.1885,
+      "step": 6704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832656612047103e-06,
+      "loss": 1.2295,
+      "step": 6705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830012712982551e-06,
+      "loss": 1.2256,
+      "step": 6706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827368896834941e-06,
+      "loss": 1.2432,
+      "step": 6707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824725163791636e-06,
+      "loss": 1.2832,
+      "step": 6708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.822081514039999e-06,
+      "loss": 1.2588,
+      "step": 6709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819437947767384e-06,
+      "loss": 1.2324,
+      "step": 6710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816794465161145e-06,
+      "loss": 1.1758,
+      "step": 6711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81415106640862e-06,
+      "loss": 1.2471,
+      "step": 6712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.811507751697149e-06,
+      "loss": 1.2148,
+      "step": 6713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808864521214065e-06,
+      "loss": 1.2178,
+      "step": 6714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806221375146692e-06,
+      "loss": 1.2373,
+      "step": 6715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803578313682348e-06,
+      "loss": 1.2354,
+      "step": 6716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800935337008348e-06,
+      "loss": 1.2705,
+      "step": 6717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.798292445312e-06,
+      "loss": 1.2744,
+      "step": 6718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795649638780609e-06,
+      "loss": 1.2314,
+      "step": 6719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793006917601463e-06,
+      "loss": 1.2402,
+      "step": 6720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790364281961857e-06,
+      "loss": 1.252,
+      "step": 6721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787721732049071e-06,
+      "loss": 1.2285,
+      "step": 6722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785079268050389e-06,
+      "loss": 1.2529,
+      "step": 6723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782436890153073e-06,
+      "loss": 1.2393,
+      "step": 6724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.779794598544394e-06,
+      "loss": 1.2305,
+      "step": 6725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.777152393411608e-06,
+      "loss": 1.2871,
+      "step": 6726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774510274941978e-06,
+      "loss": 1.2109,
+      "step": 6727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771868243322737e-06,
+      "loss": 1.2559,
+      "step": 6728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.769226298741128e-06,
+      "loss": 1.2061,
+      "step": 6729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.766584441384391e-06,
+      "loss": 1.2178,
+      "step": 6730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763942671439754e-06,
+      "loss": 1.2637,
+      "step": 6731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76130098909444e-06,
+      "loss": 1.2549,
+      "step": 6732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.758659394535657e-06,
+      "loss": 1.2412,
+      "step": 6733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75601788795062e-06,
+      "loss": 1.2246,
+      "step": 6734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753376469526535e-06,
+      "loss": 1.2002,
+      "step": 6735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.750735139450597e-06,
+      "loss": 1.2178,
+      "step": 6736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748093897909992e-06,
+      "loss": 1.2676,
+      "step": 6737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745452745091912e-06,
+      "loss": 1.2432,
+      "step": 6738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.742811681183532e-06,
+      "loss": 1.2695,
+      "step": 6739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.740170706372026e-06,
+      "loss": 1.2783,
+      "step": 6740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737529820844555e-06,
+      "loss": 1.1934,
+      "step": 6741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734889024788284e-06,
+      "loss": 1.2129,
+      "step": 6742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.732248318390361e-06,
+      "loss": 1.2402,
+      "step": 6743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729607701837938e-06,
+      "loss": 1.2402,
+      "step": 6744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726967175318152e-06,
+      "loss": 1.248,
+      "step": 6745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724326739018137e-06,
+      "loss": 1.252,
+      "step": 6746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.721686393125021e-06,
+      "loss": 1.2041,
+      "step": 6747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71904613782593e-06,
+      "loss": 1.2197,
+      "step": 6748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716405973307969e-06,
+      "loss": 1.2744,
+      "step": 6749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.713765899758255e-06,
+      "loss": 1.207,
+      "step": 6750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711125917363882e-06,
+      "loss": 1.2256,
+      "step": 6751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.708486026311955e-06,
+      "loss": 1.2568,
+      "step": 6752
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705846226789556e-06,
+      "loss": 1.2129,
+      "step": 6753
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703206518983768e-06,
+      "loss": 1.2217,
+      "step": 6754
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.700566903081666e-06,
+      "loss": 1.2275,
+      "step": 6755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69792737927033e-06,
+      "loss": 1.2295,
+      "step": 6756
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695287947736809e-06,
+      "loss": 1.252,
+      "step": 6757
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692648608668164e-06,
+      "loss": 1.2598,
+      "step": 6758
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.690009362251444e-06,
+      "loss": 1.2871,
+      "step": 6759
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687370208673702e-06,
+      "loss": 1.2236,
+      "step": 6760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.684731148121963e-06,
+      "loss": 1.2148,
+      "step": 6761
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682092180783258e-06,
+      "loss": 1.2324,
+      "step": 6762
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.679453306844612e-06,
+      "loss": 1.249,
+      "step": 6763
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676814526493052e-06,
+      "loss": 1.2588,
+      "step": 6764
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67417583991557e-06,
+      "loss": 1.251,
+      "step": 6765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671537247299183e-06,
+      "loss": 1.209,
+      "step": 6766
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668898748830884e-06,
+      "loss": 1.2051,
+      "step": 6767
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666260344697667e-06,
+      "loss": 1.2344,
+      "step": 6768
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663622035086505e-06,
+      "loss": 1.2559,
+      "step": 6769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660983820184384e-06,
+      "loss": 1.2412,
+      "step": 6770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658345700178272e-06,
+      "loss": 1.2422,
+      "step": 6771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655707675255133e-06,
+      "loss": 1.2227,
+      "step": 6772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.653069745601923e-06,
+      "loss": 1.2168,
+      "step": 6773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650431911405592e-06,
+      "loss": 1.2363,
+      "step": 6774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64779417285308e-06,
+      "loss": 1.2227,
+      "step": 6775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645156530131331e-06,
+      "loss": 1.1934,
+      "step": 6776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642518983427268e-06,
+      "loss": 1.2197,
+      "step": 6777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639881532927814e-06,
+      "loss": 1.2305,
+      "step": 6778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.637244178819888e-06,
+      "loss": 1.2344,
+      "step": 6779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6346069212904e-06,
+      "loss": 1.2812,
+      "step": 6780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631969760526248e-06,
+      "loss": 1.2148,
+      "step": 6781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62933269671433e-06,
+      "loss": 1.2422,
+      "step": 6782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626695730041534e-06,
+      "loss": 1.2578,
+      "step": 6783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.624058860694744e-06,
+      "loss": 1.2373,
+      "step": 6784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621422088860831e-06,
+      "loss": 1.2227,
+      "step": 6785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.618785414726666e-06,
+      "loss": 1.29,
+      "step": 6786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.616148838479105e-06,
+      "loss": 1.2461,
+      "step": 6787
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613512360305011e-06,
+      "loss": 1.2471,
+      "step": 6788
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610875980391223e-06,
+      "loss": 1.2314,
+      "step": 6789
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.608239698924583e-06,
+      "loss": 1.2539,
+      "step": 6790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.605603516091924e-06,
+      "loss": 1.2207,
+      "step": 6791
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602967432080079e-06,
+      "loss": 1.2852,
+      "step": 6792
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600331447075856e-06,
+      "loss": 1.2119,
+      "step": 6793
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597695561266074e-06,
+      "loss": 1.2305,
+      "step": 6794
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595059774837534e-06,
+      "loss": 1.2471,
+      "step": 6795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592424087977041e-06,
+      "loss": 1.2012,
+      "step": 6796
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589788500871379e-06,
+      "loss": 1.2441,
+      "step": 6797
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.587153013707332e-06,
+      "loss": 1.252,
+      "step": 6798
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584517626671682e-06,
+      "loss": 1.2588,
+      "step": 6799
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581882339951199e-06,
+      "loss": 1.2598,
+      "step": 6800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579247153732637e-06,
+      "loss": 1.2676,
+      "step": 6801
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57661206820276e-06,
+      "loss": 1.2412,
+      "step": 6802
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573977083548313e-06,
+      "loss": 1.2354,
+      "step": 6803
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.571342199956043e-06,
+      "loss": 1.1709,
+      "step": 6804
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56870741761267e-06,
+      "loss": 1.2041,
+      "step": 6805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566072736704933e-06,
+      "loss": 1.2451,
+      "step": 6806
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563438157419551e-06,
+      "loss": 1.2441,
+      "step": 6807
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560803679943233e-06,
+      "loss": 1.2422,
+      "step": 6808
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558169304462685e-06,
+      "loss": 1.2305,
+      "step": 6809
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.555535031164604e-06,
+      "loss": 1.2324,
+      "step": 6810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552900860235683e-06,
+      "loss": 1.2422,
+      "step": 6811
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550266791862605e-06,
+      "loss": 1.2637,
+      "step": 6812
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547632826232046e-06,
+      "loss": 1.2393,
+      "step": 6813
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544998963530672e-06,
+      "loss": 1.2734,
+      "step": 6814
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542365203945151e-06,
+      "loss": 1.21,
+      "step": 6815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539731547662134e-06,
+      "loss": 1.2695,
+      "step": 6816
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537097994868269e-06,
+      "loss": 1.1807,
+      "step": 6817
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534464545750191e-06,
+      "loss": 1.2236,
+      "step": 6818
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531831200494538e-06,
+      "loss": 1.2178,
+      "step": 6819
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529197959287936e-06,
+      "loss": 1.2646,
+      "step": 6820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.526564822316999e-06,
+      "loss": 1.2314,
+      "step": 6821
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523931789768337e-06,
+      "loss": 1.207,
+      "step": 6822
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521298861828555e-06,
+      "loss": 1.2021,
+      "step": 6823
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51866603868425e-06,
+      "loss": 1.2461,
+      "step": 6824
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516033320522004e-06,
+      "loss": 1.2559,
+      "step": 6825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5134007075284e-06,
+      "loss": 1.25,
+      "step": 6826
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510768199890013e-06,
+      "loss": 1.2158,
+      "step": 6827
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508135797793416e-06,
+      "loss": 1.248,
+      "step": 6828
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505503501425151e-06,
+      "loss": 1.2393,
+      "step": 6829
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50287131097178e-06,
+      "loss": 1.2637,
+      "step": 6830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50023922661984e-06,
+      "loss": 1.2227,
+      "step": 6831
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497607248555878e-06,
+      "loss": 1.1895,
+      "step": 6832
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494975376966408e-06,
+      "loss": 1.2285,
+      "step": 6833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492343612037955e-06,
+      "loss": 1.1826,
+      "step": 6834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489711953957038e-06,
+      "loss": 1.2354,
+      "step": 6835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48708040291016e-06,
+      "loss": 1.208,
+      "step": 6836
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484448959083811e-06,
+      "loss": 1.2363,
+      "step": 6837
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48181762266449e-06,
+      "loss": 1.2734,
+      "step": 6838
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479186393838677e-06,
+      "loss": 1.2539,
+      "step": 6839
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476555272792852e-06,
+      "loss": 1.2666,
+      "step": 6840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47392425971347e-06,
+      "loss": 1.2314,
+      "step": 6841
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471293354787003e-06,
+      "loss": 1.2129,
+      "step": 6842
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468662558199896e-06,
+      "loss": 1.2578,
+      "step": 6843
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.466031870138598e-06,
+      "loss": 1.2344,
+      "step": 6844
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.463401290789542e-06,
+      "loss": 1.248,
+      "step": 6845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46077082033916e-06,
+      "loss": 1.2373,
+      "step": 6846
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45814045897387e-06,
+      "loss": 1.25,
+      "step": 6847
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45551020688009e-06,
+      "loss": 1.2471,
+      "step": 6848
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.452880064244222e-06,
+      "loss": 1.21,
+      "step": 6849
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450250031252664e-06,
+      "loss": 1.1992,
+      "step": 6850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447620108091809e-06,
+      "loss": 1.2271,
+      "step": 6851
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444990294948042e-06,
+      "loss": 1.1914,
+      "step": 6852
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442360592007728e-06,
+      "loss": 1.2617,
+      "step": 6853
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439730999457243e-06,
+      "loss": 1.2266,
+      "step": 6854
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437101517482941e-06,
+      "loss": 1.2217,
+      "step": 6855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434472146271179e-06,
+      "loss": 1.2129,
+      "step": 6856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431842886008294e-06,
+      "loss": 1.251,
+      "step": 6857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.429213736880623e-06,
+      "loss": 1.2148,
+      "step": 6858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426584699074495e-06,
+      "loss": 1.2119,
+      "step": 6859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.423955772776233e-06,
+      "loss": 1.1733,
+      "step": 6860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421326958172142e-06,
+      "loss": 1.2334,
+      "step": 6861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41869825544853e-06,
+      "loss": 1.2686,
+      "step": 6862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41606966479169e-06,
+      "loss": 1.2305,
+      "step": 6863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413441186387916e-06,
+      "loss": 1.2295,
+      "step": 6864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410812820423489e-06,
+      "loss": 1.2598,
+      "step": 6865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40818456708467e-06,
+      "loss": 1.2588,
+      "step": 6866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405556426557731e-06,
+      "loss": 1.2568,
+      "step": 6867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402928399028928e-06,
+      "loss": 1.251,
+      "step": 6868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400300484684512e-06,
+      "loss": 1.2285,
+      "step": 6869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.397672683710713e-06,
+      "loss": 1.2314,
+      "step": 6870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395044996293773e-06,
+      "loss": 1.2705,
+      "step": 6871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392417422619912e-06,
+      "loss": 1.2012,
+      "step": 6872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.38978996287535e-06,
+      "loss": 1.2324,
+      "step": 6873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.38716261724629e-06,
+      "loss": 1.2256,
+      "step": 6874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384535385918932e-06,
+      "loss": 1.1875,
+      "step": 6875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381908269079469e-06,
+      "loss": 1.2773,
+      "step": 6876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379281266914087e-06,
+      "loss": 1.2432,
+      "step": 6877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376654379608959e-06,
+      "loss": 1.2178,
+      "step": 6878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374027607350252e-06,
+      "loss": 1.2529,
+      "step": 6879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.371400950324125e-06,
+      "loss": 1.2168,
+      "step": 6880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368774408716734e-06,
+      "loss": 1.2373,
+      "step": 6881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366147982714215e-06,
+      "loss": 1.252,
+      "step": 6882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363521672502706e-06,
+      "loss": 1.1904,
+      "step": 6883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360895478268332e-06,
+      "loss": 1.2119,
+      "step": 6884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358269400197216e-06,
+      "loss": 1.2568,
+      "step": 6885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355643438475461e-06,
+      "loss": 1.2314,
+      "step": 6886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353017593289172e-06,
+      "loss": 1.2686,
+      "step": 6887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350391864824445e-06,
+      "loss": 1.2549,
+      "step": 6888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347766253267363e-06,
+      "loss": 1.2285,
+      "step": 6889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.345140758804e-06,
+      "loss": 1.2119,
+      "step": 6890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342515381620428e-06,
+      "loss": 1.2852,
+      "step": 6891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339890121902705e-06,
+      "loss": 1.2471,
+      "step": 6892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337264979836892e-06,
+      "loss": 1.2764,
+      "step": 6893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.334639955609022e-06,
+      "loss": 1.2275,
+      "step": 6894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332015049405131e-06,
+      "loss": 1.2354,
+      "step": 6895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32939026141125e-06,
+      "loss": 1.1855,
+      "step": 6896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.326765591813401e-06,
+      "loss": 1.2627,
+      "step": 6897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324141040797587e-06,
+      "loss": 1.2314,
+      "step": 6898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321516608549808e-06,
+      "loss": 1.2051,
+      "step": 6899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318892295256065e-06,
+      "loss": 1.2266,
+      "step": 6900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.316268101102345e-06,
+      "loss": 1.2285,
+      "step": 6901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313644026274613e-06,
+      "loss": 1.2432,
+      "step": 6902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.311020070958844e-06,
+      "loss": 1.2646,
+      "step": 6903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308396235340998e-06,
+      "loss": 1.2451,
+      "step": 6904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30577251960703e-06,
+      "loss": 1.2119,
+      "step": 6905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30314892394287e-06,
+      "loss": 1.1777,
+      "step": 6906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30052544853446e-06,
+      "loss": 1.2422,
+      "step": 6907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29790209356773e-06,
+      "loss": 1.208,
+      "step": 6908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29527885922859e-06,
+      "loss": 1.2773,
+      "step": 6909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292655745702951e-06,
+      "loss": 1.2393,
+      "step": 6910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290032753176714e-06,
+      "loss": 1.2666,
+      "step": 6911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287409881835767e-06,
+      "loss": 1.2588,
+      "step": 6912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.284787131865996e-06,
+      "loss": 1.2305,
+      "step": 6913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.282164503453273e-06,
+      "loss": 1.2314,
+      "step": 6914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.279541996783466e-06,
+      "loss": 1.2578,
+      "step": 6915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276919612042429e-06,
+      "loss": 1.2705,
+      "step": 6916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274297349416014e-06,
+      "loss": 1.2168,
+      "step": 6917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271675209090057e-06,
+      "loss": 1.3086,
+      "step": 6918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269053191250388e-06,
+      "loss": 1.2148,
+      "step": 6919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266431296082836e-06,
+      "loss": 1.208,
+      "step": 6920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26380952377321e-06,
+      "loss": 1.2402,
+      "step": 6921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26118787450731e-06,
+      "loss": 1.2207,
+      "step": 6922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25856634847094e-06,
+      "loss": 1.2168,
+      "step": 6923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255944945849885e-06,
+      "loss": 1.1992,
+      "step": 6924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253323666829924e-06,
+      "loss": 1.2305,
+      "step": 6925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250702511596825e-06,
+      "loss": 1.2529,
+      "step": 6926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24808148033635e-06,
+      "loss": 1.2236,
+      "step": 6927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24546057323425e-06,
+      "loss": 1.2197,
+      "step": 6928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.242839790476276e-06,
+      "loss": 1.207,
+      "step": 6929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240219132248153e-06,
+      "loss": 1.2412,
+      "step": 6930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237598598735612e-06,
+      "loss": 1.2373,
+      "step": 6931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.234978190124365e-06,
+      "loss": 1.1953,
+      "step": 6932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232357906600131e-06,
+      "loss": 1.251,
+      "step": 6933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2297377483486e-06,
+      "loss": 1.2168,
+      "step": 6934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22711771555546e-06,
+      "loss": 1.2158,
+      "step": 6935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2244978084064e-06,
+      "loss": 1.1826,
+      "step": 6936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221878027087097e-06,
+      "loss": 1.2539,
+      "step": 6937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.219258371783199e-06,
+      "loss": 1.2461,
+      "step": 6938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216638842680374e-06,
+      "loss": 1.2197,
+      "step": 6939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214019439964264e-06,
+      "loss": 1.1924,
+      "step": 6940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21140016382051e-06,
+      "loss": 1.208,
+      "step": 6941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208781014434729e-06,
+      "loss": 1.2314,
+      "step": 6942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206161991992547e-06,
+      "loss": 1.2529,
+      "step": 6943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203543096679574e-06,
+      "loss": 1.2529,
+      "step": 6944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200924328681415e-06,
+      "loss": 1.2607,
+      "step": 6945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198305688183654e-06,
+      "loss": 1.209,
+      "step": 6946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195687175371878e-06,
+      "loss": 1.2148,
+      "step": 6947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193068790431662e-06,
+      "loss": 1.2676,
+      "step": 6948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190450533548568e-06,
+      "loss": 1.2734,
+      "step": 6949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.187832404908153e-06,
+      "loss": 1.1768,
+      "step": 6950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185214404695963e-06,
+      "loss": 1.2129,
+      "step": 6951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182596533097537e-06,
+      "loss": 1.2627,
+      "step": 6952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179978790298404e-06,
+      "loss": 1.2002,
+      "step": 6953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17736117648408e-06,
+      "loss": 1.2129,
+      "step": 6954
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174743691840076e-06,
+      "loss": 1.2217,
+      "step": 6955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.172126336551894e-06,
+      "loss": 1.2471,
+      "step": 6956
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16950911080503e-06,
+      "loss": 1.3047,
+      "step": 6957
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166892014784958e-06,
+      "loss": 1.2148,
+      "step": 6958
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.164275048677158e-06,
+      "loss": 1.2324,
+      "step": 6959
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16165821266709e-06,
+      "loss": 1.2695,
+      "step": 6960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159041506940215e-06,
+      "loss": 1.209,
+      "step": 6961
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156424931681974e-06,
+      "loss": 1.1572,
+      "step": 6962
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153808487077804e-06,
+      "loss": 1.2373,
+      "step": 6963
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15119217331313e-06,
+      "loss": 1.2197,
+      "step": 6964
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148575990573378e-06,
+      "loss": 1.2666,
+      "step": 6965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145959939043948e-06,
+      "loss": 1.2764,
+      "step": 6966
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.143344018910245e-06,
+      "loss": 1.2314,
+      "step": 6967
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140728230357654e-06,
+      "loss": 1.2168,
+      "step": 6968
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.138112573571566e-06,
+      "loss": 1.2188,
+      "step": 6969
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135497048737341e-06,
+      "loss": 1.249,
+      "step": 6970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132881656040344e-06,
+      "loss": 1.2246,
+      "step": 6971
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13026639566593e-06,
+      "loss": 1.2227,
+      "step": 6972
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127651267799447e-06,
+      "loss": 1.2256,
+      "step": 6973
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125036272626217e-06,
+      "loss": 1.2314,
+      "step": 6974
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.122421410331573e-06,
+      "loss": 1.2256,
+      "step": 6975
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11980668110083e-06,
+      "loss": 1.1885,
+      "step": 6976
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117192085119295e-06,
+      "loss": 1.2363,
+      "step": 6977
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.114577622572256e-06,
+      "loss": 1.2334,
+      "step": 6978
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111963293645008e-06,
+      "loss": 1.2305,
+      "step": 6979
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109349098522825e-06,
+      "loss": 1.2598,
+      "step": 6980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.106735037390976e-06,
+      "loss": 1.2217,
+      "step": 6981
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104121110434719e-06,
+      "loss": 1.2266,
+      "step": 6982
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1015073178393e-06,
+      "loss": 1.2207,
+      "step": 6983
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.098893659789965e-06,
+      "loss": 1.1729,
+      "step": 6984
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09628013647194e-06,
+      "loss": 1.2432,
+      "step": 6985
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093666748070443e-06,
+      "loss": 1.2324,
+      "step": 6986
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.091053494770687e-06,
+      "loss": 1.2227,
+      "step": 6987
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088440376757873e-06,
+      "loss": 1.2139,
+      "step": 6988
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085827394217196e-06,
+      "loss": 1.1582,
+      "step": 6989
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083214547333831e-06,
+      "loss": 1.2168,
+      "step": 6990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080601836292955e-06,
+      "loss": 1.2324,
+      "step": 6991
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.077989261279728e-06,
+      "loss": 1.2207,
+      "step": 6992
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075376822479309e-06,
+      "loss": 1.1992,
+      "step": 6993
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072764520076832e-06,
+      "loss": 1.2744,
+      "step": 6994
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070152354257437e-06,
+      "loss": 1.1904,
+      "step": 6995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067540325206248e-06,
+      "loss": 1.2139,
+      "step": 6996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064928433108374e-06,
+      "loss": 1.2168,
+      "step": 6997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062316678148933e-06,
+      "loss": 1.2373,
+      "step": 6998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059705060513005e-06,
+      "loss": 1.2295,
+      "step": 6999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057093580385678e-06,
+      "loss": 1.2578,
+      "step": 7000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054482237952032e-06,
+      "loss": 1.2266,
+      "step": 7001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051871033397138e-06,
+      "loss": 1.2969,
+      "step": 7002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04925996690604e-06,
+      "loss": 1.1885,
+      "step": 7003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046649038663787e-06,
+      "loss": 1.2441,
+      "step": 7004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044038248855419e-06,
+      "loss": 1.2764,
+      "step": 7005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041427597665964e-06,
+      "loss": 1.2227,
+      "step": 7006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038817085280432e-06,
+      "loss": 1.1855,
+      "step": 7007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036206711883835e-06,
+      "loss": 1.2227,
+      "step": 7008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03359647766117e-06,
+      "loss": 1.2227,
+      "step": 7009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030986382797425e-06,
+      "loss": 1.2471,
+      "step": 7010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.028376427477571e-06,
+      "loss": 1.2275,
+      "step": 7011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.025766611886581e-06,
+      "loss": 1.2393,
+      "step": 7012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02315693620941e-06,
+      "loss": 1.1992,
+      "step": 7013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020547400631011e-06,
+      "loss": 1.2139,
+      "step": 7014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.017938005336315e-06,
+      "loss": 1.2227,
+      "step": 7015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.015328750510249e-06,
+      "loss": 1.1865,
+      "step": 7016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012719636337737e-06,
+      "loss": 1.2334,
+      "step": 7017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010110663003685e-06,
+      "loss": 1.1934,
+      "step": 7018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007501830692986e-06,
+      "loss": 1.1934,
+      "step": 7019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004893139590533e-06,
+      "loss": 1.2227,
+      "step": 7020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002284589881201e-06,
+      "loss": 1.2002,
+      "step": 7021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99967618174986e-06,
+      "loss": 1.1973,
+      "step": 7022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997067915381367e-06,
+      "loss": 1.293,
+      "step": 7023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994459790960566e-06,
+      "loss": 1.2715,
+      "step": 7024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991851808672302e-06,
+      "loss": 1.2256,
+      "step": 7025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.989243968701397e-06,
+      "loss": 1.2529,
+      "step": 7026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986636271232671e-06,
+      "loss": 1.2139,
+      "step": 7027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98402871645093e-06,
+      "loss": 1.2305,
+      "step": 7028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981421304540966e-06,
+      "loss": 1.209,
+      "step": 7029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97881403568758e-06,
+      "loss": 1.209,
+      "step": 7030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976206910075539e-06,
+      "loss": 1.2803,
+      "step": 7031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97359992788961e-06,
+      "loss": 1.2227,
+      "step": 7032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970993089314548e-06,
+      "loss": 1.2471,
+      "step": 7033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96838639453511e-06,
+      "loss": 1.2188,
+      "step": 7034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965779843736022e-06,
+      "loss": 1.2441,
+      "step": 7035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963173437102011e-06,
+      "loss": 1.2822,
+      "step": 7036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960567174817798e-06,
+      "loss": 1.2021,
+      "step": 7037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957961057068088e-06,
+      "loss": 1.2041,
+      "step": 7038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955355084037571e-06,
+      "loss": 1.2236,
+      "step": 7039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952749255910934e-06,
+      "loss": 1.2217,
+      "step": 7040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950143572872854e-06,
+      "loss": 1.2295,
+      "step": 7041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947538035108002e-06,
+      "loss": 1.2148,
+      "step": 7042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944932642801017e-06,
+      "loss": 1.2246,
+      "step": 7043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942327396136553e-06,
+      "loss": 1.2285,
+      "step": 7044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939722295299243e-06,
+      "loss": 1.2744,
+      "step": 7045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937117340473713e-06,
+      "loss": 1.2031,
+      "step": 7046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934512531844569e-06,
+      "loss": 1.2754,
+      "step": 7047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931907869596418e-06,
+      "loss": 1.3213,
+      "step": 7048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92930335391385e-06,
+      "loss": 1.2861,
+      "step": 7049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92669898498145e-06,
+      "loss": 1.2334,
+      "step": 7050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924094762983788e-06,
+      "loss": 1.1562,
+      "step": 7051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.921490688105424e-06,
+      "loss": 1.2217,
+      "step": 7052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91888676053091e-06,
+      "loss": 1.2256,
+      "step": 7053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916282980444787e-06,
+      "loss": 1.2422,
+      "step": 7054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.913679348031582e-06,
+      "loss": 1.25,
+      "step": 7055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.911075863475816e-06,
+      "loss": 1.2334,
+      "step": 7056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908472526961996e-06,
+      "loss": 1.2266,
+      "step": 7057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905869338674627e-06,
+      "loss": 1.2412,
+      "step": 7058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.903266298798187e-06,
+      "loss": 1.2549,
+      "step": 7059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900663407517159e-06,
+      "loss": 1.2031,
+      "step": 7060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89806066501601e-06,
+      "loss": 1.2227,
+      "step": 7061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895458071479195e-06,
+      "loss": 1.2109,
+      "step": 7062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89285562709116e-06,
+      "loss": 1.2051,
+      "step": 7063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890253332036339e-06,
+      "loss": 1.1934,
+      "step": 7064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887651186499155e-06,
+      "loss": 1.25,
+      "step": 7065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88504919066403e-06,
+      "loss": 1.21,
+      "step": 7066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882447344715358e-06,
+      "loss": 1.252,
+      "step": 7067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879845648837536e-06,
+      "loss": 1.2324,
+      "step": 7068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.877244103214942e-06,
+      "loss": 1.2432,
+      "step": 7069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.874642708031958e-06,
+      "loss": 1.2578,
+      "step": 7070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872041463472934e-06,
+      "loss": 1.2217,
+      "step": 7071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.869440369722222e-06,
+      "loss": 1.2412,
+      "step": 7072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866839426964163e-06,
+      "loss": 1.2051,
+      "step": 7073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86423863538309e-06,
+      "loss": 1.1992,
+      "step": 7074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861637995163311e-06,
+      "loss": 1.2002,
+      "step": 7075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859037506489141e-06,
+      "loss": 1.2256,
+      "step": 7076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.856437169544875e-06,
+      "loss": 1.2783,
+      "step": 7077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853836984514802e-06,
+      "loss": 1.2539,
+      "step": 7078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851236951583185e-06,
+      "loss": 1.2471,
+      "step": 7079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8486370709343e-06,
+      "loss": 1.2725,
+      "step": 7080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846037342752397e-06,
+      "loss": 1.2549,
+      "step": 7081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84343776722172e-06,
+      "loss": 1.2266,
+      "step": 7082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840838344526498e-06,
+      "loss": 1.2451,
+      "step": 7083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83823907485095e-06,
+      "loss": 1.2197,
+      "step": 7084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835639958379292e-06,
+      "loss": 1.2119,
+      "step": 7085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833040995295723e-06,
+      "loss": 1.2363,
+      "step": 7086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.830442185784424e-06,
+      "loss": 1.2617,
+      "step": 7087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82784353002958e-06,
+      "loss": 1.2324,
+      "step": 7088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825245028215354e-06,
+      "loss": 1.2656,
+      "step": 7089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822646680525903e-06,
+      "loss": 1.2061,
+      "step": 7090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.820048487145371e-06,
+      "loss": 1.2256,
+      "step": 7091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817450448257894e-06,
+      "loss": 1.2783,
+      "step": 7092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814852564047589e-06,
+      "loss": 1.2324,
+      "step": 7093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.812254834698578e-06,
+      "loss": 1.2656,
+      "step": 7094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80965726039495e-06,
+      "loss": 1.2314,
+      "step": 7095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807059841320802e-06,
+      "loss": 1.293,
+      "step": 7096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804462577660214e-06,
+      "loss": 1.2305,
+      "step": 7097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.801865469597253e-06,
+      "loss": 1.1719,
+      "step": 7098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799268517315971e-06,
+      "loss": 1.1885,
+      "step": 7099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79667172100042e-06,
+      "loss": 1.252,
+      "step": 7100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794075080834628e-06,
+      "loss": 1.2207,
+      "step": 7101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791478597002633e-06,
+      "loss": 1.2607,
+      "step": 7102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788882269688431e-06,
+      "loss": 1.1904,
+      "step": 7103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78628609907603e-06,
+      "loss": 1.2227,
+      "step": 7104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78369008534942e-06,
+      "loss": 1.3086,
+      "step": 7105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781094228692588e-06,
+      "loss": 1.21,
+      "step": 7106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77849852928949e-06,
+      "loss": 1.1797,
+      "step": 7107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775902987324087e-06,
+      "loss": 1.2158,
+      "step": 7108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.773307602980329e-06,
+      "loss": 1.2344,
+      "step": 7109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770712376442153e-06,
+      "loss": 1.1846,
+      "step": 7110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76811730789347e-06,
+      "loss": 1.2236,
+      "step": 7111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765522397518204e-06,
+      "loss": 1.2695,
+      "step": 7112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762927645500252e-06,
+      "loss": 1.2109,
+      "step": 7113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.760333052023506e-06,
+      "loss": 1.2041,
+      "step": 7114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75773861727184e-06,
+      "loss": 1.2578,
+      "step": 7115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755144341429125e-06,
+      "loss": 1.1875,
+      "step": 7116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.752550224679217e-06,
+      "loss": 1.2607,
+      "step": 7117
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.749956267205964e-06,
+      "loss": 1.2197,
+      "step": 7118
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747362469193192e-06,
+      "loss": 1.2158,
+      "step": 7119
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.74476883082473e-06,
+      "loss": 1.2227,
+      "step": 7120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742175352284384e-06,
+      "loss": 1.2002,
+      "step": 7121
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73958203375596e-06,
+      "loss": 1.1982,
+      "step": 7122
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73698887542324e-06,
+      "loss": 1.2002,
+      "step": 7123
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734395877470005e-06,
+      "loss": 1.1875,
+      "step": 7124
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731803040080018e-06,
+      "loss": 1.2959,
+      "step": 7125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729210363437038e-06,
+      "loss": 1.2207,
+      "step": 7126
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7266178477248e-06,
+      "loss": 1.2031,
+      "step": 7127
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724025493127042e-06,
+      "loss": 1.2021,
+      "step": 7128
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721433299827482e-06,
+      "loss": 1.2227,
+      "step": 7129
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718841268009827e-06,
+      "loss": 1.2793,
+      "step": 7130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716249397857779e-06,
+      "loss": 1.248,
+      "step": 7131
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713657689555017e-06,
+      "loss": 1.2002,
+      "step": 7132
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71106614328522e-06,
+      "loss": 1.207,
+      "step": 7133
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.708474759232046e-06,
+      "loss": 1.209,
+      "step": 7134
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705883537579155e-06,
+      "loss": 1.2275,
+      "step": 7135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70329247851018e-06,
+      "loss": 1.2275,
+      "step": 7136
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700701582208747e-06,
+      "loss": 1.2451,
+      "step": 7137
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698110848858478e-06,
+      "loss": 1.2129,
+      "step": 7138
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.695520278642983e-06,
+      "loss": 1.1953,
+      "step": 7139
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692929871745843e-06,
+      "loss": 1.2363,
+      "step": 7140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690339628350643e-06,
+      "loss": 1.2266,
+      "step": 7141
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68774954864096e-06,
+      "loss": 1.2529,
+      "step": 7142
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.685159632800354e-06,
+      "loss": 1.2412,
+      "step": 7143
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68256988101236e-06,
+      "loss": 1.2168,
+      "step": 7144
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679980293460522e-06,
+      "loss": 1.2715,
+      "step": 7145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677390870328366e-06,
+      "loss": 1.2832,
+      "step": 7146
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674801611799402e-06,
+      "loss": 1.2549,
+      "step": 7147
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672212518057128e-06,
+      "loss": 1.2627,
+      "step": 7148
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.669623589285033e-06,
+      "loss": 1.1943,
+      "step": 7149
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667034825666598e-06,
+      "loss": 1.2568,
+      "step": 7150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664446227385288e-06,
+      "loss": 1.2607,
+      "step": 7151
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661857794624555e-06,
+      "loss": 1.1924,
+      "step": 7152
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659269527567839e-06,
+      "loss": 1.2471,
+      "step": 7153
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656681426398575e-06,
+      "loss": 1.2119,
+      "step": 7154
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654093491300181e-06,
+      "loss": 1.2158,
+      "step": 7155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65150572245606e-06,
+      "loss": 1.2725,
+      "step": 7156
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648918120049608e-06,
+      "loss": 1.2598,
+      "step": 7157
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.646330684264211e-06,
+      "loss": 1.2314,
+      "step": 7158
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643743415283241e-06,
+      "loss": 1.2432,
+      "step": 7159
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641156313290054e-06,
+      "loss": 1.2441,
+      "step": 7160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.638569378467999e-06,
+      "loss": 1.2754,
+      "step": 7161
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635982611000413e-06,
+      "loss": 1.21,
+      "step": 7162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633396011070622e-06,
+      "loss": 1.2744,
+      "step": 7163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630809578861933e-06,
+      "loss": 1.167,
+      "step": 7164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.628223314557649e-06,
+      "loss": 1.208,
+      "step": 7165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.625637218341057e-06,
+      "loss": 1.2432,
+      "step": 7166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.62305129039544e-06,
+      "loss": 1.2285,
+      "step": 7167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620465530904053e-06,
+      "loss": 1.2061,
+      "step": 7168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6178799400501545e-06,
+      "loss": 1.2695,
+      "step": 7169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615294518016981e-06,
+      "loss": 1.2676,
+      "step": 7170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612709264987772e-06,
+      "loss": 1.2578,
+      "step": 7171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610124181145731e-06,
+      "loss": 1.2998,
+      "step": 7172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607539266674066e-06,
+      "loss": 1.2305,
+      "step": 7173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.604954521755974e-06,
+      "loss": 1.2412,
+      "step": 7174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602369946574637e-06,
+      "loss": 1.2217,
+      "step": 7175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599785541313217e-06,
+      "loss": 1.2422,
+      "step": 7176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5972013061548725e-06,
+      "loss": 1.2656,
+      "step": 7177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594617241282751e-06,
+      "loss": 1.291,
+      "step": 7178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5920333468799876e-06,
+      "loss": 1.167,
+      "step": 7179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589449623129693e-06,
+      "loss": 1.2051,
+      "step": 7180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586866070214985e-06,
+      "loss": 1.2266,
+      "step": 7181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.584282688318954e-06,
+      "loss": 1.2256,
+      "step": 7182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581699477624689e-06,
+      "loss": 1.1797,
+      "step": 7183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579116438315257e-06,
+      "loss": 1.2109,
+      "step": 7184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576533570573721e-06,
+      "loss": 1.2412,
+      "step": 7185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573950874583128e-06,
+      "loss": 1.3057,
+      "step": 7186
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5713683505265155e-06,
+      "loss": 1.2393,
+      "step": 7187
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568785998586903e-06,
+      "loss": 1.2461,
+      "step": 7188
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566203818947303e-06,
+      "loss": 1.252,
+      "step": 7189
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563621811790715e-06,
+      "loss": 1.2119,
+      "step": 7190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561039977300127e-06,
+      "loss": 1.2373,
+      "step": 7191
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5584583156585124e-06,
+      "loss": 1.2051,
+      "step": 7192
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555876827048831e-06,
+      "loss": 1.2598,
+      "step": 7193
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553295511654036e-06,
+      "loss": 1.2354,
+      "step": 7194
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550714369657066e-06,
+      "loss": 1.2168,
+      "step": 7195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548133401240842e-06,
+      "loss": 1.2256,
+      "step": 7196
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.54555260658828e-06,
+      "loss": 1.2314,
+      "step": 7197
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5429719858822805e-06,
+      "loss": 1.2412,
+      "step": 7198
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540391539305734e-06,
+      "loss": 1.2329,
+      "step": 7199
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537811267041512e-06,
+      "loss": 1.2354,
+      "step": 7200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53523116927248e-06,
+      "loss": 1.2285,
+      "step": 7201
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532651246181488e-06,
+      "loss": 1.2471,
+      "step": 7202
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530071497951384e-06,
+      "loss": 1.2588,
+      "step": 7203
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5274919247649845e-06,
+      "loss": 1.2109,
+      "step": 7204
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.524912526805106e-06,
+      "loss": 1.2002,
+      "step": 7205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.522333304254549e-06,
+      "loss": 1.2402,
+      "step": 7206
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519754257296112e-06,
+      "loss": 1.2461,
+      "step": 7207
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51717538611256e-06,
+      "loss": 1.1982,
+      "step": 7208
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514596690886661e-06,
+      "loss": 1.2021,
+      "step": 7209
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51201817180117e-06,
+      "loss": 1.2119,
+      "step": 7210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50943982903883e-06,
+      "loss": 1.2451,
+      "step": 7211
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.506861662782359e-06,
+      "loss": 1.1914,
+      "step": 7212
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504283673214472e-06,
+      "loss": 1.2871,
+      "step": 7213
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.501705860517879e-06,
+      "loss": 1.2344,
+      "step": 7214
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499128224875267e-06,
+      "loss": 1.2354,
+      "step": 7215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496550766469305e-06,
+      "loss": 1.2432,
+      "step": 7216
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4939734854826665e-06,
+      "loss": 1.2852,
+      "step": 7217
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491396382098e-06,
+      "loss": 1.2129,
+      "step": 7218
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488819456497944e-06,
+      "loss": 1.2383,
+      "step": 7219
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486242708865127e-06,
+      "loss": 1.2354,
+      "step": 7220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483666139382161e-06,
+      "loss": 1.21,
+      "step": 7221
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481089748231647e-06,
+      "loss": 1.2832,
+      "step": 7222
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.478513535596178e-06,
+      "loss": 1.1816,
+      "step": 7223
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4759375016583244e-06,
+      "loss": 1.2197,
+      "step": 7224
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473361646600652e-06,
+      "loss": 1.1709,
+      "step": 7225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.470785970605712e-06,
+      "loss": 1.249,
+      "step": 7226
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468210473856045e-06,
+      "loss": 1.25,
+      "step": 7227
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.46563515653417e-06,
+      "loss": 1.2373,
+      "step": 7228
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463060018822603e-06,
+      "loss": 1.2148,
+      "step": 7229
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460485060903844e-06,
+      "loss": 1.2188,
+      "step": 7230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457910282960383e-06,
+      "loss": 1.1973,
+      "step": 7231
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.455335685174687e-06,
+      "loss": 1.2354,
+      "step": 7232
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452761267729225e-06,
+      "loss": 1.2549,
+      "step": 7233
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450187030806441e-06,
+      "loss": 1.2246,
+      "step": 7234
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447612974588775e-06,
+      "loss": 1.2207,
+      "step": 7235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.445039099258646e-06,
+      "loss": 1.1875,
+      "step": 7236
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442465404998467e-06,
+      "loss": 1.2461,
+      "step": 7237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439891891990633e-06,
+      "loss": 1.2373,
+      "step": 7238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437318560417537e-06,
+      "loss": 1.2637,
+      "step": 7239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.43474541046154e-06,
+      "loss": 1.2188,
+      "step": 7240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.432172442305006e-06,
+      "loss": 1.2588,
+      "step": 7241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42959965613028e-06,
+      "loss": 1.1777,
+      "step": 7242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4270270521197e-06,
+      "loss": 1.2637,
+      "step": 7243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42445463045558e-06,
+      "loss": 1.2373,
+      "step": 7244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421882391320226e-06,
+      "loss": 1.2344,
+      "step": 7245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41931033489594e-06,
+      "loss": 1.2773,
+      "step": 7246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416738461365003e-06,
+      "loss": 1.2549,
+      "step": 7247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414166770909674e-06,
+      "loss": 1.2441,
+      "step": 7248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411595263712218e-06,
+      "loss": 1.1758,
+      "step": 7249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409023939954873e-06,
+      "loss": 1.2793,
+      "step": 7250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406452799819875e-06,
+      "loss": 1.2266,
+      "step": 7251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.40388184348943e-06,
+      "loss": 1.2461,
+      "step": 7252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401311071145748e-06,
+      "loss": 1.2021,
+      "step": 7253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398740482971018e-06,
+      "loss": 1.1982,
+      "step": 7254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396170079147422e-06,
+      "loss": 1.1973,
+      "step": 7255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393599859857119e-06,
+      "loss": 1.2207,
+      "step": 7256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3910298252822595e-06,
+      "loss": 1.2148,
+      "step": 7257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.388459975604985e-06,
+      "loss": 1.2344,
+      "step": 7258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.385890311007422e-06,
+      "loss": 1.1797,
+      "step": 7259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383320831671678e-06,
+      "loss": 1.2236,
+      "step": 7260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3807515377798535e-06,
+      "loss": 1.249,
+      "step": 7261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378182429514036e-06,
+      "loss": 1.251,
+      "step": 7262
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.375613507056295e-06,
+      "loss": 1.2236,
+      "step": 7263
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3730447705886955e-06,
+      "loss": 1.2891,
+      "step": 7264
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370476220293276e-06,
+      "loss": 1.2197,
+      "step": 7265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367907856352074e-06,
+      "loss": 1.2656,
+      "step": 7266
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.36533967894711e-06,
+      "loss": 1.2705,
+      "step": 7267
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3627716882603905e-06,
+      "loss": 1.1992,
+      "step": 7268
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.360203884473905e-06,
+      "loss": 1.2021,
+      "step": 7269
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.357636267769638e-06,
+      "loss": 1.2246,
+      "step": 7270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355068838329551e-06,
+      "loss": 1.2461,
+      "step": 7271
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3525015963356095e-06,
+      "loss": 1.21,
+      "step": 7272
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.349934541969741e-06,
+      "loss": 1.2275,
+      "step": 7273
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347367675413875e-06,
+      "loss": 1.2412,
+      "step": 7274
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3448009968499274e-06,
+      "loss": 1.2236,
+      "step": 7275
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3422345064598044e-06,
+      "loss": 1.1924,
+      "step": 7276
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.339668204425383e-06,
+      "loss": 1.2285,
+      "step": 7277
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.337102090928538e-06,
+      "loss": 1.29,
+      "step": 7278
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.334536166151137e-06,
+      "loss": 1.2051,
+      "step": 7279
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3319704302750265e-06,
+      "loss": 1.251,
+      "step": 7280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329404883482028e-06,
+      "loss": 1.2451,
+      "step": 7281
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326839525953975e-06,
+      "loss": 1.2432,
+      "step": 7282
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.324274357872667e-06,
+      "loss": 1.2236,
+      "step": 7283
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321709379419902e-06,
+      "loss": 1.2471,
+      "step": 7284
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.319144590777456e-06,
+      "loss": 1.2275,
+      "step": 7285
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.316579992127097e-06,
+      "loss": 1.2275,
+      "step": 7286
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314015583650578e-06,
+      "loss": 1.2646,
+      "step": 7287
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31145136552964e-06,
+      "loss": 1.2344,
+      "step": 7288
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308887337946006e-06,
+      "loss": 1.2461,
+      "step": 7289
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.306323501081389e-06,
+      "loss": 1.2305,
+      "step": 7290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303759855117489e-06,
+      "loss": 1.2002,
+      "step": 7291
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3011964002359945e-06,
+      "loss": 1.2168,
+      "step": 7292
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298633136618571e-06,
+      "loss": 1.2256,
+      "step": 7293
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29607006444688e-06,
+      "loss": 1.2354,
+      "step": 7294
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293507183902569e-06,
+      "loss": 1.2012,
+      "step": 7295
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290944495167267e-06,
+      "loss": 1.2236,
+      "step": 7296
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288381998422589e-06,
+      "loss": 1.2305,
+      "step": 7297
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2858196938501415e-06,
+      "loss": 1.2207,
+      "step": 7298
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283257581631514e-06,
+      "loss": 1.2461,
+      "step": 7299
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280695661948287e-06,
+      "loss": 1.2471,
+      "step": 7300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278133934982019e-06,
+      "loss": 1.2383,
+      "step": 7301
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2755724009142595e-06,
+      "loss": 1.2295,
+      "step": 7302
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2730110599265465e-06,
+      "loss": 1.2402,
+      "step": 7303
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2704499122004034e-06,
+      "loss": 1.1924,
+      "step": 7304
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267888957917335e-06,
+      "loss": 1.2207,
+      "step": 7305
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.265328197258838e-06,
+      "loss": 1.2041,
+      "step": 7306
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.262767630406391e-06,
+      "loss": 1.2568,
+      "step": 7307
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260207257541468e-06,
+      "loss": 1.2637,
+      "step": 7308
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257647078845515e-06,
+      "loss": 1.2266,
+      "step": 7309
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255087094499971e-06,
+      "loss": 1.1875,
+      "step": 7310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.252527304686269e-06,
+      "loss": 1.2471,
+      "step": 7311
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249967709585821e-06,
+      "loss": 1.2051,
+      "step": 7312
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.247408309380019e-06,
+      "loss": 1.2109,
+      "step": 7313
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2448491042502465e-06,
+      "loss": 1.2793,
+      "step": 7314
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242290094377882e-06,
+      "loss": 1.2188,
+      "step": 7315
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2397312799442825e-06,
+      "loss": 1.248,
+      "step": 7316
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.23717266113078e-06,
+      "loss": 1.2324,
+      "step": 7317
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2346142381187146e-06,
+      "loss": 1.1846,
+      "step": 7318
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.232056011089397e-06,
+      "loss": 1.2256,
+      "step": 7319
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2294979802241316e-06,
+      "loss": 1.1904,
+      "step": 7320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2269401457042025e-06,
+      "loss": 1.1787,
+      "step": 7321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224382507710884e-06,
+      "loss": 1.2256,
+      "step": 7322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221825066425436e-06,
+      "loss": 1.2275,
+      "step": 7323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.219267822029108e-06,
+      "loss": 1.2471,
+      "step": 7324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2167107747031265e-06,
+      "loss": 1.2227,
+      "step": 7325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.214153924628711e-06,
+      "loss": 1.2363,
+      "step": 7326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.211597271987065e-06,
+      "loss": 1.2178,
+      "step": 7327
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.209040816959381e-06,
+      "loss": 1.2598,
+      "step": 7328
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206484559726832e-06,
+      "loss": 1.2783,
+      "step": 7329
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2039285004705795e-06,
+      "loss": 1.2734,
+      "step": 7330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2013726393717735e-06,
+      "loss": 1.2109,
+      "step": 7331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.198816976611547e-06,
+      "loss": 1.2383,
+      "step": 7332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196261512371019e-06,
+      "loss": 1.1846,
+      "step": 7333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.193706246831294e-06,
+      "loss": 1.2158,
+      "step": 7334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191151180173466e-06,
+      "loss": 1.165,
+      "step": 7335
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.188596312578613e-06,
+      "loss": 1.2412,
+      "step": 7336
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1860416442277944e-06,
+      "loss": 1.2656,
+      "step": 7337
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183487175302063e-06,
+      "loss": 1.2588,
+      "step": 7338
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180932905982452e-06,
+      "loss": 1.2432,
+      "step": 7339
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.178378836449985e-06,
+      "loss": 1.2354,
+      "step": 7340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.175824966885665e-06,
+      "loss": 1.2607,
+      "step": 7341
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173271297470487e-06,
+      "loss": 1.2217,
+      "step": 7342
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170717828385425e-06,
+      "loss": 1.2266,
+      "step": 7343
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168164559811454e-06,
+      "loss": 1.2969,
+      "step": 7344
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1656114919295135e-06,
+      "loss": 1.2627,
+      "step": 7345
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163058624920542e-06,
+      "loss": 1.2451,
+      "step": 7346
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160505958965463e-06,
+      "loss": 1.251,
+      "step": 7347
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157953494245186e-06,
+      "loss": 1.2549,
+      "step": 7348
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155401230940598e-06,
+      "loss": 1.1914,
+      "step": 7349
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152849169232576e-06,
+      "loss": 1.2324,
+      "step": 7350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150297309301994e-06,
+      "loss": 1.2285,
+      "step": 7351
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147745651329701e-06,
+      "loss": 1.2275,
+      "step": 7352
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1451941954965206e-06,
+      "loss": 1.2568,
+      "step": 7353
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1426429419832865e-06,
+      "loss": 1.1689,
+      "step": 7354
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.140091890970802e-06,
+      "loss": 1.2402,
+      "step": 7355
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1375410426398625e-06,
+      "loss": 1.2207,
+      "step": 7356
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134990397171242e-06,
+      "loss": 1.2178,
+      "step": 7357
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132439954745707e-06,
+      "loss": 1.2109,
+      "step": 7358
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129889715544007e-06,
+      "loss": 1.2822,
+      "step": 7359
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127339679746879e-06,
+      "loss": 1.2861,
+      "step": 7360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124789847535041e-06,
+      "loss": 1.21,
+      "step": 7361
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122240219089201e-06,
+      "loss": 1.2344,
+      "step": 7362
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.119690794590048e-06,
+      "loss": 1.2031,
+      "step": 7363
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.117141574218267e-06,
+      "loss": 1.2334,
+      "step": 7364
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114592558154514e-06,
+      "loss": 1.1855,
+      "step": 7365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11204374657944e-06,
+      "loss": 1.2412,
+      "step": 7366
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.109495139673679e-06,
+      "loss": 1.2393,
+      "step": 7367
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106946737617854e-06,
+      "loss": 1.208,
+      "step": 7368
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.104398540592563e-06,
+      "loss": 1.2412,
+      "step": 7369
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101850548778402e-06,
+      "loss": 1.2109,
+      "step": 7370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0993027623559465e-06,
+      "loss": 1.2012,
+      "step": 7371
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096755181505758e-06,
+      "loss": 1.2061,
+      "step": 7372
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0942078064083795e-06,
+      "loss": 1.2246,
+      "step": 7373
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091660637244348e-06,
+      "loss": 1.2354,
+      "step": 7374
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089113674194178e-06,
+      "loss": 1.1484,
+      "step": 7375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0865669174383796e-06,
+      "loss": 1.2012,
+      "step": 7376
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.084020367157432e-06,
+      "loss": 1.251,
+      "step": 7377
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081474023531813e-06,
+      "loss": 1.252,
+      "step": 7378
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078927886741982e-06,
+      "loss": 1.2383,
+      "step": 7379
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076381956968388e-06,
+      "loss": 1.2529,
+      "step": 7380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073836234391451e-06,
+      "loss": 1.2344,
+      "step": 7381
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.071290719191592e-06,
+      "loss": 1.2109,
+      "step": 7382
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068745411549213e-06,
+      "loss": 1.2412,
+      "step": 7383
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066200311644702e-06,
+      "loss": 1.2471,
+      "step": 7384
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0636554196584216e-06,
+      "loss": 1.2158,
+      "step": 7385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061110735770731e-06,
+      "loss": 1.2432,
+      "step": 7386
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058566260161974e-06,
+      "loss": 1.2334,
+      "step": 7387
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056021993012483e-06,
+      "loss": 1.2539,
+      "step": 7388
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0534779345025575e-06,
+      "loss": 1.1719,
+      "step": 7389
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050934084812504e-06,
+      "loss": 1.2314,
+      "step": 7390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048390444122601e-06,
+      "loss": 1.2783,
+      "step": 7391
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045847012613121e-06,
+      "loss": 1.2344,
+      "step": 7392
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043303790464311e-06,
+      "loss": 1.2295,
+      "step": 7393
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.040760777856411e-06,
+      "loss": 1.2393,
+      "step": 7394
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.038217974969645e-06,
+      "loss": 1.2227,
+      "step": 7395
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0356753819842216e-06,
+      "loss": 1.2393,
+      "step": 7396
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033132999080335e-06,
+      "loss": 1.2334,
+      "step": 7397
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.03059082643816e-06,
+      "loss": 1.2363,
+      "step": 7398
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028048864237863e-06,
+      "loss": 1.2373,
+      "step": 7399
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025507112659593e-06,
+      "loss": 1.2061,
+      "step": 7400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022965571883484e-06,
+      "loss": 1.2451,
+      "step": 7401
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020424242089652e-06,
+      "loss": 1.1973,
+      "step": 7402
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017883123458204e-06,
+      "loss": 1.2236,
+      "step": 7403
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015342216169227e-06,
+      "loss": 1.2227,
+      "step": 7404
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0128015204027985e-06,
+      "loss": 1.1895,
+      "step": 7405
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0102610363389726e-06,
+      "loss": 1.2529,
+      "step": 7406
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.007720764157795e-06,
+      "loss": 1.208,
+      "step": 7407
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005180704039293e-06,
+      "loss": 1.2461,
+      "step": 7408
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.00264085616349e-06,
+      "loss": 1.2266,
+      "step": 7409
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000101220710372e-06,
+      "loss": 1.1963,
+      "step": 7410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9975617978599264e-06,
+      "loss": 1.1797,
+      "step": 7411
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.995022587792126e-06,
+      "loss": 1.2451,
+      "step": 7412
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992483590686928e-06,
+      "loss": 1.2354,
+      "step": 7413
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9899448067242605e-06,
+      "loss": 1.2783,
+      "step": 7414
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.98740623608405e-06,
+      "loss": 1.1836,
+      "step": 7415
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984867878946209e-06,
+      "loss": 1.2158,
+      "step": 7416
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.982329735490632e-06,
+      "loss": 1.2041,
+      "step": 7417
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979791805897188e-06,
+      "loss": 1.2451,
+      "step": 7418
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977254090345747e-06,
+      "loss": 1.2764,
+      "step": 7419
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.974716589016157e-06,
+      "loss": 1.2646,
+      "step": 7420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97217930208825e-06,
+      "loss": 1.2275,
+      "step": 7421
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.96964222974184e-06,
+      "loss": 1.1982,
+      "step": 7422
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967105372156733e-06,
+      "loss": 1.2178,
+      "step": 7423
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964568729512713e-06,
+      "loss": 1.2158,
+      "step": 7424
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962032301989555e-06,
+      "loss": 1.2314,
+      "step": 7425
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.959496089767013e-06,
+      "loss": 1.25,
+      "step": 7426
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956960093024827e-06,
+      "loss": 1.2529,
+      "step": 7427
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954424311942727e-06,
+      "loss": 1.2109,
+      "step": 7428
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951888746700423e-06,
+      "loss": 1.1953,
+      "step": 7429
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949353397477607e-06,
+      "loss": 1.2227,
+      "step": 7430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.946818264453961e-06,
+      "loss": 1.1914,
+      "step": 7431
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94428334780915e-06,
+      "loss": 1.2607,
+      "step": 7432
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941748647722826e-06,
+      "loss": 1.2568,
+      "step": 7433
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939214164374617e-06,
+      "loss": 1.2422,
+      "step": 7434
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9366798979441455e-06,
+      "loss": 1.2783,
+      "step": 7435
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934145848611014e-06,
+      "loss": 1.2607,
+      "step": 7436
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931612016554812e-06,
+      "loss": 1.2393,
+      "step": 7437
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92907840195511e-06,
+      "loss": 1.2236,
+      "step": 7438
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926545004991466e-06,
+      "loss": 1.2344,
+      "step": 7439
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.924011825843422e-06,
+      "loss": 1.2383,
+      "step": 7440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921478864690505e-06,
+      "loss": 1.209,
+      "step": 7441
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.918946121712223e-06,
+      "loss": 1.2393,
+      "step": 7442
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916413597088074e-06,
+      "loss": 1.2783,
+      "step": 7443
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913881290997534e-06,
+      "loss": 1.1934,
+      "step": 7444
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9113492036200765e-06,
+      "loss": 1.2676,
+      "step": 7445
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90881733513514e-06,
+      "loss": 1.2119,
+      "step": 7446
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.906285685722162e-06,
+      "loss": 1.2363,
+      "step": 7447
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903754255560562e-06,
+      "loss": 1.2383,
+      "step": 7448
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9012230448297436e-06,
+      "loss": 1.2354,
+      "step": 7449
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898692053709088e-06,
+      "loss": 1.2432,
+      "step": 7450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896161282377967e-06,
+      "loss": 1.252,
+      "step": 7451
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893630731015741e-06,
+      "loss": 1.2256,
+      "step": 7452
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.891100399801751e-06,
+      "loss": 1.2158,
+      "step": 7453
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8885702889153125e-06,
+      "loss": 1.251,
+      "step": 7454
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8860403985357424e-06,
+      "loss": 1.208,
+      "step": 7455
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883510728842331e-06,
+      "loss": 1.3154,
+      "step": 7456
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.880981280014359e-06,
+      "loss": 1.2891,
+      "step": 7457
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878452052231081e-06,
+      "loss": 1.2354,
+      "step": 7458
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875923045671751e-06,
+      "loss": 1.2324,
+      "step": 7459
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873394260515593e-06,
+      "loss": 1.2549,
+      "step": 7460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.870865696941829e-06,
+      "loss": 1.1826,
+      "step": 7461
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868337355129653e-06,
+      "loss": 1.2598,
+      "step": 7462
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8658092352582495e-06,
+      "loss": 1.1787,
+      "step": 7463
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863281337506786e-06,
+      "loss": 1.209,
+      "step": 7464
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860753662054418e-06,
+      "loss": 1.2227,
+      "step": 7465
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858226209080276e-06,
+      "loss": 1.1689,
+      "step": 7466
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.855698978763484e-06,
+      "loss": 1.2129,
+      "step": 7467
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853171971283146e-06,
+      "loss": 1.2148,
+      "step": 7468
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850645186818352e-06,
+      "loss": 1.2002,
+      "step": 7469
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8481186255481724e-06,
+      "loss": 1.2334,
+      "step": 7470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845592287651666e-06,
+      "loss": 1.2578,
+      "step": 7471
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843066173307875e-06,
+      "loss": 1.1982,
+      "step": 7472
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840540282695826e-06,
+      "loss": 1.248,
+      "step": 7473
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.838014615994524e-06,
+      "loss": 1.1914,
+      "step": 7474
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835489173382968e-06,
+      "loss": 1.209,
+      "step": 7475
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83296395504013e-06,
+      "loss": 1.2012,
+      "step": 7476
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830438961144981e-06,
+      "loss": 1.2441,
+      "step": 7477
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8279141918764595e-06,
+      "loss": 1.2637,
+      "step": 7478
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.825389647413499e-06,
+      "loss": 1.1973,
+      "step": 7479
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82286532793501e-06,
+      "loss": 1.2168,
+      "step": 7480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820341233619901e-06,
+      "loss": 1.2227,
+      "step": 7481
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8178173646470435e-06,
+      "loss": 1.2578,
+      "step": 7482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815293721195305e-06,
+      "loss": 1.2578,
+      "step": 7483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812770303443542e-06,
+      "loss": 1.2324,
+      "step": 7484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.81024711157059e-06,
+      "loss": 1.2324,
+      "step": 7485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.807724145755259e-06,
+      "loss": 1.2021,
+      "step": 7486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.805201406176355e-06,
+      "loss": 1.2373,
+      "step": 7487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802678893012666e-06,
+      "loss": 1.2168,
+      "step": 7488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800156606442967e-06,
+      "loss": 1.2217,
+      "step": 7489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.797634546646001e-06,
+      "loss": 1.2246,
+      "step": 7490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795112713800513e-06,
+      "loss": 1.2646,
+      "step": 7491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792591108085225e-06,
+      "loss": 1.2354,
+      "step": 7492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7900697296788454e-06,
+      "loss": 1.2568,
+      "step": 7493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.787548578760059e-06,
+      "loss": 1.1953,
+      "step": 7494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7850276555075425e-06,
+      "loss": 1.2217,
+      "step": 7495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782506960099953e-06,
+      "loss": 1.2646,
+      "step": 7496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7799864927159334e-06,
+      "loss": 1.2715,
+      "step": 7497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777466253534105e-06,
+      "loss": 1.2207,
+      "step": 7498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774946242733081e-06,
+      "loss": 1.2588,
+      "step": 7499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772426460491452e-06,
+      "loss": 1.251,
+      "step": 7500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7699069069878e-06,
+      "loss": 1.2363,
+      "step": 7501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7673875824006795e-06,
+      "loss": 1.2178,
+      "step": 7502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.764868486908636e-06,
+      "loss": 1.2129,
+      "step": 7503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762349620690201e-06,
+      "loss": 1.1748,
+      "step": 7504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759830983923883e-06,
+      "loss": 1.2139,
+      "step": 7505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75731257678818e-06,
+      "loss": 1.249,
+      "step": 7506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754794399461568e-06,
+      "loss": 1.2539,
+      "step": 7507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7522764521225126e-06,
+      "loss": 1.209,
+      "step": 7508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.749758734949463e-06,
+      "loss": 1.1924,
+      "step": 7509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.747241248120846e-06,
+      "loss": 1.2178,
+      "step": 7510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744723991815076e-06,
+      "loss": 1.2168,
+      "step": 7511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742206966210552e-06,
+      "loss": 1.2354,
+      "step": 7512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7396901714856566e-06,
+      "loss": 1.2021,
+      "step": 7513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.737173607818752e-06,
+      "loss": 1.1992,
+      "step": 7514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734657275388188e-06,
+      "loss": 1.2627,
+      "step": 7515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7321411743722955e-06,
+      "loss": 1.2354,
+      "step": 7516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729625304949398e-06,
+      "loss": 1.2139,
+      "step": 7517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727109667297785e-06,
+      "loss": 1.2617,
+      "step": 7518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724594261595743e-06,
+      "loss": 1.2295,
+      "step": 7519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72207908802154e-06,
+      "loss": 1.2549,
+      "step": 7520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719564146753433e-06,
+      "loss": 1.2393,
+      "step": 7521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717049437969642e-06,
+      "loss": 1.2246,
+      "step": 7522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.714534961848389e-06,
+      "loss": 1.2021,
+      "step": 7523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71202071856788e-06,
+      "loss": 1.251,
+      "step": 7524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709506708306299e-06,
+      "loss": 1.2754,
+      "step": 7525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.706992931241804e-06,
+      "loss": 1.2432,
+      "step": 7526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.704479387552556e-06,
+      "loss": 1.2383,
+      "step": 7527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701966077416687e-06,
+      "loss": 1.2197,
+      "step": 7528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699453001012315e-06,
+      "loss": 1.2188,
+      "step": 7529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6969401585175435e-06,
+      "loss": 1.2451,
+      "step": 7530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694427550110453e-06,
+      "loss": 1.3076,
+      "step": 7531
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691915175969115e-06,
+      "loss": 1.2393,
+      "step": 7532
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68940303627158e-06,
+      "loss": 1.2031,
+      "step": 7533
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686891131195886e-06,
+      "loss": 1.2256,
+      "step": 7534
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.684379460920048e-06,
+      "loss": 1.2705,
+      "step": 7535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681868025622069e-06,
+      "loss": 1.2334,
+      "step": 7536
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679356825479936e-06,
+      "loss": 1.1934,
+      "step": 7537
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676845860671617e-06,
+      "loss": 1.2754,
+      "step": 7538
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.674335131375062e-06,
+      "loss": 1.2461,
+      "step": 7539
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.671824637768209e-06,
+      "loss": 1.2354,
+      "step": 7540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669314380028973e-06,
+      "loss": 1.2246,
+      "step": 7541
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.666804358335262e-06,
+      "loss": 1.2529,
+      "step": 7542
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664294572864954e-06,
+      "loss": 1.208,
+      "step": 7543
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66178502379592e-06,
+      "loss": 1.2744,
+      "step": 7544
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6592757113060106e-06,
+      "loss": 1.2305,
+      "step": 7545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656766635573068e-06,
+      "loss": 1.2617,
+      "step": 7546
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.654257796774901e-06,
+      "loss": 1.1768,
+      "step": 7547
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.651749195089312e-06,
+      "loss": 1.252,
+      "step": 7548
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64924083069409e-06,
+      "loss": 1.2393,
+      "step": 7549
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.646732703767005e-06,
+      "loss": 1.2441,
+      "step": 7550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644224814485798e-06,
+      "loss": 1.2354,
+      "step": 7551
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.641717163028207e-06,
+      "loss": 1.2148,
+      "step": 7552
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6392097495719535e-06,
+      "loss": 1.291,
+      "step": 7553
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636702574294737e-06,
+      "loss": 1.208,
+      "step": 7554
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.634195637374232e-06,
+      "loss": 1.2129,
+      "step": 7555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631688938988115e-06,
+      "loss": 1.2334,
+      "step": 7556
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629182479314031e-06,
+      "loss": 1.2373,
+      "step": 7557
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62667625852962e-06,
+      "loss": 1.2119,
+      "step": 7558
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6241702768124835e-06,
+      "loss": 1.2168,
+      "step": 7559
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621664534340231e-06,
+      "loss": 1.2324,
+      "step": 7560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619159031290442e-06,
+      "loss": 1.2266,
+      "step": 7561
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616653767840682e-06,
+      "loss": 1.2715,
+      "step": 7562
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6141487441684974e-06,
+      "loss": 1.2266,
+      "step": 7563
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.611643960451419e-06,
+      "loss": 1.2695,
+      "step": 7564
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.609139416866961e-06,
+      "loss": 1.2285,
+      "step": 7565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606635113592623e-06,
+      "loss": 1.2178,
+      "step": 7566
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60413105080588e-06,
+      "loss": 1.2432,
+      "step": 7567
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.601627228684197e-06,
+      "loss": 1.2041,
+      "step": 7568
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5991236474050215e-06,
+      "loss": 1.2324,
+      "step": 7569
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596620307145781e-06,
+      "loss": 1.2168,
+      "step": 7570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.594117208083886e-06,
+      "loss": 1.2168,
+      "step": 7571
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591614350396731e-06,
+      "loss": 1.2334,
+      "step": 7572
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589111734261695e-06,
+      "loss": 1.2432,
+      "step": 7573
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58660935985614e-06,
+      "loss": 1.1953,
+      "step": 7574
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584107227357403e-06,
+      "loss": 1.1377,
+      "step": 7575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581605336942813e-06,
+      "loss": 1.2266,
+      "step": 7576
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579103688789682e-06,
+      "loss": 1.1963,
+      "step": 7577
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576602283075299e-06,
+      "loss": 1.1914,
+      "step": 7578
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574101119976939e-06,
+      "loss": 1.2617,
+      "step": 7579
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571600199671857e-06,
+      "loss": 1.2441,
+      "step": 7580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5690995223372945e-06,
+      "loss": 1.2158,
+      "step": 7581
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56659908815048e-06,
+      "loss": 1.2168,
+      "step": 7582
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564098897288609e-06,
+      "loss": 1.2314,
+      "step": 7583
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5615989499288744e-06,
+      "loss": 1.2393,
+      "step": 7584
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.55909924624845e-06,
+      "loss": 1.2256,
+      "step": 7585
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556599786424491e-06,
+      "loss": 1.2666,
+      "step": 7586
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.554100570634127e-06,
+      "loss": 1.2373,
+      "step": 7587
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551601599054478e-06,
+      "loss": 1.2637,
+      "step": 7588
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549102871862652e-06,
+      "loss": 1.2568,
+      "step": 7589
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.546604389235733e-06,
+      "loss": 1.2031,
+      "step": 7590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544106151350781e-06,
+      "loss": 1.252,
+      "step": 7591
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.541608158384853e-06,
+      "loss": 1.2588,
+      "step": 7592
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.539110410514979e-06,
+      "loss": 1.2275,
+      "step": 7593
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536612907918176e-06,
+      "loss": 1.2051,
+      "step": 7594
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53411565077144e-06,
+      "loss": 1.1982,
+      "step": 7595
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531618639251752e-06,
+      "loss": 1.1963,
+      "step": 7596
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529121873536075e-06,
+      "loss": 1.1943,
+      "step": 7597
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526625353801357e-06,
+      "loss": 1.2646,
+      "step": 7598
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524129080224522e-06,
+      "loss": 1.2188,
+      "step": 7599
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521633052982482e-06,
+      "loss": 1.1992,
+      "step": 7600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519137272252132e-06,
+      "loss": 1.168,
+      "step": 7601
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.51664173821035e-06,
+      "loss": 1.2393,
+      "step": 7602
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514146451033987e-06,
+      "loss": 1.2744,
+      "step": 7603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511651410899891e-06,
+      "loss": 1.2188,
+      "step": 7604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50915661798488e-06,
+      "loss": 1.21,
+      "step": 7605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.506662072465766e-06,
+      "loss": 1.2148,
+      "step": 7606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5041677745193325e-06,
+      "loss": 1.2285,
+      "step": 7607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50167372432235e-06,
+      "loss": 1.2373,
+      "step": 7608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499179922051575e-06,
+      "loss": 1.2217,
+      "step": 7609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.496686367883742e-06,
+      "loss": 1.2568,
+      "step": 7610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494193061995566e-06,
+      "loss": 1.1777,
+      "step": 7611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4917000045637526e-06,
+      "loss": 1.21,
+      "step": 7612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.48920719576498e-06,
+      "loss": 1.2051,
+      "step": 7613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486714635775917e-06,
+      "loss": 1.2002,
+      "step": 7614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.484222324773209e-06,
+      "loss": 1.2178,
+      "step": 7615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.481730262933485e-06,
+      "loss": 1.2383,
+      "step": 7616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479238450433358e-06,
+      "loss": 1.2334,
+      "step": 7617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476746887449431e-06,
+      "loss": 1.2686,
+      "step": 7618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.474255574158268e-06,
+      "loss": 1.207,
+      "step": 7619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4717645107364314e-06,
+      "loss": 1.2246,
+      "step": 7620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469273697360467e-06,
+      "loss": 1.2119,
+      "step": 7621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466783134206902e-06,
+      "loss": 1.2363,
+      "step": 7622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464292821452232e-06,
+      "loss": 1.2207,
+      "step": 7623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461802759272948e-06,
+      "loss": 1.2441,
+      "step": 7624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.459312947845527e-06,
+      "loss": 1.2412,
+      "step": 7625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45682338734642e-06,
+      "loss": 1.2139,
+      "step": 7626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454334077952055e-06,
+      "loss": 1.2627,
+      "step": 7627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451845019838856e-06,
+      "loss": 1.1953,
+      "step": 7628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44935621318322e-06,
+      "loss": 1.2256,
+      "step": 7629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446867658161531e-06,
+      "loss": 1.2256,
+      "step": 7630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44437935495015e-06,
+      "loss": 1.249,
+      "step": 7631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441891303725422e-06,
+      "loss": 1.208,
+      "step": 7632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43940350466368e-06,
+      "loss": 1.2686,
+      "step": 7633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436915957941231e-06,
+      "loss": 1.2095,
+      "step": 7634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434428663734366e-06,
+      "loss": 1.1953,
+      "step": 7635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43194162221936e-06,
+      "loss": 1.209,
+      "step": 7636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429454833572471e-06,
+      "loss": 1.2432,
+      "step": 7637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42696829796994e-06,
+      "loss": 1.1855,
+      "step": 7638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42448201558798e-06,
+      "loss": 1.2559,
+      "step": 7639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4219959866028005e-06,
+      "loss": 1.2393,
+      "step": 7640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.419510211190582e-06,
+      "loss": 1.2197,
+      "step": 7641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.417024689527498e-06,
+      "loss": 1.249,
+      "step": 7642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4145394217896896e-06,
+      "loss": 1.2012,
+      "step": 7643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41205440815329e-06,
+      "loss": 1.25,
+      "step": 7644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409569648794412e-06,
+      "loss": 1.2461,
+      "step": 7645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.407085143889154e-06,
+      "loss": 1.2529,
+      "step": 7646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404600893613588e-06,
+      "loss": 1.2168,
+      "step": 7647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402116898143774e-06,
+      "loss": 1.2305,
+      "step": 7648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399633157655754e-06,
+      "loss": 1.208,
+      "step": 7649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3971496723255485e-06,
+      "loss": 1.2178,
+      "step": 7650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394666442329165e-06,
+      "loss": 1.209,
+      "step": 7651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392183467842585e-06,
+      "loss": 1.2441,
+      "step": 7652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389700749041778e-06,
+      "loss": 1.2627,
+      "step": 7653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387218286102703e-06,
+      "loss": 1.2607,
+      "step": 7654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3847360792012814e-06,
+      "loss": 1.2109,
+      "step": 7655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382254128513428e-06,
+      "loss": 1.2383,
+      "step": 7656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379772434215042e-06,
+      "loss": 1.25,
+      "step": 7657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377290996482001e-06,
+      "loss": 1.1885,
+      "step": 7658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374809815490169e-06,
+      "loss": 1.2129,
+      "step": 7659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372328891415374e-06,
+      "loss": 1.2021,
+      "step": 7660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.369848224433448e-06,
+      "loss": 1.2412,
+      "step": 7661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3673678147201946e-06,
+      "loss": 1.2246,
+      "step": 7662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364887662451402e-06,
+      "loss": 1.1953,
+      "step": 7663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.362407767802835e-06,
+      "loss": 1.2051,
+      "step": 7664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.359928130950245e-06,
+      "loss": 1.1895,
+      "step": 7665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357448752069364e-06,
+      "loss": 1.249,
+      "step": 7666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354969631335907e-06,
+      "loss": 1.1582,
+      "step": 7667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352490768925566e-06,
+      "loss": 1.1699,
+      "step": 7668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.350012165014017e-06,
+      "loss": 1.1914,
+      "step": 7669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347533819776923e-06,
+      "loss": 1.2012,
+      "step": 7670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345055733389925e-06,
+      "loss": 1.2236,
+      "step": 7671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3425779060286375e-06,
+      "loss": 1.2217,
+      "step": 7672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340100337868669e-06,
+      "loss": 1.1914,
+      "step": 7673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337623029085604e-06,
+      "loss": 1.2266,
+      "step": 7674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335145979855011e-06,
+      "loss": 1.248,
+      "step": 7675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.332669190352436e-06,
+      "loss": 1.2295,
+      "step": 7676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3301926607534096e-06,
+      "loss": 1.2188,
+      "step": 7677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327716391233441e-06,
+      "loss": 1.208,
+      "step": 7678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32524038196803e-06,
+      "loss": 1.2227,
+      "step": 7679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3227646331326434e-06,
+      "loss": 1.2236,
+      "step": 7680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3202891449027405e-06,
+      "loss": 1.2451,
+      "step": 7681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317813917453759e-06,
+      "loss": 1.2959,
+      "step": 7682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315338950961123e-06,
+      "loss": 1.2256,
+      "step": 7683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312864245600225e-06,
+      "loss": 1.209,
+      "step": 7684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310389801546453e-06,
+      "loss": 1.1729,
+      "step": 7685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307915618975164e-06,
+      "loss": 1.248,
+      "step": 7686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305441698061715e-06,
+      "loss": 1.2324,
+      "step": 7687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3029680389814214e-06,
+      "loss": 1.2148,
+      "step": 7688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3004946419095935e-06,
+      "loss": 1.2559,
+      "step": 7689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2980215070215255e-06,
+      "loss": 1.2207,
+      "step": 7690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295548634492489e-06,
+      "loss": 1.2393,
+      "step": 7691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293076024497726e-06,
+      "loss": 1.1885,
+      "step": 7692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290603677212481e-06,
+      "loss": 1.2246,
+      "step": 7693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.288131592811964e-06,
+      "loss": 1.2051,
+      "step": 7694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.285659771471378e-06,
+      "loss": 1.2227,
+      "step": 7695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283188213365889e-06,
+      "loss": 1.1953,
+      "step": 7696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.280716918670665e-06,
+      "loss": 1.1943,
+      "step": 7697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278245887560845e-06,
+      "loss": 1.2285,
+      "step": 7698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275775120211551e-06,
+      "loss": 1.25,
+      "step": 7699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273304616797884e-06,
+      "loss": 1.2676,
+      "step": 7700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27083437749493e-06,
+      "loss": 1.252,
+      "step": 7701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.268364402477756e-06,
+      "loss": 1.2451,
+      "step": 7702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.265894691921407e-06,
+      "loss": 1.2373,
+      "step": 7703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.263425246000912e-06,
+      "loss": 1.2451,
+      "step": 7704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260956064891279e-06,
+      "loss": 1.2178,
+      "step": 7705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.258487148767501e-06,
+      "loss": 1.2246,
+      "step": 7706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25601849780455e-06,
+      "loss": 1.2354,
+      "step": 7707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253550112177378e-06,
+      "loss": 1.1719,
+      "step": 7708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251081992060918e-06,
+      "loss": 1.2373,
+      "step": 7709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248614137630087e-06,
+      "loss": 1.2246,
+      "step": 7710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.246146549059785e-06,
+      "loss": 1.2432,
+      "step": 7711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.243679226524885e-06,
+      "loss": 1.2363,
+      "step": 7712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241212170200246e-06,
+      "loss": 1.251,
+      "step": 7713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.238745380260712e-06,
+      "loss": 1.2393,
+      "step": 7714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236278856881104e-06,
+      "loss": 1.2236,
+      "step": 7715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23381260023622e-06,
+      "loss": 1.2314,
+      "step": 7716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231346610500846e-06,
+      "loss": 1.2461,
+      "step": 7717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228880887849744e-06,
+      "loss": 1.1943,
+      "step": 7718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226415432457668e-06,
+      "loss": 1.2275,
+      "step": 7719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223950244499337e-06,
+      "loss": 1.21,
+      "step": 7720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2214853241494575e-06,
+      "loss": 1.248,
+      "step": 7721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219020671582724e-06,
+      "loss": 1.1963,
+      "step": 7722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2165562869738075e-06,
+      "loss": 1.2559,
+      "step": 7723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.214092170497351e-06,
+      "loss": 1.2002,
+      "step": 7724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21162832232799e-06,
+      "loss": 1.252,
+      "step": 7725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2091647426403366e-06,
+      "loss": 1.2646,
+      "step": 7726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206701431608992e-06,
+      "loss": 1.251,
+      "step": 7727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.204238389408519e-06,
+      "loss": 1.2236,
+      "step": 7728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201775616213479e-06,
+      "loss": 1.2178,
+      "step": 7729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199313112198411e-06,
+      "loss": 1.1729,
+      "step": 7730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196850877537832e-06,
+      "loss": 1.2617,
+      "step": 7731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.194388912406233e-06,
+      "loss": 1.2305,
+      "step": 7732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191927216978101e-06,
+      "loss": 1.2256,
+      "step": 7733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189465791427893e-06,
+      "loss": 1.2725,
+      "step": 7734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187004635930056e-06,
+      "loss": 1.2178,
+      "step": 7735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184543750659002e-06,
+      "loss": 1.208,
+      "step": 7736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.182083135789141e-06,
+      "loss": 1.2129,
+      "step": 7737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179622791494855e-06,
+      "loss": 1.1924,
+      "step": 7738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177162717950509e-06,
+      "loss": 1.2148,
+      "step": 7739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174702915330445e-06,
+      "loss": 1.2256,
+      "step": 7740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172243383808991e-06,
+      "loss": 1.2021,
+      "step": 7741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169784123560456e-06,
+      "loss": 1.2285,
+      "step": 7742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.167325134759128e-06,
+      "loss": 1.2207,
+      "step": 7743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164866417579269e-06,
+      "loss": 1.2236,
+      "step": 7744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1624079721951346e-06,
+      "loss": 1.2578,
+      "step": 7745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159949798780953e-06,
+      "loss": 1.2334,
+      "step": 7746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157491897510937e-06,
+      "loss": 1.2305,
+      "step": 7747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155034268559271e-06,
+      "loss": 1.209,
+      "step": 7748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152576912100132e-06,
+      "loss": 1.1934,
+      "step": 7749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150119828307674e-06,
+      "loss": 1.2773,
+      "step": 7750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1476630173560296e-06,
+      "loss": 1.2529,
+      "step": 7751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.145206479419309e-06,
+      "loss": 1.1943,
+      "step": 7752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1427502146716115e-06,
+      "loss": 1.2441,
+      "step": 7753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140294223287006e-06,
+      "loss": 1.2285,
+      "step": 7754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.137838505439562e-06,
+      "loss": 1.2793,
+      "step": 7755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135383061303301e-06,
+      "loss": 1.2275,
+      "step": 7756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132927891052247e-06,
+      "loss": 1.2314,
+      "step": 7757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130472994860398e-06,
+      "loss": 1.2148,
+      "step": 7758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.128018372901736e-06,
+      "loss": 1.1973,
+      "step": 7759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.125564025350211e-06,
+      "loss": 1.291,
+      "step": 7760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123109952379767e-06,
+      "loss": 1.2256,
+      "step": 7761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120656154164324e-06,
+      "loss": 1.2051,
+      "step": 7762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118202630877788e-06,
+      "loss": 1.1953,
+      "step": 7763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.115749382694027e-06,
+      "loss": 1.21,
+      "step": 7764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.113296409786913e-06,
+      "loss": 1.2227,
+      "step": 7765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110843712330286e-06,
+      "loss": 1.2373,
+      "step": 7766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10839129049797e-06,
+      "loss": 1.1494,
+      "step": 7767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.105939144463764e-06,
+      "loss": 1.2217,
+      "step": 7768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103487274401454e-06,
+      "loss": 1.2012,
+      "step": 7769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101035680484801e-06,
+      "loss": 1.1621,
+      "step": 7770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098584362887556e-06,
+      "loss": 1.2441,
+      "step": 7771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096133321783437e-06,
+      "loss": 1.2432,
+      "step": 7772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09368255734615e-06,
+      "loss": 1.2197,
+      "step": 7773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.091232069749384e-06,
+      "loss": 1.2744,
+      "step": 7774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088781859166803e-06,
+      "loss": 1.2197,
+      "step": 7775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086331925772053e-06,
+      "loss": 1.2285,
+      "step": 7776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083882269738761e-06,
+      "loss": 1.2266,
+      "step": 7777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081432891240535e-06,
+      "loss": 1.2617,
+      "step": 7778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078983790450961e-06,
+      "loss": 1.2012,
+      "step": 7779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076534967543606e-06,
+      "loss": 1.2266,
+      "step": 7780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.07408642269202e-06,
+      "loss": 1.251,
+      "step": 7781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071638156069729e-06,
+      "loss": 1.2578,
+      "step": 7782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069190167850245e-06,
+      "loss": 1.2783,
+      "step": 7783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0667424582070525e-06,
+      "loss": 1.2402,
+      "step": 7784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064295027313623e-06,
+      "loss": 1.1982,
+      "step": 7785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.061847875343404e-06,
+      "loss": 1.209,
+      "step": 7786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.059401002469829e-06,
+      "loss": 1.2588,
+      "step": 7787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056954408866304e-06,
+      "loss": 1.2617,
+      "step": 7788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.05450809470622e-06,
+      "loss": 1.1924,
+      "step": 7789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052062060162945e-06,
+      "loss": 1.1953,
+      "step": 7790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049616305409833e-06,
+      "loss": 1.2158,
+      "step": 7791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047170830620219e-06,
+      "loss": 1.252,
+      "step": 7792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0447256359674e-06,
+      "loss": 1.2178,
+      "step": 7793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.042280721624678e-06,
+      "loss": 1.2207,
+      "step": 7794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.039836087765319e-06,
+      "loss": 1.2061,
+      "step": 7795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037391734562581e-06,
+      "loss": 1.2344,
+      "step": 7796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034947662189683e-06,
+      "loss": 1.21,
+      "step": 7797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.032503870819846e-06,
+      "loss": 1.1436,
+      "step": 7798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030060360626258e-06,
+      "loss": 1.2246,
+      "step": 7799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027617131782093e-06,
+      "loss": 1.2227,
+      "step": 7800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0251741844604995e-06,
+      "loss": 1.2393,
+      "step": 7801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022731518834609e-06,
+      "loss": 1.2295,
+      "step": 7802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020289135077535e-06,
+      "loss": 1.208,
+      "step": 7803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017847033362369e-06,
+      "loss": 1.1992,
+      "step": 7804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015405213862181e-06,
+      "loss": 1.2266,
+      "step": 7805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012963676750023e-06,
+      "loss": 1.249,
+      "step": 7806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010522422198928e-06,
+      "loss": 1.1865,
+      "step": 7807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.008081450381906e-06,
+      "loss": 1.2334,
+      "step": 7808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0056407614719495e-06,
+      "loss": 1.166,
+      "step": 7809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.003200355642029e-06,
+      "loss": 1.2383,
+      "step": 7810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.000760233065097e-06,
+      "loss": 1.2158,
+      "step": 7811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998320393914088e-06,
+      "loss": 1.1963,
+      "step": 7812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995880838361905e-06,
+      "loss": 1.2217,
+      "step": 7813
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.993441566581446e-06,
+      "loss": 1.2031,
+      "step": 7814
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991002578745579e-06,
+      "loss": 1.25,
+      "step": 7815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988563875027158e-06,
+      "loss": 1.1992,
+      "step": 7816
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.98612545559901e-06,
+      "loss": 1.2148,
+      "step": 7817
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983687320633949e-06,
+      "loss": 1.1934,
+      "step": 7818
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.98124947030476e-06,
+      "loss": 1.2363,
+      "step": 7819
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978811904784223e-06,
+      "loss": 1.2373,
+      "step": 7820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97637462424508e-06,
+      "loss": 1.1924,
+      "step": 7821
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973937628860062e-06,
+      "loss": 1.2393,
+      "step": 7822
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97150091880188e-06,
+      "loss": 1.1963,
+      "step": 7823
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969064494243228e-06,
+      "loss": 1.2285,
+      "step": 7824
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.966628355356767e-06,
+      "loss": 1.2158,
+      "step": 7825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964192502315149e-06,
+      "loss": 1.1777,
+      "step": 7826
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961756935291005e-06,
+      "loss": 1.2188,
+      "step": 7827
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959321654456946e-06,
+      "loss": 1.1914,
+      "step": 7828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.956886659985552e-06,
+      "loss": 1.2471,
+      "step": 7829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954451952049396e-06,
+      "loss": 1.2031,
+      "step": 7830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952017530821026e-06,
+      "loss": 1.2129,
+      "step": 7831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949583396472972e-06,
+      "loss": 1.2812,
+      "step": 7832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947149549177731e-06,
+      "loss": 1.1699,
+      "step": 7833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.944715989107799e-06,
+      "loss": 1.2139,
+      "step": 7834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942282716435639e-06,
+      "loss": 1.2246,
+      "step": 7835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9398497313336975e-06,
+      "loss": 1.2607,
+      "step": 7836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9374170339744e-06,
+      "loss": 1.2051,
+      "step": 7837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934984624530149e-06,
+      "loss": 1.2148,
+      "step": 7838
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932552503173333e-06,
+      "loss": 1.21,
+      "step": 7839
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.930120670076317e-06,
+      "loss": 1.2139,
+      "step": 7840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92768912541144e-06,
+      "loss": 1.2012,
+      "step": 7841
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9252578693510285e-06,
+      "loss": 1.1943,
+      "step": 7842
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922826902067386e-06,
+      "loss": 1.2246,
+      "step": 7843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920396223732796e-06,
+      "loss": 1.2559,
+      "step": 7844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917965834519517e-06,
+      "loss": 1.2324,
+      "step": 7845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915535734599791e-06,
+      "loss": 1.2197,
+      "step": 7846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913105924145841e-06,
+      "loss": 1.1973,
+      "step": 7847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.91067640332987e-06,
+      "loss": 1.2256,
+      "step": 7848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.908247172324053e-06,
+      "loss": 1.2207,
+      "step": 7849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.905818231300551e-06,
+      "loss": 1.2432,
+      "step": 7850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.903389580431502e-06,
+      "loss": 1.1787,
+      "step": 7851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.90096121988903e-06,
+      "loss": 1.21,
+      "step": 7852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.898533149845225e-06,
+      "loss": 1.2021,
+      "step": 7853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8961053704721675e-06,
+      "loss": 1.2412,
+      "step": 7854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.893677881941913e-06,
+      "loss": 1.1934,
+      "step": 7855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.891250684426506e-06,
+      "loss": 1.209,
+      "step": 7856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88882377809795e-06,
+      "loss": 1.1982,
+      "step": 7857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886397163128243e-06,
+      "loss": 1.2578,
+      "step": 7858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.883970839689359e-06,
+      "loss": 1.2441,
+      "step": 7859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881544807953259e-06,
+      "loss": 1.1982,
+      "step": 7860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879119068091868e-06,
+      "loss": 1.2217,
+      "step": 7861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.876693620277094e-06,
+      "loss": 1.2295,
+      "step": 7862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.874268464680839e-06,
+      "loss": 1.2256,
+      "step": 7863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8718436014749715e-06,
+      "loss": 1.2471,
+      "step": 7864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869419030831332e-06,
+      "loss": 1.2217,
+      "step": 7865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866994752921759e-06,
+      "loss": 1.207,
+      "step": 7866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8645707679180584e-06,
+      "loss": 1.209,
+      "step": 7867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.862147075992023e-06,
+      "loss": 1.1787,
+      "step": 7868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.859723677315407e-06,
+      "loss": 1.2129,
+      "step": 7869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857300572059967e-06,
+      "loss": 1.2227,
+      "step": 7870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854877760397425e-06,
+      "loss": 1.1816,
+      "step": 7871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852455242499489e-06,
+      "loss": 1.2139,
+      "step": 7872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850033018537836e-06,
+      "loss": 1.2402,
+      "step": 7873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.847611088684133e-06,
+      "loss": 1.2451,
+      "step": 7874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.845189453110023e-06,
+      "loss": 1.2715,
+      "step": 7875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842768111987126e-06,
+      "loss": 1.1982,
+      "step": 7876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840347065487042e-06,
+      "loss": 1.2139,
+      "step": 7877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.837926313781347e-06,
+      "loss": 1.2168,
+      "step": 7878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835505857041605e-06,
+      "loss": 1.1719,
+      "step": 7879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833085695439358e-06,
+      "loss": 1.2451,
+      "step": 7880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830665829146109e-06,
+      "loss": 1.207,
+      "step": 7881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8282462583333645e-06,
+      "loss": 1.2324,
+      "step": 7882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.825826983172595e-06,
+      "loss": 1.2246,
+      "step": 7883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82340800383526e-06,
+      "loss": 1.2305,
+      "step": 7884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.820989320492779e-06,
+      "loss": 1.1953,
+      "step": 7885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8185709333165785e-06,
+      "loss": 1.2217,
+      "step": 7886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816152842478042e-06,
+      "loss": 1.2373,
+      "step": 7887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8137350481485476e-06,
+      "loss": 1.2539,
+      "step": 7888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811317550499433e-06,
+      "loss": 1.2139,
+      "step": 7889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.808900349702031e-06,
+      "loss": 1.2148,
+      "step": 7890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.806483445927649e-06,
+      "loss": 1.1953,
+      "step": 7891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804066839347573e-06,
+      "loss": 1.2539,
+      "step": 7892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801650530133067e-06,
+      "loss": 1.2676,
+      "step": 7893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799234518455375e-06,
+      "loss": 1.2705,
+      "step": 7894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79681880448572e-06,
+      "loss": 1.2178,
+      "step": 7895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794403388395307e-06,
+      "loss": 1.2246,
+      "step": 7896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.791988270355309e-06,
+      "loss": 1.2119,
+      "step": 7897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789573450536887e-06,
+      "loss": 1.2441,
+      "step": 7898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787158929111184e-06,
+      "loss": 1.2402,
+      "step": 7899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.784744706249312e-06,
+      "loss": 1.207,
+      "step": 7900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782330782122371e-06,
+      "loss": 1.2021,
+      "step": 7901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779917156901432e-06,
+      "loss": 1.2559,
+      "step": 7902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777503830757551e-06,
+      "loss": 1.2207,
+      "step": 7903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.775090803861763e-06,
+      "loss": 1.209,
+      "step": 7904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772678076385072e-06,
+      "loss": 1.2666,
+      "step": 7905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770265648498471e-06,
+      "loss": 1.2129,
+      "step": 7906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767853520372926e-06,
+      "loss": 1.248,
+      "step": 7907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765441692179395e-06,
+      "loss": 1.248,
+      "step": 7908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7630301640887905e-06,
+      "loss": 1.2617,
+      "step": 7909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.760618936272027e-06,
+      "loss": 1.2461,
+      "step": 7910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7582080088999814e-06,
+      "loss": 1.2314,
+      "step": 7911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755797382143524e-06,
+      "loss": 1.1963,
+      "step": 7912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753387056173489e-06,
+      "loss": 1.2393,
+      "step": 7913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750977031160691e-06,
+      "loss": 1.2295,
+      "step": 7914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.748567307275942e-06,
+      "loss": 1.2109,
+      "step": 7915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746157884690015e-06,
+      "loss": 1.1924,
+      "step": 7916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7437487635736595e-06,
+      "loss": 1.2305,
+      "step": 7917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.741339944097613e-06,
+      "loss": 1.2031,
+      "step": 7918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.73893142643259e-06,
+      "loss": 1.2236,
+      "step": 7919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736523210749283e-06,
+      "loss": 1.2197,
+      "step": 7920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734115297218353e-06,
+      "loss": 1.1816,
+      "step": 7921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731707686010459e-06,
+      "loss": 1.2197,
+      "step": 7922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7293003772962256e-06,
+      "loss": 1.2451,
+      "step": 7923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.726893371246258e-06,
+      "loss": 1.2158,
+      "step": 7924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.724486668031144e-06,
+      "loss": 1.2324,
+      "step": 7925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722080267821439e-06,
+      "loss": 1.2002,
+      "step": 7926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719674170787689e-06,
+      "loss": 1.1992,
+      "step": 7927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7172683771004135e-06,
+      "loss": 1.1709,
+      "step": 7928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71486288693011e-06,
+      "loss": 1.2246,
+      "step": 7929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712457700447258e-06,
+      "loss": 1.2002,
+      "step": 7930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7100528178223115e-06,
+      "loss": 1.168,
+      "step": 7931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7076482392257025e-06,
+      "loss": 1.21,
+      "step": 7932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.705243964827851e-06,
+      "loss": 1.2012,
+      "step": 7933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7028399947991365e-06,
+      "loss": 1.2646,
+      "step": 7934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700436329309935e-06,
+      "loss": 1.2236,
+      "step": 7935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.698032968530591e-06,
+      "loss": 1.2051,
+      "step": 7936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695629912631435e-06,
+      "loss": 1.2466,
+      "step": 7937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6932271617827674e-06,
+      "loss": 1.2324,
+      "step": 7938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6908247161548724e-06,
+      "loss": 1.2012,
+      "step": 7939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.688422575918011e-06,
+      "loss": 1.25,
+      "step": 7940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6860207412424285e-06,
+      "loss": 1.2129,
+      "step": 7941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6836192122983325e-06,
+      "loss": 1.2207,
+      "step": 7942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681217989255921e-06,
+      "loss": 1.2051,
+      "step": 7943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6788170722853745e-06,
+      "loss": 1.2256,
+      "step": 7944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.676416461556846e-06,
+      "loss": 1.2461,
+      "step": 7945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.674016157240463e-06,
+      "loss": 1.1816,
+      "step": 7946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6716161595063325e-06,
+      "loss": 1.168,
+      "step": 7947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669216468524546e-06,
+      "loss": 1.1855,
+      "step": 7948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666817084465174e-06,
+      "loss": 1.2217,
+      "step": 7949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.664418007498247e-06,
+      "loss": 1.1963,
+      "step": 7950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6620192377938e-06,
+      "loss": 1.2354,
+      "step": 7951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659620775521829e-06,
+      "loss": 1.2539,
+      "step": 7952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6572226208523186e-06,
+      "loss": 1.2158,
+      "step": 7953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6548247739552164e-06,
+      "loss": 1.2246,
+      "step": 7954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652427235000462e-06,
+      "loss": 1.2451,
+      "step": 7955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.650030004157969e-06,
+      "loss": 1.252,
+      "step": 7956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.64763308159763e-06,
+      "loss": 1.2646,
+      "step": 7957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645236467489313e-06,
+      "loss": 1.2119,
+      "step": 7958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642840162002867e-06,
+      "loss": 1.2334,
+      "step": 7959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640444165308119e-06,
+      "loss": 1.1982,
+      "step": 7960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638048477574873e-06,
+      "loss": 1.2451,
+      "step": 7961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635653098972909e-06,
+      "loss": 1.2031,
+      "step": 7962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6332580296719885e-06,
+      "loss": 1.21,
+      "step": 7963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6308632698418485e-06,
+      "loss": 1.2207,
+      "step": 7964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628468819652209e-06,
+      "loss": 1.1836,
+      "step": 7965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62607467927276e-06,
+      "loss": 1.2314,
+      "step": 7966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623680848873178e-06,
+      "loss": 1.2178,
+      "step": 7967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621287328623113e-06,
+      "loss": 1.2324,
+      "step": 7968
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618894118692198e-06,
+      "loss": 1.2041,
+      "step": 7969
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616501219250028e-06,
+      "loss": 1.1992,
+      "step": 7970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614108630466196e-06,
+      "loss": 1.2178,
+      "step": 7971
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611716352510264e-06,
+      "loss": 1.2451,
+      "step": 7972
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.60932438555177e-06,
+      "loss": 1.2822,
+      "step": 7973
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606932729760235e-06,
+      "loss": 1.1797,
+      "step": 7974
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604541385305154e-06,
+      "loss": 1.2168,
+      "step": 7975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.602150352356002e-06,
+      "loss": 1.1826,
+      "step": 7976
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.599759631082236e-06,
+      "loss": 1.1885,
+      "step": 7977
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.597369221653278e-06,
+      "loss": 1.2783,
+      "step": 7978
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594979124238535e-06,
+      "loss": 1.2305,
+      "step": 7979
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.592589339007404e-06,
+      "loss": 1.2285,
+      "step": 7980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590199866129247e-06,
+      "loss": 1.2285,
+      "step": 7981
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.587810705773397e-06,
+      "loss": 1.2139,
+      "step": 7982
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.585421858109179e-06,
+      "loss": 1.2842,
+      "step": 7983
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5830333233058895e-06,
+      "loss": 1.2383,
+      "step": 7984
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580645101532809e-06,
+      "loss": 1.2461,
+      "step": 7985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.578257192959178e-06,
+      "loss": 1.168,
+      "step": 7986
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.57586959775424e-06,
+      "loss": 1.2236,
+      "step": 7987
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5734823160871985e-06,
+      "loss": 1.1621,
+      "step": 7988
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.571095348127246e-06,
+      "loss": 1.2256,
+      "step": 7989
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568708694043537e-06,
+      "loss": 1.207,
+      "step": 7990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566322354005218e-06,
+      "loss": 1.2617,
+      "step": 7991
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.563936328181409e-06,
+      "loss": 1.209,
+      "step": 7992
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561550616741207e-06,
+      "loss": 1.2363,
+      "step": 7993
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559165219853688e-06,
+      "loss": 1.2295,
+      "step": 7994
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556780137687906e-06,
+      "loss": 1.208,
+      "step": 7995
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554395370412889e-06,
+      "loss": 1.2031,
+      "step": 7996
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.552010918197652e-06,
+      "loss": 1.1963,
+      "step": 7997
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54962678121117e-06,
+      "loss": 1.1992,
+      "step": 7998
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547242959622414e-06,
+      "loss": 1.2402,
+      "step": 7999
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544859453600323e-06,
+      "loss": 1.2715,
+      "step": 8000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.542476263313816e-06,
+      "loss": 1.2178,
+      "step": 8001
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540093388931792e-06,
+      "loss": 1.2266,
+      "step": 8002
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537710830623122e-06,
+      "loss": 1.2227,
+      "step": 8003
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.535328588556661e-06,
+      "loss": 1.2549,
+      "step": 8004
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.532946662901239e-06,
+      "loss": 1.2559,
+      "step": 8005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530565053825656e-06,
+      "loss": 1.2402,
+      "step": 8006
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.528183761498702e-06,
+      "loss": 1.21,
+      "step": 8007
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.525802786089139e-06,
+      "loss": 1.208,
+      "step": 8008
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523422127765704e-06,
+      "loss": 1.209,
+      "step": 8009
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5210417866971165e-06,
+      "loss": 1.2168,
+      "step": 8010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518661763052071e-06,
+      "loss": 1.207,
+      "step": 8011
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516282056999238e-06,
+      "loss": 1.1924,
+      "step": 8012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.513902668707274e-06,
+      "loss": 1.2197,
+      "step": 8013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511523598344795e-06,
+      "loss": 1.2275,
+      "step": 8014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.509144846080408e-06,
+      "loss": 1.2793,
+      "step": 8015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506766412082701e-06,
+      "loss": 1.1934,
+      "step": 8016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504388296520236e-06,
+      "loss": 1.2441,
+      "step": 8017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.502010499561539e-06,
+      "loss": 1.2656,
+      "step": 8018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499633021375131e-06,
+      "loss": 1.2588,
+      "step": 8019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.497255862129502e-06,
+      "loss": 1.2324,
+      "step": 8020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.494879021993127e-06,
+      "loss": 1.2686,
+      "step": 8021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492502501134438e-06,
+      "loss": 1.1924,
+      "step": 8022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490126299721872e-06,
+      "loss": 1.1973,
+      "step": 8023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487750417923827e-06,
+      "loss": 1.2275,
+      "step": 8024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485374855908685e-06,
+      "loss": 1.2559,
+      "step": 8025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482999613844795e-06,
+      "loss": 1.1953,
+      "step": 8026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.480624691900492e-06,
+      "loss": 1.2295,
+      "step": 8027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478250090244088e-06,
+      "loss": 1.2207,
+      "step": 8028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475875809043872e-06,
+      "loss": 1.2217,
+      "step": 8029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473501848468108e-06,
+      "loss": 1.2422,
+      "step": 8030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471128208685038e-06,
+      "loss": 1.2432,
+      "step": 8031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.468754889862883e-06,
+      "loss": 1.2266,
+      "step": 8032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466381892169843e-06,
+      "loss": 1.1816,
+      "step": 8033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.464009215774083e-06,
+      "loss": 1.1992,
+      "step": 8034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.461636860843762e-06,
+      "loss": 1.2568,
+      "step": 8035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459264827547005e-06,
+      "loss": 1.1885,
+      "step": 8036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45689311605192e-06,
+      "loss": 1.2158,
+      "step": 8037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45452172652659e-06,
+      "loss": 1.1826,
+      "step": 8038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452150659139075e-06,
+      "loss": 1.208,
+      "step": 8039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4497799140574116e-06,
+      "loss": 1.2393,
+      "step": 8040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.447409491449618e-06,
+      "loss": 1.2158,
+      "step": 8041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.445039391483679e-06,
+      "loss": 1.2031,
+      "step": 8042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442669614327569e-06,
+      "loss": 1.1992,
+      "step": 8043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440300160149232e-06,
+      "loss": 1.2461,
+      "step": 8044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437931029116592e-06,
+      "loss": 1.249,
+      "step": 8045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4355622213975475e-06,
+      "loss": 1.1748,
+      "step": 8046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.433193737159979e-06,
+      "loss": 1.2295,
+      "step": 8047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430825576571736e-06,
+      "loss": 1.2451,
+      "step": 8048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428457739800659e-06,
+      "loss": 1.2295,
+      "step": 8049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.426090227014544e-06,
+      "loss": 1.1641,
+      "step": 8050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423723038381181e-06,
+      "loss": 1.2334,
+      "step": 8051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421356174068338e-06,
+      "loss": 1.2559,
+      "step": 8052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.418989634243752e-06,
+      "loss": 1.21,
+      "step": 8053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416623419075138e-06,
+      "loss": 1.2236,
+      "step": 8054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.414257528730188e-06,
+      "loss": 1.1836,
+      "step": 8055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411891963376574e-06,
+      "loss": 1.2402,
+      "step": 8056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409526723181946e-06,
+      "loss": 1.1885,
+      "step": 8057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407161808313926e-06,
+      "loss": 1.2139,
+      "step": 8058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4047972189401144e-06,
+      "loss": 1.2666,
+      "step": 8059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402432955228092e-06,
+      "loss": 1.2002,
+      "step": 8060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4000690173454115e-06,
+      "loss": 1.2109,
+      "step": 8061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397705405459612e-06,
+      "loss": 1.1582,
+      "step": 8062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395342119738192e-06,
+      "loss": 1.1777,
+      "step": 8063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3929791603486435e-06,
+      "loss": 1.2324,
+      "step": 8064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390616527458428e-06,
+      "loss": 1.2227,
+      "step": 8065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388254221234985e-06,
+      "loss": 1.2217,
+      "step": 8066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.385892241845734e-06,
+      "loss": 1.2305,
+      "step": 8067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383530589458064e-06,
+      "loss": 1.2148,
+      "step": 8068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381169264239347e-06,
+      "loss": 1.1787,
+      "step": 8069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378808266356934e-06,
+      "loss": 1.2158,
+      "step": 8070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37644759597814e-06,
+      "loss": 1.2598,
+      "step": 8071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.374087253270272e-06,
+      "loss": 1.2402,
+      "step": 8072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371727238400606e-06,
+      "loss": 1.1934,
+      "step": 8073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369367551536395e-06,
+      "loss": 1.2197,
+      "step": 8074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.367008192844871e-06,
+      "loss": 1.2207,
+      "step": 8075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.364649162493243e-06,
+      "loss": 1.2354,
+      "step": 8076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3622904606486926e-06,
+      "loss": 1.2295,
+      "step": 8077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.359932087478386e-06,
+      "loss": 1.2666,
+      "step": 8078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357574043149454e-06,
+      "loss": 1.2178,
+      "step": 8079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355216327829009e-06,
+      "loss": 1.2607,
+      "step": 8080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.352858941684151e-06,
+      "loss": 1.2637,
+      "step": 8081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3505018848819505e-06,
+      "loss": 1.1992,
+      "step": 8082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348145157589439e-06,
+      "loss": 1.2393,
+      "step": 8083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345788759973647e-06,
+      "loss": 1.2402,
+      "step": 8084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3434326922015675e-06,
+      "loss": 1.2285,
+      "step": 8085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341076954440183e-06,
+      "loss": 1.252,
+      "step": 8086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338721546856429e-06,
+      "loss": 1.2412,
+      "step": 8087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336366469617247e-06,
+      "loss": 1.2178,
+      "step": 8088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334011722889537e-06,
+      "loss": 1.1865,
+      "step": 8089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331657306840183e-06,
+      "loss": 1.2607,
+      "step": 8090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329303221636037e-06,
+      "loss": 1.1963,
+      "step": 8091
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326949467443932e-06,
+      "loss": 1.1992,
+      "step": 8092
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3245960444306835e-06,
+      "loss": 1.252,
+      "step": 8093
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.322242952763076e-06,
+      "loss": 1.2295,
+      "step": 8094
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319890192607873e-06,
+      "loss": 1.2295,
+      "step": 8095
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.317537764131815e-06,
+      "loss": 1.248,
+      "step": 8096
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.315185667501618e-06,
+      "loss": 1.2217,
+      "step": 8097
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3128339028839785e-06,
+      "loss": 1.2002,
+      "step": 8098
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310482470445559e-06,
+      "loss": 1.1924,
+      "step": 8099
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3081313703530104e-06,
+      "loss": 1.2119,
+      "step": 8100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305780602772953e-06,
+      "loss": 1.2666,
+      "step": 8101
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3034301678719854e-06,
+      "loss": 1.2627,
+      "step": 8102
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.301080065816685e-06,
+      "loss": 1.2412,
+      "step": 8103
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298730296773602e-06,
+      "loss": 1.2266,
+      "step": 8104
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296380860909264e-06,
+      "loss": 1.2344,
+      "step": 8105
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2940317583901805e-06,
+      "loss": 1.1816,
+      "step": 8106
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.291682989382822e-06,
+      "loss": 1.1973,
+      "step": 8107
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289334554053654e-06,
+      "loss": 1.2334,
+      "step": 8108
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286986452569105e-06,
+      "loss": 1.2314,
+      "step": 8109
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284638685095589e-06,
+      "loss": 1.2314,
+      "step": 8110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.28229125179949e-06,
+      "loss": 1.207,
+      "step": 8111
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2799441528471694e-06,
+      "loss": 1.2451,
+      "step": 8112
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277597388404968e-06,
+      "loss": 1.1611,
+      "step": 8113
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275250958639204e-06,
+      "loss": 1.1719,
+      "step": 8114
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272904863716162e-06,
+      "loss": 1.2646,
+      "step": 8115
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270559103802108e-06,
+      "loss": 1.2275,
+      "step": 8116
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268213679063295e-06,
+      "loss": 1.1978,
+      "step": 8117
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.265868589665943e-06,
+      "loss": 1.1729,
+      "step": 8118
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2635238357762395e-06,
+      "loss": 1.1934,
+      "step": 8119
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261179417560362e-06,
+      "loss": 1.2207,
+      "step": 8120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258835335184458e-06,
+      "loss": 1.2383,
+      "step": 8121
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2564915888146585e-06,
+      "loss": 1.2188,
+      "step": 8122
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.254148178617051e-06,
+      "loss": 1.1748,
+      "step": 8123
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251805104757725e-06,
+      "loss": 1.2051,
+      "step": 8124
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.249462367402733e-06,
+      "loss": 1.208,
+      "step": 8125
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2471199667181034e-06,
+      "loss": 1.2334,
+      "step": 8126
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2447779028698375e-06,
+      "loss": 1.1611,
+      "step": 8127
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242436176023922e-06,
+      "loss": 1.2217,
+      "step": 8128
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240094786346312e-06,
+      "loss": 1.2256,
+      "step": 8129
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237753734002944e-06,
+      "loss": 1.2041,
+      "step": 8130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2354130191597265e-06,
+      "loss": 1.2627,
+      "step": 8131
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.233072641982547e-06,
+      "loss": 1.1914,
+      "step": 8132
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.230732602637269e-06,
+      "loss": 1.2217,
+      "step": 8133
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228392901289732e-06,
+      "loss": 1.2168,
+      "step": 8134
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226053538105744e-06,
+      "loss": 1.2051,
+      "step": 8135
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223714513251101e-06,
+      "loss": 1.2119,
+      "step": 8136
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221375826891568e-06,
+      "loss": 1.248,
+      "step": 8137
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219037479192888e-06,
+      "loss": 1.1953,
+      "step": 8138
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216699470320779e-06,
+      "loss": 1.2275,
+      "step": 8139
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2143618004409365e-06,
+      "loss": 1.2031,
+      "step": 8140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.21202446971903e-06,
+      "loss": 1.1914,
+      "step": 8141
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.209687478320712e-06,
+      "loss": 1.1777,
+      "step": 8142
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.207350826411595e-06,
+      "loss": 1.2051,
+      "step": 8143
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2050145141572805e-06,
+      "loss": 1.1885,
+      "step": 8144
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202678541723345e-06,
+      "loss": 1.2036,
+      "step": 8145
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200342909275339e-06,
+      "loss": 1.2383,
+      "step": 8146
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.198007616978786e-06,
+      "loss": 1.2373,
+      "step": 8147
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.195672664999191e-06,
+      "loss": 1.2637,
+      "step": 8148
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193338053502032e-06,
+      "loss": 1.2422,
+      "step": 8149
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.191003782652764e-06,
+      "loss": 1.3232,
+      "step": 8150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.18866985261681e-06,
+      "loss": 1.2363,
+      "step": 8151
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186336263559575e-06,
+      "loss": 1.1777,
+      "step": 8152
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184003015646451e-06,
+      "loss": 1.2227,
+      "step": 8153
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181670109042791e-06,
+      "loss": 1.2471,
+      "step": 8154
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.179337543913922e-06,
+      "loss": 1.1963,
+      "step": 8155
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177005320425159e-06,
+      "loss": 1.2529,
+      "step": 8156
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174673438741783e-06,
+      "loss": 1.2314,
+      "step": 8157
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.17234189902906e-06,
+      "loss": 1.2158,
+      "step": 8158
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170010701452214e-06,
+      "loss": 1.2314,
+      "step": 8159
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167679846176468e-06,
+      "loss": 1.1758,
+      "step": 8160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.165349333367007e-06,
+      "loss": 1.252,
+      "step": 8161
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.163019163188998e-06,
+      "loss": 1.2246,
+      "step": 8162
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160689335807572e-06,
+      "loss": 1.2012,
+      "step": 8163
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158359851387846e-06,
+      "loss": 1.1611,
+      "step": 8164
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156030710094913e-06,
+      "loss": 1.167,
+      "step": 8165
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.153701912093837e-06,
+      "loss": 1.248,
+      "step": 8166
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151373457549662e-06,
+      "loss": 1.2041,
+      "step": 8167
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149045346627403e-06,
+      "loss": 1.1963,
+      "step": 8168
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1467175794920555e-06,
+      "loss": 1.2119,
+      "step": 8169
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14439015630859e-06,
+      "loss": 1.2305,
+      "step": 8170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142063077241946e-06,
+      "loss": 1.2314,
+      "step": 8171
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139736342457042e-06,
+      "loss": 1.2324,
+      "step": 8172
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137409952118779e-06,
+      "loss": 1.2051,
+      "step": 8173
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135083906392026e-06,
+      "loss": 1.1904,
+      "step": 8174
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132758205441631e-06,
+      "loss": 1.2334,
+      "step": 8175
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.130432849432414e-06,
+      "loss": 1.2227,
+      "step": 8176
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.128107838529175e-06,
+      "loss": 1.2441,
+      "step": 8177
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12578317289669e-06,
+      "loss": 1.2314,
+      "step": 8178
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.123458852699701e-06,
+      "loss": 1.2314,
+      "step": 8179
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1211348781029345e-06,
+      "loss": 1.1992,
+      "step": 8180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1188112492710925e-06,
+      "loss": 1.2432,
+      "step": 8181
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1164879663688505e-06,
+      "loss": 1.25,
+      "step": 8182
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114165029560858e-06,
+      "loss": 1.2695,
+      "step": 8183
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111842439011743e-06,
+      "loss": 1.2256,
+      "step": 8184
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109520194886106e-06,
+      "loss": 1.2041,
+      "step": 8185
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107198297348528e-06,
+      "loss": 1.2178,
+      "step": 8186
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.104876746563557e-06,
+      "loss": 1.2197,
+      "step": 8187
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1025555426957155e-06,
+      "loss": 1.2217,
+      "step": 8188
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10023468590952e-06,
+      "loss": 1.2783,
+      "step": 8189
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.097914176369445e-06,
+      "loss": 1.3027,
+      "step": 8190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.095594014239946e-06,
+      "loss": 1.2256,
+      "step": 8191
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.093274199685447e-06,
+      "loss": 1.2021,
+      "step": 8192
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.090954732870358e-06,
+      "loss": 1.1816,
+      "step": 8193
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088635613959057e-06,
+      "loss": 1.2178,
+      "step": 8194
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.086316843115901e-06,
+      "loss": 1.2441,
+      "step": 8195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083998420505223e-06,
+      "loss": 1.2549,
+      "step": 8196
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081680346291326e-06,
+      "loss": 1.209,
+      "step": 8197
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079362620638494e-06,
+      "loss": 1.1934,
+      "step": 8198
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077045243710987e-06,
+      "loss": 1.2041,
+      "step": 8199
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074728215673031e-06,
+      "loss": 1.2236,
+      "step": 8200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0724115366888346e-06,
+      "loss": 1.2559,
+      "step": 8201
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0700952069225855e-06,
+      "loss": 1.1787,
+      "step": 8202
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067779226538438e-06,
+      "loss": 1.2031,
+      "step": 8203
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065463595700526e-06,
+      "loss": 1.2354,
+      "step": 8204
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063148314572961e-06,
+      "loss": 1.2227,
+      "step": 8205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060833383319823e-06,
+      "loss": 1.1846,
+      "step": 8206
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058518802105177e-06,
+      "loss": 1.2393,
+      "step": 8207
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056204571093049e-06,
+      "loss": 1.2549,
+      "step": 8208
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053890690447452e-06,
+      "loss": 1.1885,
+      "step": 8209
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051577160332371e-06,
+      "loss": 1.2139,
+      "step": 8210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049263980911765e-06,
+      "loss": 1.1787,
+      "step": 8211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.046951152349572e-06,
+      "loss": 1.2295,
+      "step": 8212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0446386748096985e-06,
+      "loss": 1.2295,
+      "step": 8213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042326548456032e-06,
+      "loss": 1.2236,
+      "step": 8214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040014773452433e-06,
+      "loss": 1.1846,
+      "step": 8215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037703349962733e-06,
+      "loss": 1.1885,
+      "step": 8216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035392278150746e-06,
+      "loss": 1.1885,
+      "step": 8217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.033081558180251e-06,
+      "loss": 1.2256,
+      "step": 8218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030771190215023e-06,
+      "loss": 1.2158,
+      "step": 8219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028461174418785e-06,
+      "loss": 1.1836,
+      "step": 8220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026151510955251e-06,
+      "loss": 1.251,
+      "step": 8221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023842199988107e-06,
+      "loss": 1.1982,
+      "step": 8222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0215332416810195e-06,
+      "loss": 1.2285,
+      "step": 8223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0192246361976096e-06,
+      "loss": 1.2568,
+      "step": 8224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0169163837015025e-06,
+      "loss": 1.2383,
+      "step": 8225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014608484356276e-06,
+      "loss": 1.252,
+      "step": 8226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0123009383255e-06,
+      "loss": 1.2275,
+      "step": 8227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009993745772699e-06,
+      "loss": 1.2217,
+      "step": 8228
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007686906861388e-06,
+      "loss": 1.2461,
+      "step": 8229
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005380421755053e-06,
+      "loss": 1.2363,
+      "step": 8230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.003074290617153e-06,
+      "loss": 1.2031,
+      "step": 8231
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000768513611126e-06,
+      "loss": 1.2422,
+      "step": 8232
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998463090900381e-06,
+      "loss": 1.2598,
+      "step": 8233
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996158022648303e-06,
+      "loss": 1.2588,
+      "step": 8234
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.993853309018256e-06,
+      "loss": 1.2314,
+      "step": 8235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991548950173568e-06,
+      "loss": 1.2139,
+      "step": 8236
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989244946277551e-06,
+      "loss": 1.1958,
+      "step": 8237
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986941297493491e-06,
+      "loss": 1.1436,
+      "step": 8238
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984638003984647e-06,
+      "loss": 1.2334,
+      "step": 8239
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9823350659142545e-06,
+      "loss": 1.1719,
+      "step": 8240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.980032483445521e-06,
+      "loss": 1.2119,
+      "step": 8241
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977730256741629e-06,
+      "loss": 1.2256,
+      "step": 8242
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975428385965746e-06,
+      "loss": 1.2061,
+      "step": 8243
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9731268712809935e-06,
+      "loss": 1.2891,
+      "step": 8244
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.970825712850483e-06,
+      "loss": 1.2051,
+      "step": 8245
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968524910837301e-06,
+      "loss": 1.2188,
+      "step": 8246
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966224465404503e-06,
+      "loss": 1.2021,
+      "step": 8247
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963924376715123e-06,
+      "loss": 1.2305,
+      "step": 8248
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961624644932166e-06,
+      "loss": 1.1875,
+      "step": 8249
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9593252702186155e-06,
+      "loss": 1.2334,
+      "step": 8250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957026252737431e-06,
+      "loss": 1.1875,
+      "step": 8251
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.954727592651537e-06,
+      "loss": 1.2256,
+      "step": 8252
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.952429290123839e-06,
+      "loss": 1.2393,
+      "step": 8253
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.950131345317225e-06,
+      "loss": 1.2852,
+      "step": 8254
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94783375839455e-06,
+      "loss": 1.1914,
+      "step": 8255
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945536529518637e-06,
+      "loss": 1.2129,
+      "step": 8256
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943239658852294e-06,
+      "loss": 1.2432,
+      "step": 8257
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.940943146558299e-06,
+      "loss": 1.2588,
+      "step": 8258
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938646992799411e-06,
+      "loss": 1.2383,
+      "step": 8259
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936351197738346e-06,
+      "loss": 1.2031,
+      "step": 8260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934055761537819e-06,
+      "loss": 1.2441,
+      "step": 8261
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931760684360505e-06,
+      "loss": 1.1777,
+      "step": 8262
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929465966369056e-06,
+      "loss": 1.2354,
+      "step": 8263
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9271716077260945e-06,
+      "loss": 1.2676,
+      "step": 8264
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9248776085942226e-06,
+      "loss": 1.251,
+      "step": 8265
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922583969136019e-06,
+      "loss": 1.209,
+      "step": 8266
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920290689514031e-06,
+      "loss": 1.2031,
+      "step": 8267
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917997769890785e-06,
+      "loss": 1.2158,
+      "step": 8268
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9157052104287785e-06,
+      "loss": 1.2598,
+      "step": 8269
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913413011290487e-06,
+      "loss": 1.2314,
+      "step": 8270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91112117263836e-06,
+      "loss": 1.25,
+      "step": 8271
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908829694634815e-06,
+      "loss": 1.251,
+      "step": 8272
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90653857744225e-06,
+      "loss": 1.1943,
+      "step": 8273
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.904247821223039e-06,
+      "loss": 1.1816,
+      "step": 8274
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901957426139526e-06,
+      "loss": 1.2441,
+      "step": 8275
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899667392354033e-06,
+      "loss": 1.1885,
+      "step": 8276
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897377720028852e-06,
+      "loss": 1.2471,
+      "step": 8277
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.895088409326254e-06,
+      "loss": 1.2334,
+      "step": 8278
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892799460408485e-06,
+      "loss": 1.2002,
+      "step": 8279
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890510873437756e-06,
+      "loss": 1.2217,
+      "step": 8280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888222648576262e-06,
+      "loss": 1.2266,
+      "step": 8281
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8859347859861695e-06,
+      "loss": 1.2129,
+      "step": 8282
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.88364728582962e-06,
+      "loss": 1.2383,
+      "step": 8283
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881360148268728e-06,
+      "loss": 1.2715,
+      "step": 8284
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879073373465583e-06,
+      "loss": 1.209,
+      "step": 8285
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.876786961582247e-06,
+      "loss": 1.2109,
+      "step": 8286
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874500912780764e-06,
+      "loss": 1.21,
+      "step": 8287
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872215227223138e-06,
+      "loss": 1.1924,
+      "step": 8288
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869929905071354e-06,
+      "loss": 1.1846,
+      "step": 8289
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.867644946487384e-06,
+      "loss": 1.2021,
+      "step": 8290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865360351633157e-06,
+      "loss": 1.2256,
+      "step": 8291
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863076120670579e-06,
+      "loss": 1.207,
+      "step": 8292
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.860792253761536e-06,
+      "loss": 1.2344,
+      "step": 8293
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8585087510678865e-06,
+      "loss": 1.1807,
+      "step": 8294
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856225612751464e-06,
+      "loss": 1.2197,
+      "step": 8295
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853942838974064e-06,
+      "loss": 1.1924,
+      "step": 8296
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851660429897477e-06,
+      "loss": 1.2227,
+      "step": 8297
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849378385683455e-06,
+      "loss": 1.1924,
+      "step": 8298
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84709670649373e-06,
+      "loss": 1.2432,
+      "step": 8299
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844815392489996e-06,
+      "loss": 1.2314,
+      "step": 8300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842534443833934e-06,
+      "loss": 1.2051,
+      "step": 8301
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840253860687194e-06,
+      "loss": 1.1992,
+      "step": 8302
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837973643211402e-06,
+      "loss": 1.2188,
+      "step": 8303
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.835693791568156e-06,
+      "loss": 1.145,
+      "step": 8304
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.83341430591903e-06,
+      "loss": 1.1992,
+      "step": 8305
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8311351864255686e-06,
+      "loss": 1.2393,
+      "step": 8306
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.828856433249299e-06,
+      "loss": 1.2803,
+      "step": 8307
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826578046551708e-06,
+      "loss": 1.1943,
+      "step": 8308
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824300026494269e-06,
+      "loss": 1.2207,
+      "step": 8309
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822022373238424e-06,
+      "loss": 1.2656,
+      "step": 8310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819745086945591e-06,
+      "loss": 1.25,
+      "step": 8311
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8174681677771615e-06,
+      "loss": 1.2461,
+      "step": 8312
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8151916158945e-06,
+      "loss": 1.2021,
+      "step": 8313
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812915431458945e-06,
+      "loss": 1.208,
+      "step": 8314
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8106396146318155e-06,
+      "loss": 1.2051,
+      "step": 8315
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80836416557439e-06,
+      "loss": 1.2432,
+      "step": 8316
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806089084447932e-06,
+      "loss": 1.1992,
+      "step": 8317
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803814371413677e-06,
+      "loss": 1.2324,
+      "step": 8318
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801540026632835e-06,
+      "loss": 1.2236,
+      "step": 8319
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799266050266587e-06,
+      "loss": 1.1797,
+      "step": 8320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.79699244247609e-06,
+      "loss": 1.1611,
+      "step": 8321
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.794719203422476e-06,
+      "loss": 1.2178,
+      "step": 8322
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792446333266847e-06,
+      "loss": 1.2178,
+      "step": 8323
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790173832170287e-06,
+      "loss": 1.2695,
+      "step": 8324
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787901700293836e-06,
+      "loss": 1.1982,
+      "step": 8325
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.78562993779853e-06,
+      "loss": 1.21,
+      "step": 8326
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783358544845367e-06,
+      "loss": 1.2119,
+      "step": 8327
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7810875215953225e-06,
+      "loss": 1.2402,
+      "step": 8328
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778816868209337e-06,
+      "loss": 1.1777,
+      "step": 8329
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776546584848336e-06,
+      "loss": 1.2305,
+      "step": 8330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774276671673213e-06,
+      "loss": 1.2246,
+      "step": 8331
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772007128844838e-06,
+      "loss": 1.2412,
+      "step": 8332
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769737956524052e-06,
+      "loss": 1.2119,
+      "step": 8333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.767469154871673e-06,
+      "loss": 1.2168,
+      "step": 8334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.765200724048489e-06,
+      "loss": 1.207,
+      "step": 8335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762932664215269e-06,
+      "loss": 1.21,
+      "step": 8336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76066497553274e-06,
+      "loss": 1.2178,
+      "step": 8337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75839765816162e-06,
+      "loss": 1.2314,
+      "step": 8338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756130712262591e-06,
+      "loss": 1.168,
+      "step": 8339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753864137996313e-06,
+      "loss": 1.2227,
+      "step": 8340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751597935523419e-06,
+      "loss": 1.2334,
+      "step": 8341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749332105004512e-06,
+      "loss": 1.2578,
+      "step": 8342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.747066646600172e-06,
+      "loss": 1.2002,
+      "step": 8343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7448015604709575e-06,
+      "loss": 1.2207,
+      "step": 8344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.742536846777386e-06,
+      "loss": 1.1719,
+      "step": 8345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7402725056799615e-06,
+      "loss": 1.249,
+      "step": 8346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.738008537339157e-06,
+      "loss": 1.2334,
+      "step": 8347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735744941915421e-06,
+      "loss": 1.2451,
+      "step": 8348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733481719569175e-06,
+      "loss": 1.1855,
+      "step": 8349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.731218870460812e-06,
+      "loss": 1.2471,
+      "step": 8350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7289563947507015e-06,
+      "loss": 1.1982,
+      "step": 8351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726694292599187e-06,
+      "loss": 1.1943,
+      "step": 8352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724432564166576e-06,
+      "loss": 1.2158,
+      "step": 8353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722171209613162e-06,
+      "loss": 1.1846,
+      "step": 8354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719910229099204e-06,
+      "loss": 1.1934,
+      "step": 8355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717649622784948e-06,
+      "loss": 1.2275,
+      "step": 8356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71538939083059e-06,
+      "loss": 1.1963,
+      "step": 8357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71312953339632e-06,
+      "loss": 1.1914,
+      "step": 8358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710870050642291e-06,
+      "loss": 1.1973,
+      "step": 8359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7086109427286374e-06,
+      "loss": 1.1768,
+      "step": 8360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.70635220981545e-06,
+      "loss": 1.1943,
+      "step": 8361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704093852062817e-06,
+      "loss": 1.2109,
+      "step": 8362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701835869630785e-06,
+      "loss": 1.2021,
+      "step": 8363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69957826267938e-06,
+      "loss": 1.1953,
+      "step": 8364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69732103136859e-06,
+      "loss": 1.2002,
+      "step": 8365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69506417585839e-06,
+      "loss": 1.2734,
+      "step": 8366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692807696308723e-06,
+      "loss": 1.248,
+      "step": 8367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690551592879505e-06,
+      "loss": 1.2432,
+      "step": 8368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.688295865730628e-06,
+      "loss": 1.2393,
+      "step": 8369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686040515021952e-06,
+      "loss": 1.2129,
+      "step": 8370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683785540913317e-06,
+      "loss": 1.2021,
+      "step": 8371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681530943564534e-06,
+      "loss": 1.2266,
+      "step": 8372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679276723135379e-06,
+      "loss": 1.2031,
+      "step": 8373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6770228797856145e-06,
+      "loss": 1.2354,
+      "step": 8374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674769413674968e-06,
+      "loss": 1.2861,
+      "step": 8375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6725163249631435e-06,
+      "loss": 1.2246,
+      "step": 8376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670263613809816e-06,
+      "loss": 1.2197,
+      "step": 8377
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668011280374636e-06,
+      "loss": 1.2393,
+      "step": 8378
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665759324817227e-06,
+      "loss": 1.2197,
+      "step": 8379
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663507747297188e-06,
+      "loss": 1.1797,
+      "step": 8380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.66125654797408e-06,
+      "loss": 1.2598,
+      "step": 8381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65900572700745e-06,
+      "loss": 1.2197,
+      "step": 8382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656755284556814e-06,
+      "loss": 1.2139,
+      "step": 8383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65450522078166e-06,
+      "loss": 1.2627,
+      "step": 8384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.652255535841449e-06,
+      "loss": 1.2021,
+      "step": 8385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650006229895617e-06,
+      "loss": 1.2305,
+      "step": 8386
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.647757303103574e-06,
+      "loss": 1.2344,
+      "step": 8387
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645508755624702e-06,
+      "loss": 1.208,
+      "step": 8388
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.64326058761835e-06,
+      "loss": 1.2383,
+      "step": 8389
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641012799243848e-06,
+      "loss": 1.1904,
+      "step": 8390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638765390660494e-06,
+      "loss": 1.2402,
+      "step": 8391
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.636518362027571e-06,
+      "loss": 1.2119,
+      "step": 8392
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634271713504318e-06,
+      "loss": 1.2148,
+      "step": 8393
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.632025445249956e-06,
+      "loss": 1.2529,
+      "step": 8394
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.629779557423678e-06,
+      "loss": 1.2412,
+      "step": 8395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627534050184655e-06,
+      "loss": 1.2539,
+      "step": 8396
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6252889236920125e-06,
+      "loss": 1.207,
+      "step": 8397
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6230441781048775e-06,
+      "loss": 1.2256,
+      "step": 8398
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6207998135823275e-06,
+      "loss": 1.2188,
+      "step": 8399
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618555830283427e-06,
+      "loss": 1.1641,
+      "step": 8400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6163122283671965e-06,
+      "loss": 1.2227,
+      "step": 8401
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.614069007992647e-06,
+      "loss": 1.1992,
+      "step": 8402
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.611826169318753e-06,
+      "loss": 1.1992,
+      "step": 8403
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609583712504465e-06,
+      "loss": 1.2197,
+      "step": 8404
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607341637708707e-06,
+      "loss": 1.2568,
+      "step": 8405
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605099945090374e-06,
+      "loss": 1.2051,
+      "step": 8406
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602858634808333e-06,
+      "loss": 1.2422,
+      "step": 8407
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600617707021431e-06,
+      "loss": 1.2188,
+      "step": 8408
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598377161888474e-06,
+      "loss": 1.208,
+      "step": 8409
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596136999568255e-06,
+      "loss": 1.2246,
+      "step": 8410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593897220219533e-06,
+      "loss": 1.1768,
+      "step": 8411
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.59165782400104e-06,
+      "loss": 1.2295,
+      "step": 8412
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589418811071483e-06,
+      "loss": 1.2334,
+      "step": 8413
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587180181589541e-06,
+      "loss": 1.2402,
+      "step": 8414
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584941935713866e-06,
+      "loss": 1.2207,
+      "step": 8415
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582704073603083e-06,
+      "loss": 1.2061,
+      "step": 8416
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580466595415785e-06,
+      "loss": 1.1631,
+      "step": 8417
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578229501310546e-06,
+      "loss": 1.2246,
+      "step": 8418
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575992791445906e-06,
+      "loss": 1.2285,
+      "step": 8419
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5737564659803814e-06,
+      "loss": 1.2012,
+      "step": 8420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571520525072461e-06,
+      "loss": 1.2402,
+      "step": 8421
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569284968880608e-06,
+      "loss": 1.1992,
+      "step": 8422
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567049797563252e-06,
+      "loss": 1.2051,
+      "step": 8423
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5648150112788045e-06,
+      "loss": 1.1885,
+      "step": 8424
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56258061018564e-06,
+      "loss": 1.2031,
+      "step": 8425
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560346594442107e-06,
+      "loss": 1.2324,
+      "step": 8426
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558112964206539e-06,
+      "loss": 1.2168,
+      "step": 8427
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555879719637233e-06,
+      "loss": 1.2881,
+      "step": 8428
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553646860892451e-06,
+      "loss": 1.2275,
+      "step": 8429
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551414388130441e-06,
+      "loss": 1.1611,
+      "step": 8430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549182301509416e-06,
+      "loss": 1.1494,
+      "step": 8431
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54695060118757e-06,
+      "loss": 1.1973,
+      "step": 8432
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544719287323049e-06,
+      "loss": 1.2012,
+      "step": 8433
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542488360074001e-06,
+      "loss": 1.1973,
+      "step": 8434
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540257819598526e-06,
+      "loss": 1.2275,
+      "step": 8435
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538027666054706e-06,
+      "loss": 1.209,
+      "step": 8436
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535797899600584e-06,
+      "loss": 1.2256,
+      "step": 8437
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.533568520394189e-06,
+      "loss": 1.207,
+      "step": 8438
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5313395285935155e-06,
+      "loss": 1.2207,
+      "step": 8439
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529110924356532e-06,
+      "loss": 1.2305,
+      "step": 8440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.526882707841181e-06,
+      "loss": 1.2383,
+      "step": 8441
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.524654879205374e-06,
+      "loss": 1.251,
+      "step": 8442
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522427438606998e-06,
+      "loss": 1.1943,
+      "step": 8443
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520200386203915e-06,
+      "loss": 1.1973,
+      "step": 8444
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517973722153951e-06,
+      "loss": 1.2246,
+      "step": 8445
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51574744661491e-06,
+      "loss": 1.2275,
+      "step": 8446
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51352155974457e-06,
+      "loss": 1.2314,
+      "step": 8447
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511296061700679e-06,
+      "loss": 1.2607,
+      "step": 8448
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509070952640957e-06,
+      "loss": 1.2295,
+      "step": 8449
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506846232723099e-06,
+      "loss": 1.2607,
+      "step": 8450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50462190210477e-06,
+      "loss": 1.1445,
+      "step": 8451
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5023979609436105e-06,
+      "loss": 1.1768,
+      "step": 8452
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500174409397227e-06,
+      "loss": 1.2256,
+      "step": 8453
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4979512476232035e-06,
+      "loss": 1.2227,
+      "step": 8454
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495728475779097e-06,
+      "loss": 1.1953,
+      "step": 8455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493506094022429e-06,
+      "loss": 1.1982,
+      "step": 8456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491284102510713e-06,
+      "loss": 1.2119,
+      "step": 8457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48906250140141e-06,
+      "loss": 1.2529,
+      "step": 8458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486841290851969e-06,
+      "loss": 1.2412,
+      "step": 8459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.484620471019805e-06,
+      "loss": 1.2363,
+      "step": 8460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482400042062312e-06,
+      "loss": 1.2227,
+      "step": 8461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480180004136841e-06,
+      "loss": 1.2305,
+      "step": 8462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.477960357400738e-06,
+      "loss": 1.1748,
+      "step": 8463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.475741102011304e-06,
+      "loss": 1.2471,
+      "step": 8464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4735222381258225e-06,
+      "loss": 1.1689,
+      "step": 8465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471303765901534e-06,
+      "loss": 1.25,
+      "step": 8466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46908568549567e-06,
+      "loss": 1.2432,
+      "step": 8467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.466867997065422e-06,
+      "loss": 1.2461,
+      "step": 8468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4646507007679574e-06,
+      "loss": 1.249,
+      "step": 8469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.462433796760418e-06,
+      "loss": 1.1982,
+      "step": 8470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460217285199915e-06,
+      "loss": 1.2207,
+      "step": 8471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458001166243532e-06,
+      "loss": 1.2266,
+      "step": 8472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.455785440048328e-06,
+      "loss": 1.2764,
+      "step": 8473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453570106771327e-06,
+      "loss": 1.2578,
+      "step": 8474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45135516656953e-06,
+      "loss": 1.2334,
+      "step": 8475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449140619599911e-06,
+      "loss": 1.1426,
+      "step": 8476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4469264660194155e-06,
+      "loss": 1.2188,
+      "step": 8477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444712705984959e-06,
+      "loss": 1.1904,
+      "step": 8478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442499339653432e-06,
+      "loss": 1.1914,
+      "step": 8479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440286367181695e-06,
+      "loss": 1.2471,
+      "step": 8480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438073788726586e-06,
+      "loss": 1.2373,
+      "step": 8481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.435861604444901e-06,
+      "loss": 1.1992,
+      "step": 8482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433649814493421e-06,
+      "loss": 1.1904,
+      "step": 8483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431438419028898e-06,
+      "loss": 1.2109,
+      "step": 8484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429227418208051e-06,
+      "loss": 1.2627,
+      "step": 8485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.427016812187576e-06,
+      "loss": 1.2119,
+      "step": 8486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424806601124135e-06,
+      "loss": 1.1943,
+      "step": 8487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422596785174368e-06,
+      "loss": 1.1943,
+      "step": 8488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420387364494888e-06,
+      "loss": 1.1992,
+      "step": 8489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41817833924227e-06,
+      "loss": 1.2178,
+      "step": 8490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415969709573069e-06,
+      "loss": 1.1963,
+      "step": 8491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413761475643808e-06,
+      "loss": 1.2617,
+      "step": 8492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411553637610996e-06,
+      "loss": 1.1982,
+      "step": 8493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.40934619563109e-06,
+      "loss": 1.21,
+      "step": 8494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407139149860536e-06,
+      "loss": 1.2139,
+      "step": 8495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404932500455747e-06,
+      "loss": 1.1963,
+      "step": 8496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402726247573111e-06,
+      "loss": 1.1768,
+      "step": 8497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400520391368976e-06,
+      "loss": 1.2246,
+      "step": 8498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.398314931999678e-06,
+      "loss": 1.1777,
+      "step": 8499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.39610986962152e-06,
+      "loss": 1.209,
+      "step": 8500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.393905204390774e-06,
+      "loss": 1.207,
+      "step": 8501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3917009364636775e-06,
+      "loss": 1.2266,
+      "step": 8502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389497065996452e-06,
+      "loss": 1.2246,
+      "step": 8503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387293593145285e-06,
+      "loss": 1.2725,
+      "step": 8504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385090518066337e-06,
+      "loss": 1.2031,
+      "step": 8505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.382887840915738e-06,
+      "loss": 1.2197,
+      "step": 8506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.380685561849595e-06,
+      "loss": 1.2021,
+      "step": 8507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37848368102398e-06,
+      "loss": 1.1943,
+      "step": 8508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376282198594944e-06,
+      "loss": 1.1475,
+      "step": 8509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374081114718502e-06,
+      "loss": 1.207,
+      "step": 8510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371880429550646e-06,
+      "loss": 1.2393,
+      "step": 8511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3696801432473365e-06,
+      "loss": 1.2178,
+      "step": 8512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367480255964512e-06,
+      "loss": 1.2129,
+      "step": 8513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3652807678580755e-06,
+      "loss": 1.2021,
+      "step": 8514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3630816790839045e-06,
+      "loss": 1.208,
+      "step": 8515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.360882989797851e-06,
+      "loss": 1.1509,
+      "step": 8516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358684700155737e-06,
+      "loss": 1.1963,
+      "step": 8517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.356486810313349e-06,
+      "loss": 1.2471,
+      "step": 8518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3542893204264525e-06,
+      "loss": 1.2363,
+      "step": 8519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3520922306507875e-06,
+      "loss": 1.1533,
+      "step": 8520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.34989554114206e-06,
+      "loss": 1.2295,
+      "step": 8521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347699252055948e-06,
+      "loss": 1.2178,
+      "step": 8522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.345503363548105e-06,
+      "loss": 1.1484,
+      "step": 8523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3433078757741494e-06,
+      "loss": 1.1846,
+      "step": 8524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3411127888896845e-06,
+      "loss": 1.2793,
+      "step": 8525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338918103050264e-06,
+      "loss": 1.2461,
+      "step": 8526
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33672381841143e-06,
+      "loss": 1.1943,
+      "step": 8527
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334529935128689e-06,
+      "loss": 1.249,
+      "step": 8528
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.332336453357532e-06,
+      "loss": 1.2031,
+      "step": 8529
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.330143373253399e-06,
+      "loss": 1.2148,
+      "step": 8530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327950694971719e-06,
+      "loss": 1.1807,
+      "step": 8531
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3257584186678845e-06,
+      "loss": 1.2256,
+      "step": 8532
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323566544497266e-06,
+      "loss": 1.2568,
+      "step": 8533
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321375072615194e-06,
+      "loss": 1.1914,
+      "step": 8534
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319184003176984e-06,
+      "loss": 1.1982,
+      "step": 8535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.316993336337917e-06,
+      "loss": 1.207,
+      "step": 8536
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.314803072253247e-06,
+      "loss": 1.2217,
+      "step": 8537
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312613211078192e-06,
+      "loss": 1.2197,
+      "step": 8538
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3104237529679495e-06,
+      "loss": 1.2295,
+      "step": 8539
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308234698077688e-06,
+      "loss": 1.2676,
+      "step": 8540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3060460465625445e-06,
+      "loss": 1.1758,
+      "step": 8541
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.30385779857763e-06,
+      "loss": 1.2168,
+      "step": 8542
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.301669954278022e-06,
+      "loss": 1.2354,
+      "step": 8543
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299482513818778e-06,
+      "loss": 1.2139,
+      "step": 8544
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297295477354923e-06,
+      "loss": 1.2002,
+      "step": 8545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.295108845041443e-06,
+      "loss": 1.1846,
+      "step": 8546
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2929226170333115e-06,
+      "loss": 1.1504,
+      "step": 8547
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290736793485464e-06,
+      "loss": 1.1982,
+      "step": 8548
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288551374552811e-06,
+      "loss": 1.2275,
+      "step": 8549
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2863663603902316e-06,
+      "loss": 1.25,
+      "step": 8550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284181751152579e-06,
+      "loss": 1.1953,
+      "step": 8551
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281997546994676e-06,
+      "loss": 1.168,
+      "step": 8552
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.279813748071321e-06,
+      "loss": 1.166,
+      "step": 8553
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277630354537272e-06,
+      "loss": 1.1904,
+      "step": 8554
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2754473665472695e-06,
+      "loss": 1.2529,
+      "step": 8555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.273264784256021e-06,
+      "loss": 1.1973,
+      "step": 8556
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271082607818209e-06,
+      "loss": 1.2188,
+      "step": 8557
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26890083738848e-06,
+      "loss": 1.2207,
+      "step": 8558
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2667194731214586e-06,
+      "loss": 1.1963,
+      "step": 8559
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264538515171739e-06,
+      "loss": 1.2334,
+      "step": 8560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262357963693886e-06,
+      "loss": 1.2012,
+      "step": 8561
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.26017781884243e-06,
+      "loss": 1.2412,
+      "step": 8562
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.257998080771882e-06,
+      "loss": 1.1934,
+      "step": 8563
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.255818749636714e-06,
+      "loss": 1.2207,
+      "step": 8564
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253639825591388e-06,
+      "loss": 1.2432,
+      "step": 8565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251461308790312e-06,
+      "loss": 1.2705,
+      "step": 8566
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2492831993878825e-06,
+      "loss": 1.2256,
+      "step": 8567
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247105497538461e-06,
+      "loss": 1.1885,
+      "step": 8568
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244928203396385e-06,
+      "loss": 1.2129,
+      "step": 8569
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2427513171159476e-06,
+      "loss": 1.2314,
+      "step": 8570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240574838851436e-06,
+      "loss": 1.166,
+      "step": 8571
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238398768757094e-06,
+      "loss": 1.1963,
+      "step": 8572
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2362231069871415e-06,
+      "loss": 1.1943,
+      "step": 8573
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.234047853695763e-06,
+      "loss": 1.165,
+      "step": 8574
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231873009037121e-06,
+      "loss": 1.2148,
+      "step": 8575
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229698573165346e-06,
+      "loss": 1.2139,
+      "step": 8576
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2275245462345405e-06,
+      "loss": 1.1885,
+      "step": 8577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225350928398778e-06,
+      "loss": 1.1792,
+      "step": 8578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.223177719812102e-06,
+      "loss": 1.2178,
+      "step": 8579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221004920628529e-06,
+      "loss": 1.2148,
+      "step": 8580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218832531002046e-06,
+      "loss": 1.2422,
+      "step": 8581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216660551086607e-06,
+      "loss": 1.2246,
+      "step": 8582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21448898103614e-06,
+      "loss": 1.2451,
+      "step": 8583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2123178210045465e-06,
+      "loss": 1.1895,
+      "step": 8584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210147071145696e-06,
+      "loss": 1.1914,
+      "step": 8585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.207976731613428e-06,
+      "loss": 1.1738,
+      "step": 8586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.205806802561556e-06,
+      "loss": 1.2305,
+      "step": 8587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203637284143862e-06,
+      "loss": 1.1367,
+      "step": 8588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2014681765141e-06,
+      "loss": 1.2471,
+      "step": 8589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199299479825998e-06,
+      "loss": 1.1865,
+      "step": 8590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1971311942332435e-06,
+      "loss": 1.2217,
+      "step": 8591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1949633198895075e-06,
+      "loss": 1.2236,
+      "step": 8592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192795856948424e-06,
+      "loss": 1.2412,
+      "step": 8593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19062880556361e-06,
+      "loss": 1.1709,
+      "step": 8594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.188462165888635e-06,
+      "loss": 1.2139,
+      "step": 8595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.186295938077051e-06,
+      "loss": 1.2285,
+      "step": 8596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18413012228238e-06,
+      "loss": 1.2295,
+      "step": 8597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181964718658115e-06,
+      "loss": 1.2324,
+      "step": 8598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1797997273577085e-06,
+      "loss": 1.1934,
+      "step": 8599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177635148534604e-06,
+      "loss": 1.2002,
+      "step": 8600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.175470982342201e-06,
+      "loss": 1.1934,
+      "step": 8601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173307228933878e-06,
+      "loss": 1.2041,
+      "step": 8602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171143888462973e-06,
+      "loss": 1.2197,
+      "step": 8603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.168980961082804e-06,
+      "loss": 1.2275,
+      "step": 8604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166818446946661e-06,
+      "loss": 1.2031,
+      "step": 8605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1646563462077975e-06,
+      "loss": 1.2158,
+      "step": 8606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1624946590194425e-06,
+      "loss": 1.21,
+      "step": 8607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160333385534796e-06,
+      "loss": 1.1855,
+      "step": 8608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.158172525907026e-06,
+      "loss": 1.1758,
+      "step": 8609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1560120802892765e-06,
+      "loss": 1.2275,
+      "step": 8610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.153852048834651e-06,
+      "loss": 1.2012,
+      "step": 8611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1516924316962346e-06,
+      "loss": 1.2363,
+      "step": 8612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14953322902708e-06,
+      "loss": 1.1631,
+      "step": 8613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.147374440980208e-06,
+      "loss": 1.2412,
+      "step": 8614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145216067708611e-06,
+      "loss": 1.1465,
+      "step": 8615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1430581093652564e-06,
+      "loss": 1.2256,
+      "step": 8616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1409005661030765e-06,
+      "loss": 1.2158,
+      "step": 8617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1387434380749795e-06,
+      "loss": 1.1748,
+      "step": 8618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1365867254338345e-06,
+      "loss": 1.2207,
+      "step": 8619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134430428332491e-06,
+      "loss": 1.2139,
+      "step": 8620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.132274546923766e-06,
+      "loss": 1.2432,
+      "step": 8621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130119081360446e-06,
+      "loss": 1.1738,
+      "step": 8622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12796403179529e-06,
+      "loss": 1.2363,
+      "step": 8623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125809398381024e-06,
+      "loss": 1.2432,
+      "step": 8624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12365518127035e-06,
+      "loss": 1.207,
+      "step": 8625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.121501380615937e-06,
+      "loss": 1.2236,
+      "step": 8626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1193479965704205e-06,
+      "loss": 1.2627,
+      "step": 8627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.117195029286413e-06,
+      "loss": 1.208,
+      "step": 8628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115042478916492e-06,
+      "loss": 1.2217,
+      "step": 8629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11289034561322e-06,
+      "loss": 1.1924,
+      "step": 8630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1107386295291065e-06,
+      "loss": 1.2065,
+      "step": 8631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108587330816648e-06,
+      "loss": 1.2031,
+      "step": 8632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106436449628307e-06,
+      "loss": 1.2314,
+      "step": 8633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.104285986116518e-06,
+      "loss": 1.1929,
+      "step": 8634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.102135940433678e-06,
+      "loss": 1.2168,
+      "step": 8635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099986312732166e-06,
+      "loss": 1.2012,
+      "step": 8636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097837103164326e-06,
+      "loss": 1.2393,
+      "step": 8637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095688311882475e-06,
+      "loss": 1.2168,
+      "step": 8638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.09353993903889e-06,
+      "loss": 1.1943,
+      "step": 8639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.09139198478583e-06,
+      "loss": 1.21,
+      "step": 8640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0892444492755215e-06,
+      "loss": 1.2373,
+      "step": 8641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087097332660158e-06,
+      "loss": 1.2314,
+      "step": 8642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.084950635091907e-06,
+      "loss": 1.209,
+      "step": 8643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.082804356722905e-06,
+      "loss": 1.2031,
+      "step": 8644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080658497705258e-06,
+      "loss": 1.1973,
+      "step": 8645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.078513058191047e-06,
+      "loss": 1.248,
+      "step": 8646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076368038332311e-06,
+      "loss": 1.1836,
+      "step": 8647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.074223438281073e-06,
+      "loss": 1.2178,
+      "step": 8648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072079258189319e-06,
+      "loss": 1.2441,
+      "step": 8649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069935498209007e-06,
+      "loss": 1.1855,
+      "step": 8650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067792158492067e-06,
+      "loss": 1.2471,
+      "step": 8651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0656492391903945e-06,
+      "loss": 1.1641,
+      "step": 8652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.06350674045586e-06,
+      "loss": 1.2393,
+      "step": 8653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061364662440306e-06,
+      "loss": 1.252,
+      "step": 8654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059223005295533e-06,
+      "loss": 1.2314,
+      "step": 8655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.057081769173324e-06,
+      "loss": 1.2305,
+      "step": 8656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.054940954225429e-06,
+      "loss": 1.2129,
+      "step": 8657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0528005606035685e-06,
+      "loss": 1.2051,
+      "step": 8658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.05066058845943e-06,
+      "loss": 1.2441,
+      "step": 8659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.048521037944674e-06,
+      "loss": 1.1914,
+      "step": 8660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.04638190921093e-06,
+      "loss": 1.2061,
+      "step": 8661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.044243202409802e-06,
+      "loss": 1.1553,
+      "step": 8662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.042104917692853e-06,
+      "loss": 1.1738,
+      "step": 8663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039967055211625e-06,
+      "loss": 1.2334,
+      "step": 8664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0378296151176275e-06,
+      "loss": 1.1846,
+      "step": 8665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0356925975623506e-06,
+      "loss": 1.2402,
+      "step": 8666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0335560026972336e-06,
+      "loss": 1.2217,
+      "step": 8667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031419830673699e-06,
+      "loss": 1.2354,
+      "step": 8668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.029284081643138e-06,
+      "loss": 1.2139,
+      "step": 8669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027148755756915e-06,
+      "loss": 1.2256,
+      "step": 8670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025013853166351e-06,
+      "loss": 1.2764,
+      "step": 8671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0228793740227555e-06,
+      "loss": 1.2061,
+      "step": 8672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020745318477396e-06,
+      "loss": 1.2383,
+      "step": 8673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018611686681515e-06,
+      "loss": 1.2305,
+      "step": 8674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.016478478786319e-06,
+      "loss": 1.1992,
+      "step": 8675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014345694942988e-06,
+      "loss": 1.1914,
+      "step": 8676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.012213335302675e-06,
+      "loss": 1.2266,
+      "step": 8677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010081400016499e-06,
+      "loss": 1.2412,
+      "step": 8678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007949889235551e-06,
+      "loss": 1.2217,
+      "step": 8679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0058188031108905e-06,
+      "loss": 1.2432,
+      "step": 8680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.003688141793547e-06,
+      "loss": 1.1982,
+      "step": 8681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0015579054345245e-06,
+      "loss": 1.2549,
+      "step": 8682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999428094184786e-06,
+      "loss": 1.207,
+      "step": 8683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997298708195274e-06,
+      "loss": 1.2285,
+      "step": 8684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.995169747616898e-06,
+      "loss": 1.2227,
+      "step": 8685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.993041212600538e-06,
+      "loss": 1.2227,
+      "step": 8686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9909131032970426e-06,
+      "loss": 1.2197,
+      "step": 8687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988785419857231e-06,
+      "loss": 1.2139,
+      "step": 8688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986658162431892e-06,
+      "loss": 1.2197,
+      "step": 8689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984531331171788e-06,
+      "loss": 1.2246,
+      "step": 8690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98240492622764e-06,
+      "loss": 1.2402,
+      "step": 8691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980278947750149e-06,
+      "loss": 1.207,
+      "step": 8692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978153395889984e-06,
+      "loss": 1.2461,
+      "step": 8693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.976028270797784e-06,
+      "loss": 1.1875,
+      "step": 8694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.973903572624154e-06,
+      "loss": 1.2021,
+      "step": 8695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971779301519672e-06,
+      "loss": 1.2412,
+      "step": 8696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.969655457634886e-06,
+      "loss": 1.1816,
+      "step": 8697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9675320411203145e-06,
+      "loss": 1.2666,
+      "step": 8698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965409052126438e-06,
+      "loss": 1.2012,
+      "step": 8699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963286490803717e-06,
+      "loss": 1.2217,
+      "step": 8700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9611643573025725e-06,
+      "loss": 1.249,
+      "step": 8701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95904265177341e-06,
+      "loss": 1.1953,
+      "step": 8702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.956921374366586e-06,
+      "loss": 1.2266,
+      "step": 8703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.954800525232438e-06,
+      "loss": 1.2178,
+      "step": 8704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952680104521269e-06,
+      "loss": 1.2051,
+      "step": 8705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9505601123833596e-06,
+      "loss": 1.1797,
+      "step": 8706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948440548968941e-06,
+      "loss": 1.1885,
+      "step": 8707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946321414428237e-06,
+      "loss": 1.2217,
+      "step": 8708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9442027089114265e-06,
+      "loss": 1.1748,
+      "step": 8709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942084432568668e-06,
+      "loss": 1.1777,
+      "step": 8710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.939966585550075e-06,
+      "loss": 1.2109,
+      "step": 8711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9378491680057415e-06,
+      "loss": 1.2881,
+      "step": 8712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935732180085732e-06,
+      "loss": 1.2139,
+      "step": 8713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933615621940074e-06,
+      "loss": 1.2119,
+      "step": 8714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.93149949371877e-06,
+      "loss": 1.2852,
+      "step": 8715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929383795571789e-06,
+      "loss": 1.252,
+      "step": 8716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9272685276490715e-06,
+      "loss": 1.1826,
+      "step": 8717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925153690100528e-06,
+      "loss": 1.2275,
+      "step": 8718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923039283076032e-06,
+      "loss": 1.2129,
+      "step": 8719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920925306725433e-06,
+      "loss": 1.1973,
+      "step": 8720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91881176119855e-06,
+      "loss": 1.2207,
+      "step": 8721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9166986466451686e-06,
+      "loss": 1.1982,
+      "step": 8722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914585963215046e-06,
+      "loss": 1.2441,
+      "step": 8723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912473711057908e-06,
+      "loss": 1.1924,
+      "step": 8724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.910361890323448e-06,
+      "loss": 1.1885,
+      "step": 8725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.908250501161334e-06,
+      "loss": 1.21,
+      "step": 8726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906139543721202e-06,
+      "loss": 1.1787,
+      "step": 8727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904029018152647e-06,
+      "loss": 1.2686,
+      "step": 8728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901918924605246e-06,
+      "loss": 1.1973,
+      "step": 8729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.899809263228538e-06,
+      "loss": 1.1885,
+      "step": 8730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.897700034172046e-06,
+      "loss": 1.123,
+      "step": 8731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89559123758524e-06,
+      "loss": 1.1895,
+      "step": 8732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8934828736175735e-06,
+      "loss": 1.2139,
+      "step": 8733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891374942418465e-06,
+      "loss": 1.2314,
+      "step": 8734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889267444137309e-06,
+      "loss": 1.2197,
+      "step": 8735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8871603789234555e-06,
+      "loss": 1.2246,
+      "step": 8736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.885053746926234e-06,
+      "loss": 1.2197,
+      "step": 8737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882947548294945e-06,
+      "loss": 1.207,
+      "step": 8738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8808417831788584e-06,
+      "loss": 1.2227,
+      "step": 8739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8787364517272e-06,
+      "loss": 1.1992,
+      "step": 8740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876631554089178e-06,
+      "loss": 1.2061,
+      "step": 8741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.87452709041397e-06,
+      "loss": 1.2295,
+      "step": 8742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.872423060850714e-06,
+      "loss": 1.1611,
+      "step": 8743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8703194655485265e-06,
+      "loss": 1.207,
+      "step": 8744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.868216304656487e-06,
+      "loss": 1.1631,
+      "step": 8745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866113578323647e-06,
+      "loss": 1.2275,
+      "step": 8746
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864011286699031e-06,
+      "loss": 1.2324,
+      "step": 8747
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.86190942993162e-06,
+      "loss": 1.1875,
+      "step": 8748
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859808008170377e-06,
+      "loss": 1.2617,
+      "step": 8749
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8577070215642284e-06,
+      "loss": 1.2314,
+      "step": 8750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855606470262072e-06,
+      "loss": 1.251,
+      "step": 8751
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853506354412774e-06,
+      "loss": 1.2305,
+      "step": 8752
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851406674165169e-06,
+      "loss": 1.1484,
+      "step": 8753
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849307429668061e-06,
+      "loss": 1.2373,
+      "step": 8754
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8472086210702275e-06,
+      "loss": 1.2676,
+      "step": 8755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845110248520403e-06,
+      "loss": 1.1963,
+      "step": 8756
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8430123121673035e-06,
+      "loss": 1.2002,
+      "step": 8757
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.84091481215961e-06,
+      "loss": 1.2354,
+      "step": 8758
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83881774864597e-06,
+      "loss": 1.2305,
+      "step": 8759
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8367211217750046e-06,
+      "loss": 1.209,
+      "step": 8760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834624931695301e-06,
+      "loss": 1.2188,
+      "step": 8761
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.832529178555416e-06,
+      "loss": 1.2061,
+      "step": 8762
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.830433862503879e-06,
+      "loss": 1.2002,
+      "step": 8763
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828338983689177e-06,
+      "loss": 1.2227,
+      "step": 8764
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826244542259779e-06,
+      "loss": 1.209,
+      "step": 8765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824150538364115e-06,
+      "loss": 1.2627,
+      "step": 8766
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822056972150595e-06,
+      "loss": 1.2207,
+      "step": 8767
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8199638437675805e-06,
+      "loss": 1.2207,
+      "step": 8768
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817871153363415e-06,
+      "loss": 1.2139,
+      "step": 8769
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815778901086408e-06,
+      "loss": 1.1826,
+      "step": 8770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8136870870848395e-06,
+      "loss": 1.2305,
+      "step": 8771
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.811595711506948e-06,
+      "loss": 1.2158,
+      "step": 8772
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.809504774500957e-06,
+      "loss": 1.2256,
+      "step": 8773
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.807414276215049e-06,
+      "loss": 1.2119,
+      "step": 8774
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8053242167973815e-06,
+      "loss": 1.207,
+      "step": 8775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8032345963960684e-06,
+      "loss": 1.1777,
+      "step": 8776
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8011454151592054e-06,
+      "loss": 1.166,
+      "step": 8777
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7990566732348535e-06,
+      "loss": 1.1973,
+      "step": 8778
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7969683707710393e-06,
+      "loss": 1.1812,
+      "step": 8779
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7948805079157637e-06,
+      "loss": 1.2031,
+      "step": 8780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.79279308481699e-06,
+      "loss": 1.2578,
+      "step": 8781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7907061016226577e-06,
+      "loss": 1.1943,
+      "step": 8782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7886195584806718e-06,
+      "loss": 1.1895,
+      "step": 8783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.786533455538899e-06,
+      "loss": 1.2441,
+      "step": 8784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.784447792945186e-06,
+      "loss": 1.1934,
+      "step": 8785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.782362570847342e-06,
+      "loss": 1.2266,
+      "step": 8786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7802777893931485e-06,
+      "loss": 1.1797,
+      "step": 8787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7781934487303517e-06,
+      "loss": 1.1885,
+      "step": 8788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7761095490066703e-06,
+      "loss": 1.2285,
+      "step": 8789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.774026090369789e-06,
+      "loss": 1.1982,
+      "step": 8790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7719430729673678e-06,
+      "loss": 1.1875,
+      "step": 8791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769860496947021e-06,
+      "loss": 1.2461,
+      "step": 8792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7677783624563447e-06,
+      "loss": 1.2373,
+      "step": 8793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7656966696429e-06,
+      "loss": 1.2061,
+      "step": 8794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.763615418654216e-06,
+      "loss": 1.1855,
+      "step": 8795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7615346096377925e-06,
+      "loss": 1.1973,
+      "step": 8796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.759454242741094e-06,
+      "loss": 1.2373,
+      "step": 8797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7573743181115572e-06,
+      "loss": 1.1924,
+      "step": 8798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.755294835896589e-06,
+      "loss": 1.21,
+      "step": 8799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7532157962435555e-06,
+      "loss": 1.1777,
+      "step": 8800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7511371992998026e-06,
+      "loss": 1.2227,
+      "step": 8801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7490590452126364e-06,
+      "loss": 1.2061,
+      "step": 8802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746981334129345e-06,
+      "loss": 1.1855,
+      "step": 8803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.744904066197166e-06,
+      "loss": 1.207,
+      "step": 8804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7428272415633193e-06,
+      "loss": 1.1797,
+      "step": 8805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7407508603749877e-06,
+      "loss": 1.1953,
+      "step": 8806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7386749227793294e-06,
+      "loss": 1.2373,
+      "step": 8807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.736599428923455e-06,
+      "loss": 1.292,
+      "step": 8808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.734524378954465e-06,
+      "loss": 1.2158,
+      "step": 8809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732449773019414e-06,
+      "loss": 1.2178,
+      "step": 8810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7303756112653344e-06,
+      "loss": 1.21,
+      "step": 8811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7283018938392135e-06,
+      "loss": 1.1953,
+      "step": 8812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7262286208880192e-06,
+      "loss": 1.2061,
+      "step": 8813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7241557925586847e-06,
+      "loss": 1.1836,
+      "step": 8814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7220834089981117e-06,
+      "loss": 1.248,
+      "step": 8815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7200114703531685e-06,
+      "loss": 1.2373,
+      "step": 8816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7179399767706934e-06,
+      "loss": 1.166,
+      "step": 8817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7158689283974946e-06,
+      "loss": 1.2002,
+      "step": 8818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.71379832538035e-06,
+      "loss": 1.2363,
+      "step": 8819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7117281678659957e-06,
+      "loss": 1.1797,
+      "step": 8820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7096584560011464e-06,
+      "loss": 1.2002,
+      "step": 8821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7075891899324845e-06,
+      "loss": 1.1963,
+      "step": 8822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7055203698066566e-06,
+      "loss": 1.2461,
+      "step": 8823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.703451995770282e-06,
+      "loss": 1.248,
+      "step": 8824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7013840679699452e-06,
+      "loss": 1.2285,
+      "step": 8825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6993165865521995e-06,
+      "loss": 1.1797,
+      "step": 8826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.697249551663572e-06,
+      "loss": 1.1973,
+      "step": 8827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6951829634505452e-06,
+      "loss": 1.1865,
+      "step": 8828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6931168220595825e-06,
+      "loss": 1.2227,
+      "step": 8829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.691051127637112e-06,
+      "loss": 1.1953,
+      "step": 8830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6889858803295265e-06,
+      "loss": 1.2324,
+      "step": 8831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6869210802831924e-06,
+      "loss": 1.1943,
+      "step": 8832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.684856727644441e-06,
+      "loss": 1.2627,
+      "step": 8833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6827928225595734e-06,
+      "loss": 1.2139,
+      "step": 8834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.680729365174862e-06,
+      "loss": 1.165,
+      "step": 8835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6786663556365364e-06,
+      "loss": 1.2393,
+      "step": 8836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6766037940908063e-06,
+      "loss": 1.2129,
+      "step": 8837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6745416806838408e-06,
+      "loss": 1.2422,
+      "step": 8838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6724800155617925e-06,
+      "loss": 1.2295,
+      "step": 8839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6704187988707607e-06,
+      "loss": 1.2393,
+      "step": 8840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6683580307568277e-06,
+      "loss": 1.2119,
+      "step": 8841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6662977113660413e-06,
+      "loss": 1.2031,
+      "step": 8842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6642378408444178e-06,
+      "loss": 1.2119,
+      "step": 8843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6621784193379306e-06,
+      "loss": 1.2285,
+      "step": 8844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6601194469925404e-06,
+      "loss": 1.2314,
+      "step": 8845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6580609239541644e-06,
+      "loss": 1.1729,
+      "step": 8846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6560028503686916e-06,
+      "loss": 1.2568,
+      "step": 8847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.653945226381973e-06,
+      "loss": 1.1895,
+      "step": 8848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6518880521398337e-06,
+      "loss": 1.207,
+      "step": 8849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.649831327788067e-06,
+      "loss": 1.2627,
+      "step": 8850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.647775053472431e-06,
+      "loss": 1.2783,
+      "step": 8851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6457192293386555e-06,
+      "loss": 1.1885,
+      "step": 8852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6436638555324367e-06,
+      "loss": 1.1602,
+      "step": 8853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6416089321994375e-06,
+      "loss": 1.2676,
+      "step": 8854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6395544594852918e-06,
+      "loss": 1.2217,
+      "step": 8855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6375004375356015e-06,
+      "loss": 1.2158,
+      "step": 8856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6354468664959297e-06,
+      "loss": 1.2148,
+      "step": 8857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6333937465118165e-06,
+      "loss": 1.1611,
+      "step": 8858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.631341077728764e-06,
+      "loss": 1.1816,
+      "step": 8859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6292888602922483e-06,
+      "loss": 1.21,
+      "step": 8860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6272370943477065e-06,
+      "loss": 1.2188,
+      "step": 8861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6251857800405497e-06,
+      "loss": 1.207,
+      "step": 8862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6231349175161535e-06,
+      "loss": 1.2041,
+      "step": 8863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6210845069198654e-06,
+      "loss": 1.2031,
+      "step": 8864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.619034548396991e-06,
+      "loss": 1.2646,
+      "step": 8865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6169850420928153e-06,
+      "loss": 1.2666,
+      "step": 8866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6149359881525837e-06,
+      "loss": 1.2344,
+      "step": 8867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.612887386721521e-06,
+      "loss": 1.2275,
+      "step": 8868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.610839237944801e-06,
+      "loss": 1.2266,
+      "step": 8869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.608791541967581e-06,
+      "loss": 1.1855,
+      "step": 8870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606744298934981e-06,
+      "loss": 1.2266,
+      "step": 8871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604697508992091e-06,
+      "loss": 1.2344,
+      "step": 8872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6026511722839608e-06,
+      "loss": 1.2314,
+      "step": 8873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6006052889556154e-06,
+      "loss": 1.2012,
+      "step": 8874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.598559859152051e-06,
+      "loss": 1.2109,
+      "step": 8875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5965148830182295e-06,
+      "loss": 1.2227,
+      "step": 8876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5944703606990695e-06,
+      "loss": 1.1914,
+      "step": 8877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5924262923394694e-06,
+      "loss": 1.2461,
+      "step": 8878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.590382678084294e-06,
+      "loss": 1.1885,
+      "step": 8879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588339518078373e-06,
+      "loss": 1.2031,
+      "step": 8880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5862968124665065e-06,
+      "loss": 1.165,
+      "step": 8881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5842545613934586e-06,
+      "loss": 1.2139,
+      "step": 8882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.582212765003965e-06,
+      "loss": 1.2002,
+      "step": 8883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58017142344273e-06,
+      "loss": 1.1914,
+      "step": 8884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5781305368544186e-06,
+      "loss": 1.207,
+      "step": 8885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5760901053836694e-06,
+      "loss": 1.2559,
+      "step": 8886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5740501291750894e-06,
+      "loss": 1.2617,
+      "step": 8887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.572010608373251e-06,
+      "loss": 1.1797,
+      "step": 8888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5699715431226943e-06,
+      "loss": 1.1992,
+      "step": 8889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5679329335679293e-06,
+      "loss": 1.2002,
+      "step": 8890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5658947798534304e-06,
+      "loss": 1.1699,
+      "step": 8891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5638570821236463e-06,
+      "loss": 1.2061,
+      "step": 8892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.561819840522981e-06,
+      "loss": 1.2363,
+      "step": 8893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.559783055195817e-06,
+      "loss": 1.2324,
+      "step": 8894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.557746726286502e-06,
+      "loss": 1.1836,
+      "step": 8895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5557108539393494e-06,
+      "loss": 1.21,
+      "step": 8896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5536754382986414e-06,
+      "loss": 1.2275,
+      "step": 8897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.551640479508629e-06,
+      "loss": 1.1875,
+      "step": 8898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5496059777135285e-06,
+      "loss": 1.2051,
+      "step": 8899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5475719330575286e-06,
+      "loss": 1.2119,
+      "step": 8900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.545538345684776e-06,
+      "loss": 1.2285,
+      "step": 8901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5435052157393933e-06,
+      "loss": 1.1885,
+      "step": 8902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.541472543365465e-06,
+      "loss": 1.1758,
+      "step": 8903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5394403287070566e-06,
+      "loss": 1.2637,
+      "step": 8904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.537408571908182e-06,
+      "loss": 1.2334,
+      "step": 8905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.535377273112834e-06,
+      "loss": 1.2275,
+      "step": 8906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5333464324649715e-06,
+      "loss": 1.2324,
+      "step": 8907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.531316050108522e-06,
+      "loss": 1.1963,
+      "step": 8908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5292861261873723e-06,
+      "loss": 1.1992,
+      "step": 8909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5272566608453838e-06,
+      "loss": 1.2441,
+      "step": 8910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.525227654226392e-06,
+      "loss": 1.1875,
+      "step": 8911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.523199106474191e-06,
+      "loss": 1.1914,
+      "step": 8912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5211710177325374e-06,
+      "loss": 1.1758,
+      "step": 8913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5191433881451655e-06,
+      "loss": 1.1865,
+      "step": 8914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5171162178557727e-06,
+      "loss": 1.2129,
+      "step": 8915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.515089507008025e-06,
+      "loss": 1.2236,
+      "step": 8916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5130632557455547e-06,
+      "loss": 1.2188,
+      "step": 8917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.511037464211964e-06,
+      "loss": 1.1914,
+      "step": 8918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5090121325508176e-06,
+      "loss": 1.21,
+      "step": 8919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5069872609056565e-06,
+      "loss": 1.1914,
+      "step": 8920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.504962849419976e-06,
+      "loss": 1.2148,
+      "step": 8921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.502938898237249e-06,
+      "loss": 1.1729,
+      "step": 8922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5009154075009123e-06,
+      "loss": 1.2451,
+      "step": 8923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4988923773543713e-06,
+      "loss": 1.2188,
+      "step": 8924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.496869807940999e-06,
+      "loss": 1.1836,
+      "step": 8925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4948476994041324e-06,
+      "loss": 1.1963,
+      "step": 8926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4928260518870807e-06,
+      "loss": 1.1914,
+      "step": 8927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4908048655331194e-06,
+      "loss": 1.2227,
+      "step": 8928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4887841404854837e-06,
+      "loss": 1.2275,
+      "step": 8929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.486763876887387e-06,
+      "loss": 1.2168,
+      "step": 8930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4847440748820026e-06,
+      "loss": 1.2061,
+      "step": 8931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.482724734612477e-06,
+      "loss": 1.2363,
+      "step": 8932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4807058562219176e-06,
+      "loss": 1.1934,
+      "step": 8933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4786874398534044e-06,
+      "loss": 1.208,
+      "step": 8934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4766694856499816e-06,
+      "loss": 1.2227,
+      "step": 8935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.474651993754665e-06,
+      "loss": 1.2578,
+      "step": 8936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.472634964310427e-06,
+      "loss": 1.2275,
+      "step": 8937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4706183974602193e-06,
+      "loss": 1.252,
+      "step": 8938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.468602293346951e-06,
+      "loss": 1.2119,
+      "step": 8939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4665866521135137e-06,
+      "loss": 1.2324,
+      "step": 8940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4645714739027457e-06,
+      "loss": 1.2148,
+      "step": 8941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.462556758857466e-06,
+      "loss": 1.2217,
+      "step": 8942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.460542507120457e-06,
+      "loss": 1.2031,
+      "step": 8943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.458528718834473e-06,
+      "loss": 1.2168,
+      "step": 8944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.456515394142219e-06,
+      "loss": 1.1816,
+      "step": 8945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4545025331863925e-06,
+      "loss": 1.2051,
+      "step": 8946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4524901361096384e-06,
+      "loss": 1.1777,
+      "step": 8947
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4504782030545803e-06,
+      "loss": 1.1953,
+      "step": 8948
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4484667341637966e-06,
+      "loss": 1.2676,
+      "step": 8949
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.446455729579843e-06,
+      "loss": 1.168,
+      "step": 8950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4444451894452402e-06,
+      "loss": 1.2197,
+      "step": 8951
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4424351139024734e-06,
+      "loss": 1.2832,
+      "step": 8952
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.440425503093998e-06,
+      "loss": 1.2451,
+      "step": 8953
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4384163571622355e-06,
+      "loss": 1.2197,
+      "step": 8954
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4364076762495725e-06,
+      "loss": 1.1562,
+      "step": 8955
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4343994604983686e-06,
+      "loss": 1.2246,
+      "step": 8956
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4323917100509386e-06,
+      "loss": 1.1904,
+      "step": 8957
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4303844250495742e-06,
+      "loss": 1.2207,
+      "step": 8958
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4283776056365334e-06,
+      "loss": 1.2344,
+      "step": 8959
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.426371251954038e-06,
+      "loss": 1.2559,
+      "step": 8960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4243653641442797e-06,
+      "loss": 1.2422,
+      "step": 8961
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4223599423494136e-06,
+      "loss": 1.2021,
+      "step": 8962
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.420354986711566e-06,
+      "loss": 1.2402,
+      "step": 8963
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.418350497372829e-06,
+      "loss": 1.2617,
+      "step": 8964
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4163464744752562e-06,
+      "loss": 1.2246,
+      "step": 8965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4143429181608745e-06,
+      "loss": 1.1982,
+      "step": 8966
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.412339828571677e-06,
+      "loss": 1.1631,
+      "step": 8967
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.410337205849621e-06,
+      "loss": 1.2051,
+      "step": 8968
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.408335050136634e-06,
+      "loss": 1.2178,
+      "step": 8969
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4063333615746074e-06,
+      "loss": 1.2119,
+      "step": 8970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.404332140305401e-06,
+      "loss": 1.2246,
+      "step": 8971
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402331386470845e-06,
+      "loss": 1.2031,
+      "step": 8972
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.400331100212726e-06,
+      "loss": 1.2148,
+      "step": 8973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3983312816728063e-06,
+      "loss": 1.1953,
+      "step": 8974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3963319309928124e-06,
+      "loss": 1.1631,
+      "step": 8975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3943330483144456e-06,
+      "loss": 1.2432,
+      "step": 8976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3923346337793574e-06,
+      "loss": 1.2021,
+      "step": 8977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3903366875291776e-06,
+      "loss": 1.2334,
+      "step": 8978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3883392097055035e-06,
+      "loss": 1.1934,
+      "step": 8979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3863422004498967e-06,
+      "loss": 1.2676,
+      "step": 8980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.384345659903877e-06,
+      "loss": 1.1992,
+      "step": 8981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3823495882089475e-06,
+      "loss": 1.1963,
+      "step": 8982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.380353985506568e-06,
+      "loss": 1.2207,
+      "step": 8983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.378358851938165e-06,
+      "loss": 1.2012,
+      "step": 8984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.376364187645138e-06,
+      "loss": 1.2363,
+      "step": 8985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374369992768843e-06,
+      "loss": 1.1943,
+      "step": 8986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.37237626745061e-06,
+      "loss": 1.1963,
+      "step": 8987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.370383011831735e-06,
+      "loss": 1.209,
+      "step": 8988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3683902260534797e-06,
+      "loss": 1.2275,
+      "step": 8989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.366397910257073e-06,
+      "loss": 1.2061,
+      "step": 8990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3644060645837094e-06,
+      "loss": 1.208,
+      "step": 8991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3624146891745523e-06,
+      "loss": 1.1738,
+      "step": 8992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3604237841707334e-06,
+      "loss": 1.1816,
+      "step": 8993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3584333497133403e-06,
+      "loss": 1.249,
+      "step": 8994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3564433859434396e-06,
+      "loss": 1.248,
+      "step": 8995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.354453893002059e-06,
+      "loss": 1.2139,
+      "step": 8996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3524648710301944e-06,
+      "loss": 1.2305,
+      "step": 8997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.350476320168806e-06,
+      "loss": 1.2246,
+      "step": 8998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348488240558826e-06,
+      "loss": 1.2324,
+      "step": 8999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3465006323411453e-06,
+      "loss": 1.1904,
+      "step": 9000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3445134956566316e-06,
+      "loss": 1.2012,
+      "step": 9001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3425268306461056e-06,
+      "loss": 1.251,
+      "step": 9002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3405406374503657e-06,
+      "loss": 1.1865,
+      "step": 9003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3385549162101692e-06,
+      "loss": 1.2393,
+      "step": 9004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3365696670662563e-06,
+      "loss": 1.2041,
+      "step": 9005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.334584890159308e-06,
+      "loss": 1.1875,
+      "step": 9006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3326005856299926e-06,
+      "loss": 1.2236,
+      "step": 9007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3306167536189338e-06,
+      "loss": 1.21,
+      "step": 9008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.328633394266733e-06,
+      "loss": 1.2246,
+      "step": 9009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.32665050771394e-06,
+      "loss": 1.2285,
+      "step": 9010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3246680941010843e-06,
+      "loss": 1.2354,
+      "step": 9011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.322686153568666e-06,
+      "loss": 1.1797,
+      "step": 9012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3207046862571457e-06,
+      "loss": 1.1777,
+      "step": 9013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.318723692306941e-06,
+      "loss": 1.1846,
+      "step": 9014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316743171858451e-06,
+      "loss": 1.1709,
+      "step": 9015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.314763125052032e-06,
+      "loss": 1.2207,
+      "step": 9016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3127835520280117e-06,
+      "loss": 1.2373,
+      "step": 9017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3108044529266826e-06,
+      "loss": 1.1992,
+      "step": 9018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3088258278883035e-06,
+      "loss": 1.2266,
+      "step": 9019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.306847677053098e-06,
+      "loss": 1.2188,
+      "step": 9020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.304870000561261e-06,
+      "loss": 1.1943,
+      "step": 9021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3028927985529445e-06,
+      "loss": 1.2451,
+      "step": 9022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.300916071168276e-06,
+      "loss": 1.2236,
+      "step": 9023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2989398185473464e-06,
+      "loss": 1.2334,
+      "step": 9024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2969640408302117e-06,
+      "loss": 1.1885,
+      "step": 9025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2949887381568957e-06,
+      "loss": 1.2471,
+      "step": 9026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2930139106673886e-06,
+      "loss": 1.1621,
+      "step": 9027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.291039558501645e-06,
+      "loss": 1.2041,
+      "step": 9028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2890656817995914e-06,
+      "loss": 1.2061,
+      "step": 9029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2870922807011084e-06,
+      "loss": 1.2354,
+      "step": 9030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.285119355346056e-06,
+      "loss": 1.1826,
+      "step": 9031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.283146905874255e-06,
+      "loss": 1.2139,
+      "step": 9032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2811749324254915e-06,
+      "loss": 1.1602,
+      "step": 9033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2792034351395196e-06,
+      "loss": 1.1973,
+      "step": 9034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2772324141560597e-06,
+      "loss": 1.2383,
+      "step": 9035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.275261869614799e-06,
+      "loss": 1.2021,
+      "step": 9036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2732918016553903e-06,
+      "loss": 1.1982,
+      "step": 9037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.271322210417448e-06,
+      "loss": 1.2246,
+      "step": 9038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2693530960405594e-06,
+      "loss": 1.2207,
+      "step": 9039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2673844586642733e-06,
+      "loss": 1.2197,
+      "step": 9040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2654162984281147e-06,
+      "loss": 1.1963,
+      "step": 9041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2634486154715583e-06,
+      "loss": 1.2178,
+      "step": 9042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2614814099340574e-06,
+      "loss": 1.2637,
+      "step": 9043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2595146819550285e-06,
+      "loss": 1.1602,
+      "step": 9044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.257548431673854e-06,
+      "loss": 1.2197,
+      "step": 9045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2555826592298776e-06,
+      "loss": 1.21,
+      "step": 9046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2536173647624135e-06,
+      "loss": 1.1895,
+      "step": 9047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2516525484107474e-06,
+      "loss": 1.2578,
+      "step": 9048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2496882103141258e-06,
+      "loss": 1.1904,
+      "step": 9049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2477243506117563e-06,
+      "loss": 1.21,
+      "step": 9050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.245760969442819e-06,
+      "loss": 1.2207,
+      "step": 9051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2437980669464595e-06,
+      "loss": 1.1431,
+      "step": 9052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2418356432617893e-06,
+      "loss": 1.2197,
+      "step": 9053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2398736985278847e-06,
+      "loss": 1.2363,
+      "step": 9054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2379122328837877e-06,
+      "loss": 1.2549,
+      "step": 9055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.235951246468507e-06,
+      "loss": 1.2129,
+      "step": 9056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2339907394210222e-06,
+      "loss": 1.2217,
+      "step": 9057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2320307118802687e-06,
+      "loss": 1.1895,
+      "step": 9058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2300711639851547e-06,
+      "loss": 1.1855,
+      "step": 9059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.228112095874554e-06,
+      "loss": 1.208,
+      "step": 9060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2261535076873064e-06,
+      "loss": 1.2139,
+      "step": 9061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2241953995622155e-06,
+      "loss": 1.1934,
+      "step": 9062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2222377716380527e-06,
+      "loss": 1.1582,
+      "step": 9063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2202806240535565e-06,
+      "loss": 1.2207,
+      "step": 9064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.218323956947431e-06,
+      "loss": 1.1533,
+      "step": 9065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2163677704583405e-06,
+      "loss": 1.1846,
+      "step": 9066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2144120647249224e-06,
+      "loss": 1.1807,
+      "step": 9067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.212456839885777e-06,
+      "loss": 1.2168,
+      "step": 9068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2105020960794707e-06,
+      "loss": 1.1855,
+      "step": 9069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2085478334445373e-06,
+      "loss": 1.2354,
+      "step": 9070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.206594052119475e-06,
+      "loss": 1.1973,
+      "step": 9071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204640752242747e-06,
+      "loss": 1.1426,
+      "step": 9072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.202687933952787e-06,
+      "loss": 1.1914,
+      "step": 9073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2007355973879872e-06,
+      "loss": 1.1963,
+      "step": 9074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.19878374268671e-06,
+      "loss": 1.2295,
+      "step": 9075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1968323699872805e-06,
+      "loss": 1.2412,
+      "step": 9076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1948814794280027e-06,
+      "loss": 1.2598,
+      "step": 9077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1929310711471264e-06,
+      "loss": 1.2139,
+      "step": 9078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.19098114528288e-06,
+      "loss": 1.2119,
+      "step": 9079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.189031701973454e-06,
+      "loss": 1.1992,
+      "step": 9080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1870827413570103e-06,
+      "loss": 1.2373,
+      "step": 9081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185134263571663e-06,
+      "loss": 1.2715,
+      "step": 9082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1831862687555026e-06,
+      "loss": 1.1807,
+      "step": 9083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1812387570465885e-06,
+      "loss": 1.2266,
+      "step": 9084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1792917285829415e-06,
+      "loss": 1.2246,
+      "step": 9085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1773451835025414e-06,
+      "loss": 1.2266,
+      "step": 9086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1753991219433412e-06,
+      "loss": 1.25,
+      "step": 9087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1734535440432603e-06,
+      "loss": 1.1719,
+      "step": 9088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1715084499401796e-06,
+      "loss": 1.2705,
+      "step": 9089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1695638397719485e-06,
+      "loss": 1.2197,
+      "step": 9090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.167619713676383e-06,
+      "loss": 1.1758,
+      "step": 9091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1656760717912615e-06,
+      "loss": 1.1846,
+      "step": 9092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.163732914254334e-06,
+      "loss": 1.2461,
+      "step": 9093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.161790241203304e-06,
+      "loss": 1.2422,
+      "step": 9094
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1598480527758534e-06,
+      "loss": 1.2461,
+      "step": 9095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157906349109625e-06,
+      "loss": 1.2373,
+      "step": 9096
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1559651303422266e-06,
+      "loss": 1.1914,
+      "step": 9097
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1540243966112327e-06,
+      "loss": 1.2324,
+      "step": 9098
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152084148054182e-06,
+      "loss": 1.2246,
+      "step": 9099
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1501443848085812e-06,
+      "loss": 1.2158,
+      "step": 9100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1482051070119035e-06,
+      "loss": 1.1895,
+      "step": 9101
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1462663148015793e-06,
+      "loss": 1.1768,
+      "step": 9102
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.144328008315014e-06,
+      "loss": 1.2002,
+      "step": 9103
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1423901876895745e-06,
+      "loss": 1.2119,
+      "step": 9104
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1404528530625955e-06,
+      "loss": 1.1753,
+      "step": 9105
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.138516004571376e-06,
+      "loss": 1.21,
+      "step": 9106
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.136579642353178e-06,
+      "loss": 1.1738,
+      "step": 9107
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1346437665452336e-06,
+      "loss": 1.21,
+      "step": 9108
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13270837728474e-06,
+      "loss": 1.21,
+      "step": 9109
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1307734747088525e-06,
+      "loss": 1.2393,
+      "step": 9110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1288390589547e-06,
+      "loss": 1.2402,
+      "step": 9111
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.126905130159372e-06,
+      "loss": 1.25,
+      "step": 9112
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1249716884599367e-06,
+      "loss": 1.168,
+      "step": 9113
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1230387339934055e-06,
+      "loss": 1.2539,
+      "step": 9114
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.121106266896771e-06,
+      "loss": 1.1797,
+      "step": 9115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1191742873069852e-06,
+      "loss": 1.1592,
+      "step": 9116
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1172427953609698e-06,
+      "loss": 1.1963,
+      "step": 9117
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.115311791195609e-06,
+      "loss": 1.2236,
+      "step": 9118
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.113381274947751e-06,
+      "loss": 1.1836,
+      "step": 9119
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.111451246754215e-06,
+      "loss": 1.2549,
+      "step": 9120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109521706751778e-06,
+      "loss": 1.1748,
+      "step": 9121
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.107592655077193e-06,
+      "loss": 1.2441,
+      "step": 9122
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1056640918671623e-06,
+      "loss": 1.2148,
+      "step": 9123
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1037360172583697e-06,
+      "loss": 1.1982,
+      "step": 9124
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1018084313874543e-06,
+      "loss": 1.1953,
+      "step": 9125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0998813343910273e-06,
+      "loss": 1.1904,
+      "step": 9126
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0979547264056586e-06,
+      "loss": 1.2139,
+      "step": 9127
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.09602860756789e-06,
+      "loss": 1.2402,
+      "step": 9128
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0941029780142227e-06,
+      "loss": 1.2139,
+      "step": 9129
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0921778378811317e-06,
+      "loss": 1.2363,
+      "step": 9130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0902531873050432e-06,
+      "loss": 1.2432,
+      "step": 9131
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.088329026422361e-06,
+      "loss": 1.1895,
+      "step": 9132
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0864053553694506e-06,
+      "loss": 1.1885,
+      "step": 9133
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0844821742826424e-06,
+      "loss": 1.2129,
+      "step": 9134
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.082559483298232e-06,
+      "loss": 1.2422,
+      "step": 9135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.08063728255248e-06,
+      "loss": 1.1738,
+      "step": 9136
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.078715572181612e-06,
+      "loss": 1.1963,
+      "step": 9137
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.076794352321825e-06,
+      "loss": 1.2148,
+      "step": 9138
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.074873623109268e-06,
+      "loss": 1.2305,
+      "step": 9139
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.072953384680066e-06,
+      "loss": 1.1914,
+      "step": 9140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.071033637170303e-06,
+      "loss": 1.209,
+      "step": 9141
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.069114380716042e-06,
+      "loss": 1.2236,
+      "step": 9142
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0671956154532887e-06,
+      "loss": 1.2031,
+      "step": 9143
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0652773415180326e-06,
+      "loss": 1.1201,
+      "step": 9144
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0633595590462185e-06,
+      "loss": 1.2031,
+      "step": 9145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.061442268173764e-06,
+      "loss": 1.1689,
+      "step": 9146
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0595254690365415e-06,
+      "loss": 1.2363,
+      "step": 9147
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.057609161770394e-06,
+      "loss": 1.2314,
+      "step": 9148
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055693346511137e-06,
+      "loss": 1.1904,
+      "step": 9149
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0537780233945423e-06,
+      "loss": 1.1855,
+      "step": 9150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0518631925563446e-06,
+      "loss": 1.1826,
+      "step": 9151
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0499488541322496e-06,
+      "loss": 1.1904,
+      "step": 9152
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.048035008257928e-06,
+      "loss": 1.2061,
+      "step": 9153
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046121655069012e-06,
+      "loss": 1.2119,
+      "step": 9154
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0442087947011022e-06,
+      "loss": 1.2705,
+      "step": 9155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0422964272897627e-06,
+      "loss": 1.2197,
+      "step": 9156
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0403845529705223e-06,
+      "loss": 1.1797,
+      "step": 9157
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0384731718788797e-06,
+      "loss": 1.2256,
+      "step": 9158
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0365622841502864e-06,
+      "loss": 1.1143,
+      "step": 9159
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0346518899201706e-06,
+      "loss": 1.1943,
+      "step": 9160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.032741989323923e-06,
+      "loss": 1.1973,
+      "step": 9161
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0308325824968976e-06,
+      "loss": 1.1768,
+      "step": 9162
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.028923669574413e-06,
+      "loss": 1.1963,
+      "step": 9163
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.027015250691755e-06,
+      "loss": 1.1787,
+      "step": 9164
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0251073259841734e-06,
+      "loss": 1.2402,
+      "step": 9165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.023199895586885e-06,
+      "loss": 1.207,
+      "step": 9166
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0212929596350614e-06,
+      "loss": 1.2422,
+      "step": 9167
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0193865182638536e-06,
+      "loss": 1.2119,
+      "step": 9168
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.017480571608369e-06,
+      "loss": 1.2051,
+      "step": 9169
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.015575119803682e-06,
+      "loss": 1.2549,
+      "step": 9170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.013670162984833e-06,
+      "loss": 1.2129,
+      "step": 9171
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.011765701286825e-06,
+      "loss": 1.2227,
+      "step": 9172
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.009861734844627e-06,
+      "loss": 1.2021,
+      "step": 9173
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007958263793177e-06,
+      "loss": 1.2295,
+      "step": 9174
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0060552882673667e-06,
+      "loss": 1.1904,
+      "step": 9175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0041528084020644e-06,
+      "loss": 1.251,
+      "step": 9176
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.002250824332095e-06,
+      "loss": 1.2158,
+      "step": 9177
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0003493361922597e-06,
+      "loss": 1.2637,
+      "step": 9178
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9984483441173094e-06,
+      "loss": 1.209,
+      "step": 9179
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9965478482419706e-06,
+      "loss": 1.1914,
+      "step": 9180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9946478487009312e-06,
+      "loss": 1.1982,
+      "step": 9181
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9927483456288462e-06,
+      "loss": 1.2305,
+      "step": 9182
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.990849339160328e-06,
+      "loss": 1.2148,
+      "step": 9183
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9889508294299572e-06,
+      "loss": 1.1982,
+      "step": 9184
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.98705281657229e-06,
+      "loss": 1.2363,
+      "step": 9185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.985155300721837e-06,
+      "loss": 1.1768,
+      "step": 9186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9832582820130697e-06,
+      "loss": 1.1523,
+      "step": 9187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981361760580431e-06,
+      "loss": 1.2207,
+      "step": 9188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.979465736558329e-06,
+      "loss": 1.2334,
+      "step": 9189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9775702100811354e-06,
+      "loss": 1.2402,
+      "step": 9190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.975675181283184e-06,
+      "loss": 1.1992,
+      "step": 9191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9737806502987766e-06,
+      "loss": 1.1445,
+      "step": 9192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9718866172621786e-06,
+      "loss": 1.2041,
+      "step": 9193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9699930823076217e-06,
+      "loss": 1.2393,
+      "step": 9194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9681000455692964e-06,
+      "loss": 1.1699,
+      "step": 9195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9662075071813645e-06,
+      "loss": 1.2139,
+      "step": 9196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9643154672779505e-06,
+      "loss": 1.2588,
+      "step": 9197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.962423925993142e-06,
+      "loss": 1.2471,
+      "step": 9198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9605328834609924e-06,
+      "loss": 1.2432,
+      "step": 9199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.958642339815522e-06,
+      "loss": 1.2344,
+      "step": 9200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9567522951907112e-06,
+      "loss": 1.1797,
+      "step": 9201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9548627497205105e-06,
+      "loss": 1.165,
+      "step": 9202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9529737035388272e-06,
+      "loss": 1.1895,
+      "step": 9203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9510851567795405e-06,
+      "loss": 1.1836,
+      "step": 9204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9491971095764915e-06,
+      "loss": 1.2061,
+      "step": 9205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.947309562063485e-06,
+      "loss": 1.1943,
+      "step": 9206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.945422514374293e-06,
+      "loss": 1.1533,
+      "step": 9207
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.94353596664265e-06,
+      "loss": 1.249,
+      "step": 9208
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.941649919002255e-06,
+      "loss": 1.2568,
+      "step": 9209
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.939764371586775e-06,
+      "loss": 1.2295,
+      "step": 9210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.937879324529834e-06,
+      "loss": 1.2383,
+      "step": 9211
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9359947779650267e-06,
+      "loss": 1.2002,
+      "step": 9212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9341107320259086e-06,
+      "loss": 1.2217,
+      "step": 9213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.932227186846011e-06,
+      "loss": 1.2197,
+      "step": 9214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9303441425588096e-06,
+      "loss": 1.2139,
+      "step": 9215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9284615992977627e-06,
+      "loss": 1.21,
+      "step": 9216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.926579557196282e-06,
+      "loss": 1.1562,
+      "step": 9217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9246980163877536e-06,
+      "loss": 1.2559,
+      "step": 9218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.922816977005515e-06,
+      "loss": 1.1689,
+      "step": 9219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9209364391828744e-06,
+      "loss": 1.208,
+      "step": 9220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9190564030531133e-06,
+      "loss": 1.2295,
+      "step": 9221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9171768687494693e-06,
+      "loss": 1.1475,
+      "step": 9222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9152978364051376e-06,
+      "loss": 1.2148,
+      "step": 9223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.91341930615329e-06,
+      "loss": 1.2305,
+      "step": 9224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.911541278127057e-06,
+      "loss": 1.1816,
+      "step": 9225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.909663752459534e-06,
+      "loss": 1.2207,
+      "step": 9226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.90778672928378e-06,
+      "loss": 1.1748,
+      "step": 9227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.905910208732823e-06,
+      "loss": 1.2793,
+      "step": 9228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9040341909396486e-06,
+      "loss": 1.2324,
+      "step": 9229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9021586760372143e-06,
+      "loss": 1.251,
+      "step": 9230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9002836641584333e-06,
+      "loss": 1.1865,
+      "step": 9231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898409155436188e-06,
+      "loss": 1.2178,
+      "step": 9232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.896535150003326e-06,
+      "loss": 1.2441,
+      "step": 9233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894661647992657e-06,
+      "loss": 1.2285,
+      "step": 9234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8927886495369573e-06,
+      "loss": 1.2217,
+      "step": 9235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8909161547689667e-06,
+      "loss": 1.2529,
+      "step": 9236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.889044163821386e-06,
+      "loss": 1.2627,
+      "step": 9237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8871726768268893e-06,
+      "loss": 1.2832,
+      "step": 9238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.885301693918101e-06,
+      "loss": 1.1748,
+      "step": 9239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.883431215227621e-06,
+      "loss": 1.2227,
+      "step": 9240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.88156124088801e-06,
+      "loss": 1.1895,
+      "step": 9241
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879691771031793e-06,
+      "loss": 1.2041,
+      "step": 9242
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8778228057914594e-06,
+      "loss": 1.209,
+      "step": 9243
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875954345299462e-06,
+      "loss": 1.2217,
+      "step": 9244
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.87408638968822e-06,
+      "loss": 1.1914,
+      "step": 9245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8722189390901167e-06,
+      "loss": 1.2148,
+      "step": 9246
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8703519936374926e-06,
+      "loss": 1.2285,
+      "step": 9247
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.868485553462661e-06,
+      "loss": 1.1895,
+      "step": 9248
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8666196186978946e-06,
+      "loss": 1.1875,
+      "step": 9249
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.864754189475436e-06,
+      "loss": 1.293,
+      "step": 9250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8628892659274897e-06,
+      "loss": 1.2441,
+      "step": 9251
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.861024848186217e-06,
+      "loss": 1.2422,
+      "step": 9252
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8591609363837515e-06,
+      "loss": 1.2012,
+      "step": 9253
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8572975306521877e-06,
+      "loss": 1.2236,
+      "step": 9254
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.855434631123589e-06,
+      "loss": 1.1807,
+      "step": 9255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.853572237929969e-06,
+      "loss": 1.2334,
+      "step": 9256
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.851710351203326e-06,
+      "loss": 1.2041,
+      "step": 9257
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8498489710756084e-06,
+      "loss": 1.1855,
+      "step": 9258
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8479880976787343e-06,
+      "loss": 1.2295,
+      "step": 9259
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.846127731144578e-06,
+      "loss": 1.2192,
+      "step": 9260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8442678716049866e-06,
+      "loss": 1.2383,
+      "step": 9261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8424085191917693e-06,
+      "loss": 1.1895,
+      "step": 9262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8405496740366966e-06,
+      "loss": 1.1982,
+      "step": 9263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8386913362715053e-06,
+      "loss": 1.21,
+      "step": 9264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.836833506027896e-06,
+      "loss": 1.2227,
+      "step": 9265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8349761834375345e-06,
+      "loss": 1.2041,
+      "step": 9266
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.833119368632049e-06,
+      "loss": 1.2334,
+      "step": 9267
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831263061743028e-06,
+      "loss": 1.1816,
+      "step": 9268
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8294072629020306e-06,
+      "loss": 1.2031,
+      "step": 9269
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8275519722405765e-06,
+      "loss": 1.1982,
+      "step": 9270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.825697189890151e-06,
+      "loss": 1.2041,
+      "step": 9271
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.823842915982201e-06,
+      "loss": 1.2637,
+      "step": 9272
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8219891506481412e-06,
+      "loss": 1.2002,
+      "step": 9273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8201358940193457e-06,
+      "loss": 1.2285,
+      "step": 9274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8182831462271575e-06,
+      "loss": 1.1436,
+      "step": 9275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.816430907402877e-06,
+      "loss": 1.1816,
+      "step": 9276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.814579177677773e-06,
+      "loss": 1.1992,
+      "step": 9277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.812727957183076e-06,
+      "loss": 1.2266,
+      "step": 9278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.81087724604999e-06,
+      "loss": 1.2344,
+      "step": 9279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8090270444096666e-06,
+      "loss": 1.2461,
+      "step": 9280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8071773523932312e-06,
+      "loss": 1.1758,
+      "step": 9281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8053281701317725e-06,
+      "loss": 1.2061,
+      "step": 9282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.803479497756345e-06,
+      "loss": 1.209,
+      "step": 9283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8016313353979575e-06,
+      "loss": 1.1631,
+      "step": 9284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.799783683187589e-06,
+      "loss": 1.2021,
+      "step": 9285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.79793654125619e-06,
+      "loss": 1.2432,
+      "step": 9286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.796089909734665e-06,
+      "loss": 1.2041,
+      "step": 9287
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.79424378875388e-06,
+      "loss": 1.2148,
+      "step": 9288
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.792398178444673e-06,
+      "loss": 1.1973,
+      "step": 9289
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.790553078937841e-06,
+      "loss": 1.168,
+      "step": 9290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.788708490364147e-06,
+      "loss": 1.2217,
+      "step": 9291
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7868644128543166e-06,
+      "loss": 1.2119,
+      "step": 9292
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7850208465390404e-06,
+      "loss": 1.2256,
+      "step": 9293
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.78317779154897e-06,
+      "loss": 1.248,
+      "step": 9294
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7813352480147273e-06,
+      "loss": 1.2275,
+      "step": 9295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7794932160668866e-06,
+      "loss": 1.1953,
+      "step": 9296
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7776516958359956e-06,
+      "loss": 1.1963,
+      "step": 9297
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7758106874525626e-06,
+      "loss": 1.1973,
+      "step": 9298
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7739701910470595e-06,
+      "loss": 1.209,
+      "step": 9299
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7721302067499234e-06,
+      "loss": 1.2588,
+      "step": 9300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.770290734691553e-06,
+      "loss": 1.1943,
+      "step": 9301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.768451775002311e-06,
+      "loss": 1.2129,
+      "step": 9302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7666133278125296e-06,
+      "loss": 1.1865,
+      "step": 9303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.76477539325249e-06,
+      "loss": 1.1846,
+      "step": 9304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.762937971452453e-06,
+      "loss": 1.2217,
+      "step": 9305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7611010625426347e-06,
+      "loss": 1.2363,
+      "step": 9306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7592646666532165e-06,
+      "loss": 1.165,
+      "step": 9307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7574287839143454e-06,
+      "loss": 1.1465,
+      "step": 9308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.755593414456128e-06,
+      "loss": 1.2178,
+      "step": 9309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.753758558408639e-06,
+      "loss": 1.1797,
+      "step": 9310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7519242159019166e-06,
+      "loss": 1.1973,
+      "step": 9311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7500903870659535e-06,
+      "loss": 1.2227,
+      "step": 9312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748257072030718e-06,
+      "loss": 1.2295,
+      "step": 9313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7464242709261334e-06,
+      "loss": 1.1904,
+      "step": 9314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.744591983882099e-06,
+      "loss": 1.167,
+      "step": 9315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.742760211028459e-06,
+      "loss": 1.2041,
+      "step": 9316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7409289524950356e-06,
+      "loss": 1.2002,
+      "step": 9317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.739098208411609e-06,
+      "loss": 1.2158,
+      "step": 9318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.737267978907927e-06,
+      "loss": 1.1924,
+      "step": 9319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7354382641136934e-06,
+      "loss": 1.2295,
+      "step": 9320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.733609064158578e-06,
+      "loss": 1.2246,
+      "step": 9321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.731780379172222e-06,
+      "loss": 1.1797,
+      "step": 9322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7299522092842256e-06,
+      "loss": 1.1582,
+      "step": 9323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7281245546241452e-06,
+      "loss": 1.1846,
+      "step": 9324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7262974153215084e-06,
+      "loss": 1.249,
+      "step": 9325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7244707915058057e-06,
+      "loss": 1.1855,
+      "step": 9326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.722644683306489e-06,
+      "loss": 1.1934,
+      "step": 9327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7208190908529743e-06,
+      "loss": 1.1758,
+      "step": 9328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7189940142746405e-06,
+      "loss": 1.1396,
+      "step": 9329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.717169453700833e-06,
+      "loss": 1.1689,
+      "step": 9330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.715345409260861e-06,
+      "loss": 1.2021,
+      "step": 9331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713521881083986e-06,
+      "loss": 1.1992,
+      "step": 9332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711698869299447e-06,
+      "loss": 1.1904,
+      "step": 9333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7098763740364387e-06,
+      "loss": 1.2412,
+      "step": 9334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7080543954241223e-06,
+      "loss": 1.2383,
+      "step": 9335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7062329335916215e-06,
+      "loss": 1.2168,
+      "step": 9336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7044119886680233e-06,
+      "loss": 1.1895,
+      "step": 9337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7025915607823773e-06,
+      "loss": 1.1787,
+      "step": 9338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7007716500637006e-06,
+      "loss": 1.1924,
+      "step": 9339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.698952256640964e-06,
+      "loss": 1.1895,
+      "step": 9340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.697133380643111e-06,
+      "loss": 1.1807,
+      "step": 9341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.695315022199044e-06,
+      "loss": 1.2217,
+      "step": 9342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6934971814376322e-06,
+      "loss": 1.2246,
+      "step": 9343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6916798584877046e-06,
+      "loss": 1.2012,
+      "step": 9344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6898630534780544e-06,
+      "loss": 1.1943,
+      "step": 9345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68804676653744e-06,
+      "loss": 1.2285,
+      "step": 9346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.686230997794582e-06,
+      "loss": 1.2148,
+      "step": 9347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6844157473781605e-06,
+      "loss": 1.2695,
+      "step": 9348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6826010154168236e-06,
+      "loss": 1.2334,
+      "step": 9349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.680786802039179e-06,
+      "loss": 1.1426,
+      "step": 9350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.678973107373809e-06,
+      "loss": 1.2334,
+      "step": 9351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6771599315492416e-06,
+      "loss": 1.1934,
+      "step": 9352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.675347274693977e-06,
+      "loss": 1.1914,
+      "step": 9353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6735351369364803e-06,
+      "loss": 1.2256,
+      "step": 9354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6717235184051792e-06,
+      "loss": 1.2139,
+      "step": 9355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6699124192284587e-06,
+      "loss": 1.1934,
+      "step": 9356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6681018395346704e-06,
+      "loss": 1.1748,
+      "step": 9357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6662917794521357e-06,
+      "loss": 1.2178,
+      "step": 9358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6644822391091352e-06,
+      "loss": 1.208,
+      "step": 9359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.662673218633902e-06,
+      "loss": 1.2412,
+      "step": 9360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6608647181546465e-06,
+      "loss": 1.1943,
+      "step": 9361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6590567377995357e-06,
+      "loss": 1.2061,
+      "step": 9362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6572492776967016e-06,
+      "loss": 1.2109,
+      "step": 9363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.65544233797424e-06,
+      "loss": 1.165,
+      "step": 9364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653635918760207e-06,
+      "loss": 1.208,
+      "step": 9365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6518300201826243e-06,
+      "loss": 1.251,
+      "step": 9366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6500246423694786e-06,
+      "loss": 1.1719,
+      "step": 9367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.64821978544871e-06,
+      "loss": 1.167,
+      "step": 9368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.646415449548233e-06,
+      "loss": 1.2021,
+      "step": 9369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6446116347959206e-06,
+      "loss": 1.2227,
+      "step": 9370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6428083413196093e-06,
+      "loss": 1.1973,
+      "step": 9371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.641005569247097e-06,
+      "loss": 1.2363,
+      "step": 9372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.639203318706146e-06,
+      "loss": 1.1992,
+      "step": 9373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.637401589824484e-06,
+      "loss": 1.2354,
+      "step": 9374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6356003827298016e-06,
+      "loss": 1.2275,
+      "step": 9375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6337996975497428e-06,
+      "loss": 1.2031,
+      "step": 9376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6319995344119254e-06,
+      "loss": 1.1885,
+      "step": 9377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.630199893443929e-06,
+      "loss": 1.2227,
+      "step": 9378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.628400774773291e-06,
+      "loss": 1.2158,
+      "step": 9379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6266021785275173e-06,
+      "loss": 1.248,
+      "step": 9380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6248041048340732e-06,
+      "loss": 1.1611,
+      "step": 9381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6230065538203888e-06,
+      "loss": 1.208,
+      "step": 9382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6212095256138547e-06,
+      "loss": 1.2344,
+      "step": 9383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6194130203418312e-06,
+      "loss": 1.2295,
+      "step": 9384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.61761703813163e-06,
+      "loss": 1.2188,
+      "step": 9385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6158215791105323e-06,
+      "loss": 1.2598,
+      "step": 9386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.614026643405787e-06,
+      "loss": 1.1992,
+      "step": 9387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6122322311446024e-06,
+      "loss": 1.2119,
+      "step": 9388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.610438342454142e-06,
+      "loss": 1.1758,
+      "step": 9389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6086449774615418e-06,
+      "loss": 1.2285,
+      "step": 9390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.606852136293898e-06,
+      "loss": 1.1816,
+      "step": 9391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6050598190782704e-06,
+      "loss": 1.2314,
+      "step": 9392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6032680259416733e-06,
+      "loss": 1.2227,
+      "step": 9393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.601476757011099e-06,
+      "loss": 1.2197,
+      "step": 9394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5996860124134914e-06,
+      "loss": 1.2041,
+      "step": 9395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597895792275764e-06,
+      "loss": 1.2236,
+      "step": 9396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.596106096724784e-06,
+      "loss": 1.2207,
+      "step": 9397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.59431692588739e-06,
+      "loss": 1.2188,
+      "step": 9398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5925282798903794e-06,
+      "loss": 1.2012,
+      "step": 9399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.590740158860514e-06,
+      "loss": 1.2275,
+      "step": 9400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5889525629245184e-06,
+      "loss": 1.2178,
+      "step": 9401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5871654922090784e-06,
+      "loss": 1.2305,
+      "step": 9402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.585378946840844e-06,
+      "loss": 1.207,
+      "step": 9403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.583592926946431e-06,
+      "loss": 1.2549,
+      "step": 9404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5818074326524088e-06,
+      "loss": 1.2021,
+      "step": 9405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.580022464085318e-06,
+      "loss": 1.1934,
+      "step": 9406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5782380213716585e-06,
+      "loss": 1.1836,
+      "step": 9407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.576454104637893e-06,
+      "loss": 1.2334,
+      "step": 9408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.574670714010451e-06,
+      "loss": 1.1846,
+      "step": 9409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5728878496157174e-06,
+      "loss": 1.2061,
+      "step": 9410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5711055115800452e-06,
+      "loss": 1.2227,
+      "step": 9411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5693237000297534e-06,
+      "loss": 1.2031,
+      "step": 9412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.56754241509111e-06,
+      "loss": 1.2393,
+      "step": 9413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5657616568903586e-06,
+      "loss": 1.1914,
+      "step": 9414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5639814255536997e-06,
+      "loss": 1.2012,
+      "step": 9415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.562201721207306e-06,
+      "loss": 1.21,
+      "step": 9416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5604225439772954e-06,
+      "loss": 1.21,
+      "step": 9417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.558643893989763e-06,
+      "loss": 1.166,
+      "step": 9418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.55686577137076e-06,
+      "loss": 1.209,
+      "step": 9419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5550881762463064e-06,
+      "loss": 1.2109,
+      "step": 9420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.553311108742372e-06,
+      "loss": 1.2383,
+      "step": 9421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.551534568984899e-06,
+      "loss": 1.1982,
+      "step": 9422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5497585570997975e-06,
+      "loss": 1.2529,
+      "step": 9423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5479830732129317e-06,
+      "loss": 1.1885,
+      "step": 9424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546208117450125e-06,
+      "loss": 1.2227,
+      "step": 9425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5444336899371702e-06,
+      "loss": 1.1895,
+      "step": 9426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542659790799823e-06,
+      "loss": 1.2021,
+      "step": 9427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5408864201638005e-06,
+      "loss": 1.2148,
+      "step": 9428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.539113578154774e-06,
+      "loss": 1.1934,
+      "step": 9429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.537341264898392e-06,
+      "loss": 1.2197,
+      "step": 9430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5355694805202567e-06,
+      "loss": 1.2168,
+      "step": 9431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5337982251459383e-06,
+      "loss": 1.1846,
+      "step": 9432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5320274989009564e-06,
+      "loss": 1.1465,
+      "step": 9433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.530257301910808e-06,
+      "loss": 1.2207,
+      "step": 9434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5284876343009447e-06,
+      "loss": 1.1963,
+      "step": 9435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.526718496196785e-06,
+      "loss": 1.1875,
+      "step": 9436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5249498877237066e-06,
+      "loss": 1.1729,
+      "step": 9437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5231818090070505e-06,
+      "loss": 1.2607,
+      "step": 9438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5214142601721202e-06,
+      "loss": 1.2031,
+      "step": 9439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.519647241344185e-06,
+      "loss": 1.2227,
+      "step": 9440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5178807526484682e-06,
+      "loss": 1.2021,
+      "step": 9441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5161147942101637e-06,
+      "loss": 1.1875,
+      "step": 9442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5143493661544238e-06,
+      "loss": 1.1606,
+      "step": 9443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.512584468606365e-06,
+      "loss": 1.2129,
+      "step": 9444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.510820101691065e-06,
+      "loss": 1.1753,
+      "step": 9445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5090562655335637e-06,
+      "loss": 1.2432,
+      "step": 9446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507292960258867e-06,
+      "loss": 1.2109,
+      "step": 9447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.50553018599194e-06,
+      "loss": 1.2461,
+      "step": 9448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5037679428577064e-06,
+      "loss": 1.1924,
+      "step": 9449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.502006230981059e-06,
+      "loss": 1.1743,
+      "step": 9450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5002450504868457e-06,
+      "loss": 1.2129,
+      "step": 9451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.498484401499892e-06,
+      "loss": 1.2041,
+      "step": 9452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.496724284144967e-06,
+      "loss": 1.2139,
+      "step": 9453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.49496469854681e-06,
+      "loss": 1.1855,
+      "step": 9454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4932056448301243e-06,
+      "loss": 1.2041,
+      "step": 9455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.491447123119577e-06,
+      "loss": 1.2627,
+      "step": 9456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4896891335397886e-06,
+      "loss": 1.2412,
+      "step": 9457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4879316762153484e-06,
+      "loss": 1.2207,
+      "step": 9458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4861747512708113e-06,
+      "loss": 1.1846,
+      "step": 9459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4844183588306926e-06,
+      "loss": 1.208,
+      "step": 9460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.48266249901946e-06,
+      "loss": 1.1748,
+      "step": 9461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4809071719615553e-06,
+      "loss": 1.1494,
+      "step": 9462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4791523777813784e-06,
+      "loss": 1.1934,
+      "step": 9463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4773981166032913e-06,
+      "loss": 1.1924,
+      "step": 9464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.475644388551618e-06,
+      "loss": 1.2607,
+      "step": 9465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4738911937506447e-06,
+      "loss": 1.2285,
+      "step": 9466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.472138532324623e-06,
+      "loss": 1.1523,
+      "step": 9467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.470386404397763e-06,
+      "loss": 1.1787,
+      "step": 9468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4686348100942335e-06,
+      "loss": 1.1982,
+      "step": 9469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.466883749538175e-06,
+      "loss": 1.1943,
+      "step": 9470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4651332228536817e-06,
+      "loss": 1.1865,
+      "step": 9471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4633832301648165e-06,
+      "loss": 1.1733,
+      "step": 9472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4616337715955985e-06,
+      "loss": 1.1924,
+      "step": 9473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.459884847270013e-06,
+      "loss": 1.2549,
+      "step": 9474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.458136457312006e-06,
+      "loss": 1.2568,
+      "step": 9475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4563886018454896e-06,
+      "loss": 1.2246,
+      "step": 9476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4546412809943277e-06,
+      "loss": 1.1924,
+      "step": 9477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.452894494882355e-06,
+      "loss": 1.166,
+      "step": 9478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4511482436333667e-06,
+      "loss": 1.249,
+      "step": 9479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4494025273711196e-06,
+      "loss": 1.249,
+      "step": 9480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4476573462193345e-06,
+      "loss": 1.1553,
+      "step": 9481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4459127003016882e-06,
+      "loss": 1.1807,
+      "step": 9482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4441685897418265e-06,
+      "loss": 1.2021,
+      "step": 9483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4424250146633564e-06,
+      "loss": 1.208,
+      "step": 9484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44068197518984e-06,
+      "loss": 1.1943,
+      "step": 9485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438939471444809e-06,
+      "loss": 1.1865,
+      "step": 9486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4371975035517514e-06,
+      "loss": 1.1982,
+      "step": 9487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4354560716341293e-06,
+      "loss": 1.1973,
+      "step": 9488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4337151758153478e-06,
+      "loss": 1.1875,
+      "step": 9489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4319748162187893e-06,
+      "loss": 1.2197,
+      "step": 9490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4302349929677914e-06,
+      "loss": 1.1992,
+      "step": 9491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4284957061856595e-06,
+      "loss": 1.2061,
+      "step": 9492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4267569559956496e-06,
+      "loss": 1.2295,
+      "step": 9493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425018742520987e-06,
+      "loss": 1.1846,
+      "step": 9494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.423281065884865e-06,
+      "loss": 1.1973,
+      "step": 9495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4215439262104334e-06,
+      "loss": 1.2031,
+      "step": 9496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4198073236207965e-06,
+      "loss": 1.1855,
+      "step": 9497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4180712582390296e-06,
+      "loss": 1.2207,
+      "step": 9498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4163357301881687e-06,
+      "loss": 1.2061,
+      "step": 9499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4146007395912095e-06,
+      "loss": 1.293,
+      "step": 9500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.412866286571113e-06,
+      "loss": 1.2139,
+      "step": 9501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4111323712507974e-06,
+      "loss": 1.2842,
+      "step": 9502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4093989937531447e-06,
+      "loss": 1.1914,
+      "step": 9503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.407666154201005e-06,
+      "loss": 1.2246,
+      "step": 9504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.405933852717176e-06,
+      "loss": 1.1924,
+      "step": 9505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4042020894244313e-06,
+      "loss": 1.2168,
+      "step": 9506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4024708644454975e-06,
+      "loss": 1.2051,
+      "step": 9507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4007401779030694e-06,
+      "loss": 1.2422,
+      "step": 9508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3990100299198004e-06,
+      "loss": 1.1953,
+      "step": 9509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3972804206183044e-06,
+      "loss": 1.2559,
+      "step": 9510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.395551350121159e-06,
+      "loss": 1.2354,
+      "step": 9511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.393822818550907e-06,
+      "loss": 1.2305,
+      "step": 9512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3920948260300447e-06,
+      "loss": 1.2188,
+      "step": 9513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.390367372681035e-06,
+      "loss": 1.1768,
+      "step": 9514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3886404586263045e-06,
+      "loss": 1.2129,
+      "step": 9515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3869140839882354e-06,
+      "loss": 1.2002,
+      "step": 9516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3851882488891853e-06,
+      "loss": 1.2275,
+      "step": 9517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3834629534514553e-06,
+      "loss": 1.2402,
+      "step": 9518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3817381977973197e-06,
+      "loss": 1.1445,
+      "step": 9519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380013982049012e-06,
+      "loss": 1.1875,
+      "step": 9520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3782903063287288e-06,
+      "loss": 1.2139,
+      "step": 9521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3765671707586234e-06,
+      "loss": 1.1953,
+      "step": 9522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3748445754608127e-06,
+      "loss": 1.1719,
+      "step": 9523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.373122520557384e-06,
+      "loss": 1.2344,
+      "step": 9524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3714010061703785e-06,
+      "loss": 1.2168,
+      "step": 9525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3696800324217927e-06,
+      "loss": 1.2168,
+      "step": 9526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3679595994335967e-06,
+      "loss": 1.1777,
+      "step": 9527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3662397073277175e-06,
+      "loss": 1.1631,
+      "step": 9528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3645203562260465e-06,
+      "loss": 1.207,
+      "step": 9529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.362801546250424e-06,
+      "loss": 1.2246,
+      "step": 9530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.361083277522672e-06,
+      "loss": 1.2236,
+      "step": 9531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.359365550164562e-06,
+      "loss": 1.1992,
+      "step": 9532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3576483642978297e-06,
+      "loss": 1.1973,
+      "step": 9533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3559317200441677e-06,
+      "loss": 1.2236,
+      "step": 9534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.354215617525237e-06,
+      "loss": 1.2285,
+      "step": 9535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352500056862658e-06,
+      "loss": 1.2051,
+      "step": 9536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350785038178013e-06,
+      "loss": 1.2393,
+      "step": 9537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3490705615928434e-06,
+      "loss": 1.2344,
+      "step": 9538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.347356627228656e-06,
+      "loss": 1.2188,
+      "step": 9539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.345643235206916e-06,
+      "loss": 1.2461,
+      "step": 9540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.343930385649055e-06,
+      "loss": 1.1865,
+      "step": 9541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.342218078676457e-06,
+      "loss": 1.2188,
+      "step": 9542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3405063144104744e-06,
+      "loss": 1.2344,
+      "step": 9543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3387950929724226e-06,
+      "loss": 1.168,
+      "step": 9544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3370844144835724e-06,
+      "loss": 1.1807,
+      "step": 9545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3353742790651623e-06,
+      "loss": 1.2041,
+      "step": 9546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3336646868383883e-06,
+      "loss": 1.1973,
+      "step": 9547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.331955637924409e-06,
+      "loss": 1.2168,
+      "step": 9548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.330247132444349e-06,
+      "loss": 1.1602,
+      "step": 9549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.328539170519282e-06,
+      "loss": 1.2412,
+      "step": 9550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3268317522702542e-06,
+      "loss": 1.2051,
+      "step": 9551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.32512487781827e-06,
+      "loss": 1.1973,
+      "step": 9552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3234185472843016e-06,
+      "loss": 1.2578,
+      "step": 9553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3217127607892685e-06,
+      "loss": 1.2119,
+      "step": 9554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3200075184540628e-06,
+      "loss": 1.2021,
+      "step": 9555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3183028203995352e-06,
+      "loss": 1.2217,
+      "step": 9556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3165986667465003e-06,
+      "loss": 1.1289,
+      "step": 9557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3148950576157246e-06,
+      "loss": 1.2178,
+      "step": 9558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.313191993127946e-06,
+      "loss": 1.2158,
+      "step": 9559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3114894734038616e-06,
+      "loss": 1.207,
+      "step": 9560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.309787498564132e-06,
+      "loss": 1.2598,
+      "step": 9561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3080860687293693e-06,
+      "loss": 1.1787,
+      "step": 9562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.306385184020158e-06,
+      "loss": 1.2617,
+      "step": 9563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.304684844557038e-06,
+      "loss": 1.2002,
+      "step": 9564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3029850504605157e-06,
+      "loss": 1.1514,
+      "step": 9565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3012858018510474e-06,
+      "loss": 1.2256,
+      "step": 9566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2995870988490666e-06,
+      "loss": 1.1943,
+      "step": 9567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2978889415749584e-06,
+      "loss": 1.2285,
+      "step": 9568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2961913301490725e-06,
+      "loss": 1.2314,
+      "step": 9569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2944942646917135e-06,
+      "loss": 1.2129,
+      "step": 9570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2927977453231554e-06,
+      "loss": 1.1885,
+      "step": 9571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2911017721636307e-06,
+      "loss": 1.1689,
+      "step": 9572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2894063453333325e-06,
+      "loss": 1.21,
+      "step": 9573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2877114649524155e-06,
+      "loss": 1.1836,
+      "step": 9574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286017131140996e-06,
+      "loss": 1.2266,
+      "step": 9575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2843233440191525e-06,
+      "loss": 1.21,
+      "step": 9576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2826301037069244e-06,
+      "loss": 1.2412,
+      "step": 9577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2809374103243065e-06,
+      "loss": 1.1797,
+      "step": 9578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2792452639912633e-06,
+      "loss": 1.1934,
+      "step": 9579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2775536648277164e-06,
+      "loss": 1.1748,
+      "step": 9580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.27586261295355e-06,
+      "loss": 1.2197,
+      "step": 9581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.274172108488609e-06,
+      "loss": 1.2451,
+      "step": 9582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2724821515526983e-06,
+      "loss": 1.166,
+      "step": 9583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2707927422655872e-06,
+      "loss": 1.1826,
+      "step": 9584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2691038807470044e-06,
+      "loss": 1.2051,
+      "step": 9585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2674155671166342e-06,
+      "loss": 1.1719,
+      "step": 9586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265727801494132e-06,
+      "loss": 1.2207,
+      "step": 9587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.264040583999105e-06,
+      "loss": 1.2344,
+      "step": 9588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2623539147511353e-06,
+      "loss": 1.2402,
+      "step": 9589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2606677938697484e-06,
+      "loss": 1.1982,
+      "step": 9590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2589822214744427e-06,
+      "loss": 1.2021,
+      "step": 9591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257297197684675e-06,
+      "loss": 1.2285,
+      "step": 9592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2556127226198653e-06,
+      "loss": 1.1943,
+      "step": 9593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.253928796399386e-06,
+      "loss": 1.2168,
+      "step": 9594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2522454191425767e-06,
+      "loss": 1.2139,
+      "step": 9595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2505625909687458e-06,
+      "loss": 1.1973,
+      "step": 9596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2488803119971535e-06,
+      "loss": 1.1807,
+      "step": 9597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2471985823470175e-06,
+      "loss": 1.1855,
+      "step": 9598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2455174021375257e-06,
+      "loss": 1.2383,
+      "step": 9599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2438367714878216e-06,
+      "loss": 1.2178,
+      "step": 9600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2421566905170155e-06,
+      "loss": 1.2432,
+      "step": 9601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2404771593441667e-06,
+      "loss": 1.1924,
+      "step": 9602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2387981780883096e-06,
+      "loss": 1.1592,
+      "step": 9603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2371197468684336e-06,
+      "loss": 1.2021,
+      "step": 9604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2354418658034895e-06,
+      "loss": 1.2627,
+      "step": 9605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.233764535012384e-06,
+      "loss": 1.2051,
+      "step": 9606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2320877546139918e-06,
+      "loss": 1.2246,
+      "step": 9607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2304115247271474e-06,
+      "loss": 1.2295,
+      "step": 9608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.228735845470643e-06,
+      "loss": 1.207,
+      "step": 9609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.227060716963236e-06,
+      "loss": 1.2334,
+      "step": 9610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.225386139323643e-06,
+      "loss": 1.2295,
+      "step": 9611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2237121126705395e-06,
+      "loss": 1.1699,
+      "step": 9612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2220386371225665e-06,
+      "loss": 1.1768,
+      "step": 9613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2203657127983193e-06,
+      "loss": 1.1777,
+      "step": 9614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.218693339816359e-06,
+      "loss": 1.1982,
+      "step": 9615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.217021518295207e-06,
+      "loss": 1.2236,
+      "step": 9616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.215350248353346e-06,
+      "loss": 1.2188,
+      "step": 9617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2136795301092174e-06,
+      "loss": 1.1914,
+      "step": 9618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2120093636812266e-06,
+      "loss": 1.2041,
+      "step": 9619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2103397491877376e-06,
+      "loss": 1.2217,
+      "step": 9620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2086706867470774e-06,
+      "loss": 1.2109,
+      "step": 9621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2070021764775284e-06,
+      "loss": 1.1953,
+      "step": 9622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.20533421849734e-06,
+      "loss": 1.1631,
+      "step": 9623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2036668129247174e-06,
+      "loss": 1.1582,
+      "step": 9624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2019999598778385e-06,
+      "loss": 1.1973,
+      "step": 9625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200333659474825e-06,
+      "loss": 1.168,
+      "step": 9626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1986679118337684e-06,
+      "loss": 1.166,
+      "step": 9627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1970027170727215e-06,
+      "loss": 1.2109,
+      "step": 9628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1953380753097e-06,
+      "loss": 1.1709,
+      "step": 9629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.19367398666267e-06,
+      "loss": 1.1758,
+      "step": 9630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1920104512495667e-06,
+      "loss": 1.1895,
+      "step": 9631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190347469188289e-06,
+      "loss": 1.2852,
+      "step": 9632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.188685040596693e-06,
+      "loss": 1.1826,
+      "step": 9633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.18702316559259e-06,
+      "loss": 1.2188,
+      "step": 9634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1853618442937586e-06,
+      "loss": 1.1982,
+      "step": 9635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183701076817938e-06,
+      "loss": 1.209,
+      "step": 9636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.182040863282825e-06,
+      "loss": 1.2598,
+      "step": 9637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.18038120380608e-06,
+      "loss": 1.2334,
+      "step": 9638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1787220985053224e-06,
+      "loss": 1.2002,
+      "step": 9639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1770635474981328e-06,
+      "loss": 1.21,
+      "step": 9640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1754055509020566e-06,
+      "loss": 1.2275,
+      "step": 9641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1737481088345893e-06,
+      "loss": 1.2061,
+      "step": 9642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1720912214131972e-06,
+      "loss": 1.1875,
+      "step": 9643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170434888755303e-06,
+      "loss": 1.1924,
+      "step": 9644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1687791109782906e-06,
+      "loss": 1.167,
+      "step": 9645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.167123888199506e-06,
+      "loss": 1.2373,
+      "step": 9646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.165469220536255e-06,
+      "loss": 1.1846,
+      "step": 9647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.163815108105802e-06,
+      "loss": 1.209,
+      "step": 9648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162161551025377e-06,
+      "loss": 1.2168,
+      "step": 9649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.160508549412167e-06,
+      "loss": 1.2119,
+      "step": 9650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1588561033833153e-06,
+      "loss": 1.208,
+      "step": 9651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1572042130559345e-06,
+      "loss": 1.1719,
+      "step": 9652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1555528785470915e-06,
+      "loss": 1.1895,
+      "step": 9653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1539020999738235e-06,
+      "loss": 1.2285,
+      "step": 9654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152251877453113e-06,
+      "loss": 1.1758,
+      "step": 9655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1506022111019144e-06,
+      "loss": 1.2129,
+      "step": 9656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148953101037138e-06,
+      "loss": 1.1963,
+      "step": 9657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1473045473756615e-06,
+      "loss": 1.2422,
+      "step": 9658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.14565655023431e-06,
+      "loss": 1.2148,
+      "step": 9659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1440091097298777e-06,
+      "loss": 1.2041,
+      "step": 9660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1423622259791253e-06,
+      "loss": 1.2236,
+      "step": 9661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140715899098765e-06,
+      "loss": 1.1797,
+      "step": 9662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1390701292054694e-06,
+      "loss": 1.2168,
+      "step": 9663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1374249164158735e-06,
+      "loss": 1.2285,
+      "step": 9664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135780260846576e-06,
+      "loss": 1.2021,
+      "step": 9665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1341361626141354e-06,
+      "loss": 1.21,
+      "step": 9666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1324926218350616e-06,
+      "loss": 1.2686,
+      "step": 9667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130849638625838e-06,
+      "loss": 1.2256,
+      "step": 9668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1292072131029027e-06,
+      "loss": 1.2725,
+      "step": 9669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127565345382656e-06,
+      "loss": 1.2275,
+      "step": 9670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1259240355814503e-06,
+      "loss": 1.2148,
+      "step": 9671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1242832838156103e-06,
+      "loss": 1.1943,
+      "step": 9672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.122643090201414e-06,
+      "loss": 1.1914,
+      "step": 9673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1210034548551028e-06,
+      "loss": 1.2256,
+      "step": 9674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1193643778928776e-06,
+      "loss": 1.2549,
+      "step": 9675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.117725859430898e-06,
+      "loss": 1.2285,
+      "step": 9676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1160878995852885e-06,
+      "loss": 1.2266,
+      "step": 9677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1144504984721324e-06,
+      "loss": 1.2285,
+      "step": 9678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1128136562074675e-06,
+      "loss": 1.249,
+      "step": 9679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1111773729072984e-06,
+      "loss": 1.1846,
+      "step": 9680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1095416486875885e-06,
+      "loss": 1.2217,
+      "step": 9681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.107906483664263e-06,
+      "loss": 1.2256,
+      "step": 9682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.106271877953203e-06,
+      "loss": 1.1895,
+      "step": 9683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1046378316702566e-06,
+      "loss": 1.2559,
+      "step": 9684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1030043449312264e-06,
+      "loss": 1.2217,
+      "step": 9685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.101371417851881e-06,
+      "loss": 1.2041,
+      "step": 9686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.099739050547939e-06,
+      "loss": 1.1934,
+      "step": 9687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0981072431350903e-06,
+      "loss": 1.1748,
+      "step": 9688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.096475995728978e-06,
+      "loss": 1.1709,
+      "step": 9689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.094845308445217e-06,
+      "loss": 1.2148,
+      "step": 9690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093215181399365e-06,
+      "loss": 1.2041,
+      "step": 9691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0915856147069537e-06,
+      "loss": 1.2412,
+      "step": 9692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089956608483468e-06,
+      "loss": 1.2207,
+      "step": 9693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0883281628443587e-06,
+      "loss": 1.1992,
+      "step": 9694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0867002779050284e-06,
+      "loss": 1.2275,
+      "step": 9695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.085072953780847e-06,
+      "loss": 1.1963,
+      "step": 9696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.083446190587145e-06,
+      "loss": 1.2188,
+      "step": 9697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0818199884392143e-06,
+      "loss": 1.2051,
+      "step": 9698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.080194347452297e-06,
+      "loss": 1.2139,
+      "step": 9699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0785692677416035e-06,
+      "loss": 1.2295,
+      "step": 9700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0769447494223048e-06,
+      "loss": 1.1675,
+      "step": 9701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075320792609534e-06,
+      "loss": 1.2305,
+      "step": 9702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0736973974183707e-06,
+      "loss": 1.1973,
+      "step": 9703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072074563963873e-06,
+      "loss": 1.1992,
+      "step": 9704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0704522923610504e-06,
+      "loss": 1.2021,
+      "step": 9705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.068830582724873e-06,
+      "loss": 1.2178,
+      "step": 9706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0672094351702688e-06,
+      "loss": 1.1963,
+      "step": 9707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.065588849812129e-06,
+      "loss": 1.2031,
+      "step": 9708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0639688267653055e-06,
+      "loss": 1.2109,
+      "step": 9709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0623493661446083e-06,
+      "loss": 1.2246,
+      "step": 9710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.06073046806481e-06,
+      "loss": 1.1885,
+      "step": 9711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0591121326406405e-06,
+      "loss": 1.249,
+      "step": 9712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0574943599867914e-06,
+      "loss": 1.1738,
+      "step": 9713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0558771502179174e-06,
+      "loss": 1.2314,
+      "step": 9714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.054260503448624e-06,
+      "loss": 1.2109,
+      "step": 9715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052644419793486e-06,
+      "loss": 1.2383,
+      "step": 9716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0510288993670357e-06,
+      "loss": 1.1924,
+      "step": 9717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0494139422837634e-06,
+      "loss": 1.2256,
+      "step": 9718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0477995486581227e-06,
+      "loss": 1.2432,
+      "step": 9719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.046185718604524e-06,
+      "loss": 1.2168,
+      "step": 9720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.044572452237341e-06,
+      "loss": 1.1934,
+      "step": 9721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0429597496709074e-06,
+      "loss": 1.2188,
+      "step": 9722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.041347611019511e-06,
+      "loss": 1.2168,
+      "step": 9723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.039736036397405e-06,
+      "loss": 1.1953,
+      "step": 9724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0381250259188e-06,
+      "loss": 1.1621,
+      "step": 9725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.036514579697876e-06,
+      "loss": 1.2607,
+      "step": 9726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0349046978487573e-06,
+      "loss": 1.1973,
+      "step": 9727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0332953804855395e-06,
+      "loss": 1.2246,
+      "step": 9728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0316866277222737e-06,
+      "loss": 1.2227,
+      "step": 9729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0300784396729754e-06,
+      "loss": 1.2451,
+      "step": 9730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.028470816451611e-06,
+      "loss": 1.2266,
+      "step": 9731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.026863758172114e-06,
+      "loss": 1.2139,
+      "step": 9732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0252572649483803e-06,
+      "loss": 1.1914,
+      "step": 9733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023651336894262e-06,
+      "loss": 1.2656,
+      "step": 9734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0220459741235677e-06,
+      "loss": 1.1748,
+      "step": 9735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0204411767500697e-06,
+      "loss": 1.2178,
+      "step": 9736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0188369448875023e-06,
+      "loss": 1.2666,
+      "step": 9737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0172332786495575e-06,
+      "loss": 1.2158,
+      "step": 9738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015630178149881e-06,
+      "loss": 1.2002,
+      "step": 9739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.014027643502091e-06,
+      "loss": 1.1914,
+      "step": 9740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.012425674819757e-06,
+      "loss": 1.2021,
+      "step": 9741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010824272216413e-06,
+      "loss": 1.1807,
+      "step": 9742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0092234358055463e-06,
+      "loss": 1.1797,
+      "step": 9743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0076231657006086e-06,
+      "loss": 1.2129,
+      "step": 9744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.006023462015012e-06,
+      "loss": 1.166,
+      "step": 9745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0044243248621275e-06,
+      "loss": 1.1689,
+      "step": 9746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0028257543552864e-06,
+      "loss": 1.2295,
+      "step": 9747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001227750607778e-06,
+      "loss": 1.1855,
+      "step": 9748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9996303137328534e-06,
+      "loss": 1.2051,
+      "step": 9749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9980334438437253e-06,
+      "loss": 1.2197,
+      "step": 9750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.99643714105356e-06,
+      "loss": 1.2119,
+      "step": 9751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.994841405475487e-06,
+      "loss": 1.2109,
+      "step": 9752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993246237222599e-06,
+      "loss": 1.1963,
+      "step": 9753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991651636407944e-06,
+      "loss": 1.2227,
+      "step": 9754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9900576031445316e-06,
+      "loss": 1.2061,
+      "step": 9755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9884641375453307e-06,
+      "loss": 1.1748,
+      "step": 9756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9868712397232704e-06,
+      "loss": 1.1982,
+      "step": 9757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9852789097912405e-06,
+      "loss": 1.1787,
+      "step": 9758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9836871478620867e-06,
+      "loss": 1.2197,
+      "step": 9759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9820959540486184e-06,
+      "loss": 1.2119,
+      "step": 9760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9805053284636e-06,
+      "loss": 1.1699,
+      "step": 9761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.978915271219768e-06,
+      "loss": 1.165,
+      "step": 9762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.977325782429801e-06,
+      "loss": 1.1895,
+      "step": 9763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.97573686220635e-06,
+      "loss": 1.2148,
+      "step": 9764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9741485106620205e-06,
+      "loss": 1.2051,
+      "step": 9765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9725607279093817e-06,
+      "loss": 1.2354,
+      "step": 9766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9709735140609543e-06,
+      "loss": 1.167,
+      "step": 9767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969386869229224e-06,
+      "loss": 1.2129,
+      "step": 9768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9678007935266417e-06,
+      "loss": 1.2021,
+      "step": 9769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9662152870656137e-06,
+      "loss": 1.2285,
+      "step": 9770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9646303499584975e-06,
+      "loss": 1.2412,
+      "step": 9771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.963045982317622e-06,
+      "loss": 1.1553,
+      "step": 9772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9614621842552693e-06,
+      "loss": 1.2607,
+      "step": 9773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.959878955883687e-06,
+      "loss": 1.2178,
+      "step": 9774
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.95829629731507e-06,
+      "loss": 1.1768,
+      "step": 9775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9567142086615896e-06,
+      "loss": 1.1602,
+      "step": 9776
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.955132690035366e-06,
+      "loss": 1.2188,
+      "step": 9777
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9535517415484826e-06,
+      "loss": 1.2305,
+      "step": 9778
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951971363312978e-06,
+      "loss": 1.1953,
+      "step": 9779
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9503915554408535e-06,
+      "loss": 1.209,
+      "step": 9780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.948812318044071e-06,
+      "loss": 1.21,
+      "step": 9781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9472336512345523e-06,
+      "loss": 1.2021,
+      "step": 9782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9456555551241772e-06,
+      "loss": 1.2041,
+      "step": 9783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.944078029824783e-06,
+      "loss": 1.2188,
+      "step": 9784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942501075448172e-06,
+      "loss": 1.2236,
+      "step": 9785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9409246921061e-06,
+      "loss": 1.2275,
+      "step": 9786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93934887991029e-06,
+      "loss": 1.2314,
+      "step": 9787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9377736389724135e-06,
+      "loss": 1.1943,
+      "step": 9788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9361989694041094e-06,
+      "loss": 1.1475,
+      "step": 9789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9346248713169747e-06,
+      "loss": 1.1973,
+      "step": 9790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933051344822571e-06,
+      "loss": 1.2715,
+      "step": 9791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9314783900324064e-06,
+      "loss": 1.1855,
+      "step": 9792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929906007057959e-06,
+      "loss": 1.1592,
+      "step": 9793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9283341960106648e-06,
+      "loss": 1.1787,
+      "step": 9794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9267629570019186e-06,
+      "loss": 1.2178,
+      "step": 9795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925192290143071e-06,
+      "loss": 1.1553,
+      "step": 9796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9236221955454327e-06,
+      "loss": 1.2051,
+      "step": 9797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9220526733202826e-06,
+      "loss": 1.208,
+      "step": 9798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.920483723578853e-06,
+      "loss": 1.2002,
+      "step": 9799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9189153464323298e-06,
+      "loss": 1.2354,
+      "step": 9800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9173475419918665e-06,
+      "loss": 1.1719,
+      "step": 9801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9157803103685723e-06,
+      "loss": 1.1924,
+      "step": 9802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.914213651673521e-06,
+      "loss": 1.1943,
+      "step": 9803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9126475660177323e-06,
+      "loss": 1.1943,
+      "step": 9804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9110820535122044e-06,
+      "loss": 1.2441,
+      "step": 9805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9095171142678802e-06,
+      "loss": 1.2246,
+      "step": 9806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.907952748395672e-06,
+      "loss": 1.2246,
+      "step": 9807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9063889560064385e-06,
+      "loss": 1.1982,
+      "step": 9808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.904825737211009e-06,
+      "loss": 1.2129,
+      "step": 9809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.90326309212017e-06,
+      "loss": 1.2393,
+      "step": 9810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.901701020844665e-06,
+      "loss": 1.1689,
+      "step": 9811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.900139523495198e-06,
+      "loss": 1.208,
+      "step": 9812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8985786001824325e-06,
+      "loss": 1.208,
+      "step": 9813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.897018251016991e-06,
+      "loss": 1.2188,
+      "step": 9814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8954584761094574e-06,
+      "loss": 1.2256,
+      "step": 9815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8938992755703678e-06,
+      "loss": 1.1826,
+      "step": 9816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.892340649510227e-06,
+      "loss": 1.1816,
+      "step": 9817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8907825980394922e-06,
+      "loss": 1.1631,
+      "step": 9818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8892251212685842e-06,
+      "loss": 1.251,
+      "step": 9819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.88766821930788e-06,
+      "loss": 1.2012,
+      "step": 9820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8861118922677179e-06,
+      "loss": 1.207,
+      "step": 9821
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8845561402583957e-06,
+      "loss": 1.2051,
+      "step": 9822
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8830009633901703e-06,
+      "loss": 1.1572,
+      "step": 9823
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8814463617732515e-06,
+      "loss": 1.2314,
+      "step": 9824
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.879892335517819e-06,
+      "loss": 1.2559,
+      "step": 9825
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8783388847340022e-06,
+      "loss": 1.2148,
+      "step": 9826
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8767860095319013e-06,
+      "loss": 1.2119,
+      "step": 9827
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8752337100215622e-06,
+      "loss": 1.1953,
+      "step": 9828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8736819863129973e-06,
+      "loss": 1.2061,
+      "step": 9829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8721308385161796e-06,
+      "loss": 1.1992,
+      "step": 9830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.870580266741039e-06,
+      "loss": 1.1924,
+      "step": 9831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8690302710974594e-06,
+      "loss": 1.209,
+      "step": 9832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8674808516952914e-06,
+      "loss": 1.208,
+      "step": 9833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8659320086443444e-06,
+      "loss": 1.1875,
+      "step": 9834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8643837420543865e-06,
+      "loss": 1.2197,
+      "step": 9835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8628360520351384e-06,
+      "loss": 1.1748,
+      "step": 9836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8612889386962863e-06,
+      "loss": 1.2324,
+      "step": 9837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8597424021474742e-06,
+      "loss": 1.2363,
+      "step": 9838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8581964424983079e-06,
+      "loss": 1.2285,
+      "step": 9839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.856651059858342e-06,
+      "loss": 1.1924,
+      "step": 9840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8551062543371056e-06,
+      "loss": 1.1973,
+      "step": 9841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8535620260440756e-06,
+      "loss": 1.167,
+      "step": 9842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8520183750886954e-06,
+      "loss": 1.2012,
+      "step": 9843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8504753015803567e-06,
+      "loss": 1.2041,
+      "step": 9844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8489328056284196e-06,
+      "loss": 1.2041,
+      "step": 9845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.847390887342202e-06,
+      "loss": 1.2119,
+      "step": 9846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.845849546830979e-06,
+      "loss": 1.2285,
+      "step": 9847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8443087842039841e-06,
+      "loss": 1.21,
+      "step": 9848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8427685995704137e-06,
+      "loss": 1.249,
+      "step": 9849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8412289930394178e-06,
+      "loss": 1.2168,
+      "step": 9850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.839689964720114e-06,
+      "loss": 1.1934,
+      "step": 9851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8381515147215644e-06,
+      "loss": 1.248,
+      "step": 9852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8366136431528036e-06,
+      "loss": 1.2471,
+      "step": 9853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8350763501228198e-06,
+      "loss": 1.1572,
+      "step": 9854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8335396357405622e-06,
+      "loss": 1.2207,
+      "step": 9855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.832003500114936e-06,
+      "loss": 1.1699,
+      "step": 9856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8304679433548068e-06,
+      "loss": 1.1885,
+      "step": 9857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8289329655690014e-06,
+      "loss": 1.2061,
+      "step": 9858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8273985668663052e-06,
+      "loss": 1.2061,
+      "step": 9859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8258647473554547e-06,
+      "loss": 1.2432,
+      "step": 9860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8243315071451561e-06,
+      "loss": 1.2188,
+      "step": 9861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.822798846344066e-06,
+      "loss": 1.1797,
+      "step": 9862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8212667650608128e-06,
+      "loss": 1.2031,
+      "step": 9863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8197352634039655e-06,
+      "loss": 1.1768,
+      "step": 9864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8182043414820672e-06,
+      "loss": 1.1895,
+      "step": 9865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.816673999403611e-06,
+      "loss": 1.1943,
+      "step": 9866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8151442372770567e-06,
+      "loss": 1.2256,
+      "step": 9867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8136150552108123e-06,
+      "loss": 1.2607,
+      "step": 9868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8120864533132521e-06,
+      "loss": 1.2139,
+      "step": 9869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8105584316927128e-06,
+      "loss": 1.1895,
+      "step": 9870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8090309904574833e-06,
+      "loss": 1.209,
+      "step": 9871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8075041297158113e-06,
+      "loss": 1.2129,
+      "step": 9872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.805977849575905e-06,
+      "loss": 1.2031,
+      "step": 9873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8044521501459333e-06,
+      "loss": 1.1748,
+      "step": 9874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8029270315340253e-06,
+      "loss": 1.1924,
+      "step": 9875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8014024938482576e-06,
+      "loss": 1.1631,
+      "step": 9876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7998785371966833e-06,
+      "loss": 1.1709,
+      "step": 9877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7983551616872996e-06,
+      "loss": 1.1758,
+      "step": 9878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.796832367428073e-06,
+      "loss": 1.1689,
+      "step": 9879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.795310154526918e-06,
+      "loss": 1.2139,
+      "step": 9880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7937885230917174e-06,
+      "loss": 1.2148,
+      "step": 9881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7922674732303068e-06,
+      "loss": 1.1973,
+      "step": 9882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7907470050504849e-06,
+      "loss": 1.1826,
+      "step": 9883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7892271186600062e-06,
+      "loss": 1.2256,
+      "step": 9884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.787707814166586e-06,
+      "loss": 1.2197,
+      "step": 9885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7861890916778956e-06,
+      "loss": 1.1494,
+      "step": 9886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7846709513015715e-06,
+      "loss": 1.1934,
+      "step": 9887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7831533931451983e-06,
+      "loss": 1.1865,
+      "step": 9888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7816364173163269e-06,
+      "loss": 1.1738,
+      "step": 9889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.780120023922466e-06,
+      "loss": 1.1797,
+      "step": 9890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7786042130710823e-06,
+      "loss": 1.166,
+      "step": 9891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7770889848696004e-06,
+      "loss": 1.2217,
+      "step": 9892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7755743394254066e-06,
+      "loss": 1.21,
+      "step": 9893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7740602768458416e-06,
+      "loss": 1.2275,
+      "step": 9894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7725467972382104e-06,
+      "loss": 1.1318,
+      "step": 9895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7710339007097687e-06,
+      "loss": 1.2285,
+      "step": 9896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7695215873677375e-06,
+      "loss": 1.2529,
+      "step": 9897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7680098573192905e-06,
+      "loss": 1.2041,
+      "step": 9898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7664987106715736e-06,
+      "loss": 1.2148,
+      "step": 9899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7649881475316721e-06,
+      "loss": 1.2295,
+      "step": 9900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7634781680066436e-06,
+      "loss": 1.1641,
+      "step": 9901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7619687722034995e-06,
+      "loss": 1.1904,
+      "step": 9902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7604599602292138e-06,
+      "loss": 1.1729,
+      "step": 9903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7589517321907112e-06,
+      "loss": 1.2168,
+      "step": 9904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7574440881948785e-06,
+      "loss": 1.2295,
+      "step": 9905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7559370283485677e-06,
+      "loss": 1.1709,
+      "step": 9906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7544305527585836e-06,
+      "loss": 1.1777,
+      "step": 9907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7529246615316864e-06,
+      "loss": 1.208,
+      "step": 9908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7514193547746006e-06,
+      "loss": 1.2305,
+      "step": 9909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7499146325940063e-06,
+      "loss": 1.2314,
+      "step": 9910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7484104950965464e-06,
+      "loss": 1.1533,
+      "step": 9911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7469069423888118e-06,
+      "loss": 1.1699,
+      "step": 9912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7454039745773665e-06,
+      "loss": 1.2207,
+      "step": 9913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.743901591768723e-06,
+      "loss": 1.1904,
+      "step": 9914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.742399794069355e-06,
+      "loss": 1.1426,
+      "step": 9915
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7408985815856972e-06,
+      "loss": 1.1777,
+      "step": 9916
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7393979544241357e-06,
+      "loss": 1.1826,
+      "step": 9917
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7378979126910224e-06,
+      "loss": 1.2285,
+      "step": 9918
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.736398456492665e-06,
+      "loss": 1.2012,
+      "step": 9919
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7348995859353302e-06,
+      "loss": 1.1631,
+      "step": 9920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7334013011252427e-06,
+      "loss": 1.2148,
+      "step": 9921
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7319036021685865e-06,
+      "loss": 1.1777,
+      "step": 9922
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7304064891715023e-06,
+      "loss": 1.2129,
+      "step": 9923
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.728909962240094e-06,
+      "loss": 1.1348,
+      "step": 9924
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7274140214804137e-06,
+      "loss": 1.2275,
+      "step": 9925
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.725918666998484e-06,
+      "loss": 1.2178,
+      "step": 9926
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.724423898900275e-06,
+      "loss": 1.1758,
+      "step": 9927
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7229297172917304e-06,
+      "loss": 1.1846,
+      "step": 9928
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7214361222787345e-06,
+      "loss": 1.2383,
+      "step": 9929
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7199431139671408e-06,
+      "loss": 1.2549,
+      "step": 9930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7184506924627587e-06,
+      "loss": 1.1963,
+      "step": 9931
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7169588578713593e-06,
+      "loss": 1.2197,
+      "step": 9932
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7154676102986623e-06,
+      "loss": 1.1885,
+      "step": 9933
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7139769498503545e-06,
+      "loss": 1.2461,
+      "step": 9934
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7124868766320813e-06,
+      "loss": 1.1787,
+      "step": 9935
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.710997390749447e-06,
+      "loss": 1.21,
+      "step": 9936
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.709508492308004e-06,
+      "loss": 1.2246,
+      "step": 9937
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7080201814132736e-06,
+      "loss": 1.1836,
+      "step": 9938
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7065324581707331e-06,
+      "loss": 1.252,
+      "step": 9939
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7050453226858187e-06,
+      "loss": 1.2305,
+      "step": 9940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.703558775063918e-06,
+      "loss": 1.1904,
+      "step": 9941
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7020728154103882e-06,
+      "loss": 1.2021,
+      "step": 9942
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.700587443830537e-06,
+      "loss": 1.1758,
+      "step": 9943
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6991026604296357e-06,
+      "loss": 1.1768,
+      "step": 9944
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6976184653129058e-06,
+      "loss": 1.2012,
+      "step": 9945
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.696134858585533e-06,
+      "loss": 1.1855,
+      "step": 9946
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6946518403526635e-06,
+      "loss": 1.0938,
+      "step": 9947
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6931694107193962e-06,
+      "loss": 1.1514,
+      "step": 9948
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.691687569790792e-06,
+      "loss": 1.2246,
+      "step": 9949
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6902063176718685e-06,
+      "loss": 1.2217,
+      "step": 9950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6887256544676013e-06,
+      "loss": 1.21,
+      "step": 9951
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.687245580282929e-06,
+      "loss": 1.1865,
+      "step": 9952
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6857660952227383e-06,
+      "loss": 1.2002,
+      "step": 9953
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6842871993918819e-06,
+      "loss": 1.2744,
+      "step": 9954
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6828088928951714e-06,
+      "loss": 1.2422,
+      "step": 9955
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.681331175837373e-06,
+      "loss": 1.2109,
+      "step": 9956
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798540483232119e-06,
+      "loss": 1.2178,
+      "step": 9957
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.678377510457373e-06,
+      "loss": 1.2305,
+      "step": 9958
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6769015623444973e-06,
+      "loss": 1.2129,
+      "step": 9959
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6754262040891888e-06,
+      "loss": 1.1992,
+      "step": 9960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6739514357960008e-06,
+      "loss": 1.2188,
+      "step": 9961
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6724772575694525e-06,
+      "loss": 1.1924,
+      "step": 9962
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6710036695140165e-06,
+      "loss": 1.2539,
+      "step": 9963
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.669530671734133e-06,
+      "loss": 1.1777,
+      "step": 9964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6680582643341858e-06,
+      "loss": 1.2236,
+      "step": 9965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6665864474185267e-06,
+      "loss": 1.2451,
+      "step": 9966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6651152210914645e-06,
+      "loss": 1.1885,
+      "step": 9967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6636445854572658e-06,
+      "loss": 1.2432,
+      "step": 9968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6621745406201506e-06,
+      "loss": 1.1914,
+      "step": 9969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.660705086684301e-06,
+      "loss": 1.209,
+      "step": 9970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6592362237538605e-06,
+      "loss": 1.1904,
+      "step": 9971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6577679519329292e-06,
+      "loss": 1.2295,
+      "step": 9972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6563002713255583e-06,
+      "loss": 1.2031,
+      "step": 9973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6548331820357633e-06,
+      "loss": 1.251,
+      "step": 9974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6533666841675177e-06,
+      "loss": 1.2285,
+      "step": 9975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6519007778247543e-06,
+      "loss": 1.166,
+      "step": 9976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650435463111355e-06,
+      "loss": 1.2402,
+      "step": 9977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6489707401311738e-06,
+      "loss": 1.2158,
+      "step": 9978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647506608988012e-06,
+      "loss": 1.2402,
+      "step": 9979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6460430697856367e-06,
+      "loss": 1.21,
+      "step": 9980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6445801226277624e-06,
+      "loss": 1.2148,
+      "step": 9981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.643117767618072e-06,
+      "loss": 1.1924,
+      "step": 9982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6416560048602015e-06,
+      "loss": 1.2148,
+      "step": 9983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6401948344577466e-06,
+      "loss": 1.251,
+      "step": 9984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6387342565142594e-06,
+      "loss": 1.1895,
+      "step": 9985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6372742711332511e-06,
+      "loss": 1.2051,
+      "step": 9986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6358148784181927e-06,
+      "loss": 1.186,
+      "step": 9987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.634356078472511e-06,
+      "loss": 1.2119,
+      "step": 9988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.632897871399588e-06,
+      "loss": 1.2363,
+      "step": 9989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6314402573027687e-06,
+      "loss": 1.208,
+      "step": 9990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.629983236285354e-06,
+      "loss": 1.168,
+      "step": 9991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6285268084506023e-06,
+      "loss": 1.2529,
+      "step": 9992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6270709739017321e-06,
+      "loss": 1.2256,
+      "step": 9993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6256157327419175e-06,
+      "loss": 1.2178,
+      "step": 9994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6241610850742918e-06,
+      "loss": 1.2529,
+      "step": 9995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6227070310019466e-06,
+      "loss": 1.1592,
+      "step": 9996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6212535706279275e-06,
+      "loss": 1.2051,
+      "step": 9997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6198007040552432e-06,
+      "loss": 1.1807,
+      "step": 9998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6183484313868559e-06,
+      "loss": 1.1885,
+      "step": 9999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.616896752725695e-06,
+      "loss": 1.208,
+      "step": 10000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6154456681746344e-06,
+      "loss": 1.1895,
+      "step": 10001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6139951778365138e-06,
+      "loss": 1.1943,
+      "step": 10002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6125452818141296e-06,
+      "loss": 1.1885,
+      "step": 10003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6110959802102388e-06,
+      "loss": 1.2012,
+      "step": 10004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.609647273127548e-06,
+      "loss": 1.2061,
+      "step": 10005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6081991606687275e-06,
+      "loss": 1.2344,
+      "step": 10006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6067516429364082e-06,
+      "loss": 1.2129,
+      "step": 10007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6053047200331784e-06,
+      "loss": 1.2285,
+      "step": 10008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6038583920615747e-06,
+      "loss": 1.1523,
+      "step": 10009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6024126591241007e-06,
+      "loss": 1.2148,
+      "step": 10010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6009675213232168e-06,
+      "loss": 1.2168,
+      "step": 10011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5995229787613397e-06,
+      "loss": 1.2715,
+      "step": 10012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5980790315408401e-06,
+      "loss": 1.1709,
+      "step": 10013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5966356797640547e-06,
+      "loss": 1.1943,
+      "step": 10014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5951929235332731e-06,
+      "loss": 1.2383,
+      "step": 10015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5937507629507454e-06,
+      "loss": 1.2139,
+      "step": 10016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5923091981186722e-06,
+      "loss": 1.1597,
+      "step": 10017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5908682291392208e-06,
+      "loss": 1.1826,
+      "step": 10018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5894278561145104e-06,
+      "loss": 1.1973,
+      "step": 10019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5879880791466228e-06,
+      "loss": 1.1836,
+      "step": 10020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5865488983375931e-06,
+      "loss": 1.1426,
+      "step": 10021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5851103137894164e-06,
+      "loss": 1.2285,
+      "step": 10022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5836723256040442e-06,
+      "loss": 1.1846,
+      "step": 10023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5822349338833908e-06,
+      "loss": 1.251,
+      "step": 10024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5807981387293193e-06,
+      "loss": 1.1768,
+      "step": 10025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5793619402436555e-06,
+      "loss": 1.2275,
+      "step": 10026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5779263385281852e-06,
+      "loss": 1.1582,
+      "step": 10027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5764913336846477e-06,
+      "loss": 1.1689,
+      "step": 10028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.575056925814743e-06,
+      "loss": 1.207,
+      "step": 10029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5736231150201265e-06,
+      "loss": 1.1641,
+      "step": 10030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5721899014024123e-06,
+      "loss": 1.2217,
+      "step": 10031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5707572850631747e-06,
+      "loss": 1.1553,
+      "step": 10032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5693252661039393e-06,
+      "loss": 1.1392,
+      "step": 10033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.567893844626195e-06,
+      "loss": 1.2168,
+      "step": 10034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5664630207313836e-06,
+      "loss": 1.2314,
+      "step": 10035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5650327945209153e-06,
+      "loss": 1.2314,
+      "step": 10036
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5636031660961427e-06,
+      "loss": 1.1924,
+      "step": 10037
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5621741355583863e-06,
+      "loss": 1.2188,
+      "step": 10038
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5607457030089212e-06,
+      "loss": 1.1963,
+      "step": 10039
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5593178685489818e-06,
+      "loss": 1.1719,
+      "step": 10040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.557890632279756e-06,
+      "loss": 1.167,
+      "step": 10041
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5564639943023906e-06,
+      "loss": 1.2324,
+      "step": 10042
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.555037954717995e-06,
+      "loss": 1.2129,
+      "step": 10043
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5536125136276347e-06,
+      "loss": 1.2002,
+      "step": 10044
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5521876711323258e-06,
+      "loss": 1.207,
+      "step": 10045
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5507634273330475e-06,
+      "loss": 1.2412,
+      "step": 10046
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5493397823307365e-06,
+      "loss": 1.2471,
+      "step": 10047
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.547916736226287e-06,
+      "loss": 1.2324,
+      "step": 10048
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5464942891205492e-06,
+      "loss": 1.1602,
+      "step": 10049
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5450724411143337e-06,
+      "loss": 1.2344,
+      "step": 10050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5436511923084052e-06,
+      "loss": 1.1816,
+      "step": 10051
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.542230542803488e-06,
+      "loss": 1.1787,
+      "step": 10052
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5408104927002666e-06,
+      "loss": 1.2295,
+      "step": 10053
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.539391042099373e-06,
+      "loss": 1.2119,
+      "step": 10054
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5379721911014078e-06,
+      "loss": 1.2188,
+      "step": 10055
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5365539398069252e-06,
+      "loss": 1.2119,
+      "step": 10056
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5351362883164344e-06,
+      "loss": 1.2148,
+      "step": 10057
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5337192367304066e-06,
+      "loss": 1.1943,
+      "step": 10058
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5323027851492677e-06,
+      "loss": 1.2275,
+      "step": 10059
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.530886933673401e-06,
+      "loss": 1.2363,
+      "step": 10060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5294716824031497e-06,
+      "loss": 1.2109,
+      "step": 10061
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5280570314388087e-06,
+      "loss": 1.2891,
+      "step": 10062
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.526642980880637e-06,
+      "loss": 1.2559,
+      "step": 10063
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5252295308288467e-06,
+      "loss": 1.2041,
+      "step": 10064
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5238166813836108e-06,
+      "loss": 1.1709,
+      "step": 10065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.522404432645056e-06,
+      "loss": 1.2246,
+      "step": 10066
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.52099278471327e-06,
+      "loss": 1.2422,
+      "step": 10067
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.519581737688295e-06,
+      "loss": 1.2441,
+      "step": 10068
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5181712916701353e-06,
+      "loss": 1.2021,
+      "step": 10069
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5167614467587433e-06,
+      "loss": 1.207,
+      "step": 10070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5153522030540357e-06,
+      "loss": 1.2148,
+      "step": 10071
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5139435606558894e-06,
+      "loss": 1.21,
+      "step": 10072
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5125355196641356e-06,
+      "loss": 1.1934,
+      "step": 10073
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5111280801785554e-06,
+      "loss": 1.1807,
+      "step": 10074
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5097212422988994e-06,
+      "loss": 1.1572,
+      "step": 10075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5083150061248675e-06,
+      "loss": 1.2139,
+      "step": 10076
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5069093717561234e-06,
+      "loss": 1.1729,
+      "step": 10077
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5055043392922774e-06,
+      "loss": 1.2451,
+      "step": 10078
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.50409990883291e-06,
+      "loss": 1.2129,
+      "step": 10079
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5026960804775515e-06,
+      "loss": 1.2246,
+      "step": 10080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5012928543256933e-06,
+      "loss": 1.2275,
+      "step": 10081
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.499890230476777e-06,
+      "loss": 1.2061,
+      "step": 10082
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4984882090302088e-06,
+      "loss": 1.1924,
+      "step": 10083
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4970867900853515e-06,
+      "loss": 1.1406,
+      "step": 10084
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4956859737415218e-06,
+      "loss": 1.1855,
+      "step": 10085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4942857600979954e-06,
+      "loss": 1.2217,
+      "step": 10086
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.492886149254007e-06,
+      "loss": 1.2168,
+      "step": 10087
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.491487141308745e-06,
+      "loss": 1.166,
+      "step": 10088
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4900887363613603e-06,
+      "loss": 1.1475,
+      "step": 10089
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4886909345109535e-06,
+      "loss": 1.2305,
+      "step": 10090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4872937358565887e-06,
+      "loss": 1.1924,
+      "step": 10091
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4858971404972855e-06,
+      "loss": 1.1982,
+      "step": 10092
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4845011485320204e-06,
+      "loss": 1.2188,
+      "step": 10093
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4831057600597266e-06,
+      "loss": 1.1953,
+      "step": 10094
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4817109751792957e-06,
+      "loss": 1.1807,
+      "step": 10095
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4803167939895768e-06,
+      "loss": 1.1904,
+      "step": 10096
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4789232165893775e-06,
+      "loss": 1.2217,
+      "step": 10097
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4775302430774541e-06,
+      "loss": 1.1523,
+      "step": 10098
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4761378735525312e-06,
+      "loss": 1.1943,
+      "step": 10099
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.474746108113283e-06,
+      "loss": 1.1768,
+      "step": 10100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4733549468583497e-06,
+      "loss": 1.1963,
+      "step": 10101
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4719643898863179e-06,
+      "loss": 1.2168,
+      "step": 10102
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4705744372957365e-06,
+      "loss": 1.1729,
+      "step": 10103
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4691850891851123e-06,
+      "loss": 1.2461,
+      "step": 10104
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.467796345652911e-06,
+      "loss": 1.2637,
+      "step": 10105
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4664082067975482e-06,
+      "loss": 1.1982,
+      "step": 10106
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.465020672717401e-06,
+      "loss": 1.1895,
+      "step": 10107
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4636337435108072e-06,
+      "loss": 1.2061,
+      "step": 10108
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4622474192760604e-06,
+      "loss": 1.209,
+      "step": 10109
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4608617001114033e-06,
+      "loss": 1.207,
+      "step": 10110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.459476586115045e-06,
+      "loss": 1.2734,
+      "step": 10111
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4580920773851482e-06,
+      "loss": 1.2197,
+      "step": 10112
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4567081740198351e-06,
+      "loss": 1.2227,
+      "step": 10113
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4553248761171757e-06,
+      "loss": 1.1787,
+      "step": 10114
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4539421837752122e-06,
+      "loss": 1.2422,
+      "step": 10115
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4525600970919317e-06,
+      "loss": 1.209,
+      "step": 10116
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4511786161652874e-06,
+      "loss": 1.2012,
+      "step": 10117
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4497977410931775e-06,
+      "loss": 1.2148,
+      "step": 10118
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484174719734688e-06,
+      "loss": 1.1904,
+      "step": 10119
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470378089039805e-06,
+      "loss": 1.1631,
+      "step": 10120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456587519824894e-06,
+      "loss": 1.2432,
+      "step": 10121
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4442803013067285e-06,
+      "loss": 1.2109,
+      "step": 10122
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4429024569743888e-06,
+      "loss": 1.2803,
+      "step": 10123
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4415252190831187e-06,
+      "loss": 1.2256,
+      "step": 10124
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4401485877305244e-06,
+      "loss": 1.2021,
+      "step": 10125
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4387725630141637e-06,
+      "loss": 1.1953,
+      "step": 10126
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4373971450315561e-06,
+      "loss": 1.1729,
+      "step": 10127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4360223338801804e-06,
+      "loss": 1.1982,
+      "step": 10128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4346481296574665e-06,
+      "loss": 1.1934,
+      "step": 10129
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4332745324608065e-06,
+      "loss": 1.2041,
+      "step": 10130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.431901542387546e-06,
+      "loss": 1.2041,
+      "step": 10131
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.430529159534988e-06,
+      "loss": 1.1396,
+      "step": 10132
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4291573840003968e-06,
+      "loss": 1.2393,
+      "step": 10133
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4277862158809852e-06,
+      "loss": 1.2012,
+      "step": 10134
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4264156552739293e-06,
+      "loss": 1.1348,
+      "step": 10135
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4250457022763598e-06,
+      "loss": 1.1973,
+      "step": 10136
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4236763569853706e-06,
+      "loss": 1.2432,
+      "step": 10137
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4223076194980013e-06,
+      "loss": 1.1807,
+      "step": 10138
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.420939489911256e-06,
+      "loss": 1.1426,
+      "step": 10139
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4195719683220933e-06,
+      "loss": 1.2109,
+      "step": 10140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.418205054827434e-06,
+      "loss": 1.2012,
+      "step": 10141
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4168387495241442e-06,
+      "loss": 1.2363,
+      "step": 10142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4154730525090542e-06,
+      "loss": 1.2471,
+      "step": 10143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.414107963878958e-06,
+      "loss": 1.1816,
+      "step": 10144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4127434837305953e-06,
+      "loss": 1.2021,
+      "step": 10145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.411379612160666e-06,
+      "loss": 1.251,
+      "step": 10146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410016349265827e-06,
+      "loss": 1.1934,
+      "step": 10147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4086536951426944e-06,
+      "loss": 1.2783,
+      "step": 10148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4072916498878408e-06,
+      "loss": 1.2012,
+      "step": 10149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4059302135977882e-06,
+      "loss": 1.2197,
+      "step": 10150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4045693863690279e-06,
+      "loss": 1.2598,
+      "step": 10151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4032091682979997e-06,
+      "loss": 1.2217,
+      "step": 10152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4018495594811032e-06,
+      "loss": 1.2012,
+      "step": 10153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004905600146912e-06,
+      "loss": 1.1807,
+      "step": 10154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3991321699950767e-06,
+      "loss": 1.1963,
+      "step": 10155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3977743895185302e-06,
+      "loss": 1.1865,
+      "step": 10156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3964172186812752e-06,
+      "loss": 1.2188,
+      "step": 10157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3950606575794966e-06,
+      "loss": 1.2139,
+      "step": 10158
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3937047063093335e-06,
+      "loss": 1.2246,
+      "step": 10159
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3923493649668806e-06,
+      "loss": 1.2266,
+      "step": 10160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.390994633648195e-06,
+      "loss": 1.2188,
+      "step": 10161
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3896405124492796e-06,
+      "loss": 1.2129,
+      "step": 10162
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3882870014661054e-06,
+      "loss": 1.2402,
+      "step": 10163
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3869341007945957e-06,
+      "loss": 1.1953,
+      "step": 10164
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3855818105306284e-06,
+      "loss": 1.2266,
+      "step": 10165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3842301307700424e-06,
+      "loss": 1.1914,
+      "step": 10166
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3828790616086297e-06,
+      "loss": 1.1846,
+      "step": 10167
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3815286031421405e-06,
+      "loss": 1.1943,
+      "step": 10168
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.380178755466285e-06,
+      "loss": 1.2119,
+      "step": 10169
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.378829518676722e-06,
+      "loss": 1.2207,
+      "step": 10170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.377480892869074e-06,
+      "loss": 1.2041,
+      "step": 10171
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.376132878138916e-06,
+      "loss": 1.1924,
+      "step": 10172
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.374785474581788e-06,
+      "loss": 1.1904,
+      "step": 10173
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3734386822931733e-06,
+      "loss": 1.165,
+      "step": 10174
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3720925013685226e-06,
+      "loss": 1.2012,
+      "step": 10175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3707469319032384e-06,
+      "loss": 1.2529,
+      "step": 10176
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3694019739926845e-06,
+      "loss": 1.2695,
+      "step": 10177
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3680576277321721e-06,
+      "loss": 1.1904,
+      "step": 10178
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3667138932169765e-06,
+      "loss": 1.21,
+      "step": 10179
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3653707705423314e-06,
+      "loss": 1.1553,
+      "step": 10180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3640282598034227e-06,
+      "loss": 1.1875,
+      "step": 10181
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3626863610953955e-06,
+      "loss": 1.207,
+      "step": 10182
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.361345074513345e-06,
+      "loss": 1.1904,
+      "step": 10183
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3600044001523317e-06,
+      "loss": 1.2891,
+      "step": 10184
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3586643381073682e-06,
+      "loss": 1.2432,
+      "step": 10185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3573248884734257e-06,
+      "loss": 1.1455,
+      "step": 10186
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3559860513454303e-06,
+      "loss": 1.2148,
+      "step": 10187
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3546478268182639e-06,
+      "loss": 1.1865,
+      "step": 10188
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3533102149867695e-06,
+      "loss": 1.1738,
+      "step": 10189
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3519732159457422e-06,
+      "loss": 1.2002,
+      "step": 10190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3506368297899342e-06,
+      "loss": 1.2305,
+      "step": 10191
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3493010566140542e-06,
+      "loss": 1.166,
+      "step": 10192
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3479658965127707e-06,
+      "loss": 1.1826,
+      "step": 10193
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3466313495807049e-06,
+      "loss": 1.2422,
+      "step": 10194
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3452974159124365e-06,
+      "loss": 1.1973,
+      "step": 10195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3439640956025024e-06,
+      "loss": 1.1836,
+      "step": 10196
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3426313887453934e-06,
+      "loss": 1.2178,
+      "step": 10197
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3412992954355607e-06,
+      "loss": 1.1924,
+      "step": 10198
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3399678157674067e-06,
+      "loss": 1.1689,
+      "step": 10199
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3386369498352936e-06,
+      "loss": 1.2207,
+      "step": 10200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3373066977335415e-06,
+      "loss": 1.2002,
+      "step": 10201
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3359770595564247e-06,
+      "loss": 1.1738,
+      "step": 10202
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3346480353981739e-06,
+      "loss": 1.166,
+      "step": 10203
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3333196253529778e-06,
+      "loss": 1.1924,
+      "step": 10204
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3319918295149804e-06,
+      "loss": 1.2305,
+      "step": 10205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3306646479782847e-06,
+      "loss": 1.2422,
+      "step": 10206
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.329338080836945e-06,
+      "loss": 1.209,
+      "step": 10207
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3280121281849723e-06,
+      "loss": 1.2041,
+      "step": 10208
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3266867901163428e-06,
+      "loss": 1.207,
+      "step": 10209
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3253620667249834e-06,
+      "loss": 1.2676,
+      "step": 10210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3240379581047725e-06,
+      "loss": 1.1709,
+      "step": 10211
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3227144643495504e-06,
+      "loss": 1.1816,
+      "step": 10212
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3213915855531157e-06,
+      "loss": 1.1924,
+      "step": 10213
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3200693218092208e-06,
+      "loss": 1.2275,
+      "step": 10214
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.318747673211569e-06,
+      "loss": 1.209,
+      "step": 10215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3174266398538315e-06,
+      "loss": 1.1797,
+      "step": 10216
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3161062218296283e-06,
+      "loss": 1.1924,
+      "step": 10217
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3147864192325388e-06,
+      "loss": 1.1943,
+      "step": 10218
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3134672321560937e-06,
+      "loss": 1.1787,
+      "step": 10219
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.312148660693785e-06,
+      "loss": 1.2109,
+      "step": 10220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3108307049390601e-06,
+      "loss": 1.1885,
+      "step": 10221
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.309513364985323e-06,
+      "loss": 1.2188,
+      "step": 10222
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3081966409259327e-06,
+      "loss": 1.1914,
+      "step": 10223
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3068805328542067e-06,
+      "loss": 1.2227,
+      "step": 10224
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3055650408634157e-06,
+      "loss": 1.1748,
+      "step": 10225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3042501650467921e-06,
+      "loss": 1.167,
+      "step": 10226
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3029359054975166e-06,
+      "loss": 1.2256,
+      "step": 10227
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3016222623087315e-06,
+      "loss": 1.2295,
+      "step": 10228
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.300309235573537e-06,
+      "loss": 1.2568,
+      "step": 10229
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.298996825384985e-06,
+      "loss": 1.1953,
+      "step": 10230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.297685031836088e-06,
+      "loss": 1.1865,
+      "step": 10231
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.29637385501981e-06,
+      "loss": 1.2246,
+      "step": 10232
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2950632950290764e-06,
+      "loss": 1.2148,
+      "step": 10233
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2937533519567691e-06,
+      "loss": 1.2607,
+      "step": 10234
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.292444025895717e-06,
+      "loss": 1.2314,
+      "step": 10235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2911353169387154e-06,
+      "loss": 1.2002,
+      "step": 10236
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2898272251785127e-06,
+      "loss": 1.2256,
+      "step": 10237
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2885197507078128e-06,
+      "loss": 1.207,
+      "step": 10238
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2872128936192762e-06,
+      "loss": 1.2627,
+      "step": 10239
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2859066540055197e-06,
+      "loss": 1.2754,
+      "step": 10240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2846010319591185e-06,
+      "loss": 1.1904,
+      "step": 10241
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2832960275726014e-06,
+      "loss": 1.1836,
+      "step": 10242
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2819916409384514e-06,
+      "loss": 1.167,
+      "step": 10243
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2806878721491088e-06,
+      "loss": 1.1807,
+      "step": 10244
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2793847212969767e-06,
+      "loss": 1.1602,
+      "step": 10245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.278082188474411e-06,
+      "loss": 1.2188,
+      "step": 10246
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2767802737737145e-06,
+      "loss": 1.2471,
+      "step": 10247
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.275478977287159e-06,
+      "loss": 1.167,
+      "step": 10248
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2741782991069662e-06,
+      "loss": 1.1953,
+      "step": 10249
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2728782393253158e-06,
+      "loss": 1.2021,
+      "step": 10250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2715787980343385e-06,
+      "loss": 1.1797,
+      "step": 10251
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2702799753261319e-06,
+      "loss": 1.2373,
+      "step": 10252
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.268981771292741e-06,
+      "loss": 1.1807,
+      "step": 10253
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2676841860261712e-06,
+      "loss": 1.208,
+      "step": 10254
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2663872196183779e-06,
+      "loss": 1.1758,
+      "step": 10255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2650908721612798e-06,
+      "loss": 1.2236,
+      "step": 10256
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2637951437467488e-06,
+      "loss": 1.1943,
+      "step": 10257
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2625000344666128e-06,
+      "loss": 1.1748,
+      "step": 10258
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.261205544412657e-06,
+      "loss": 1.1992,
+      "step": 10259
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2599116736766215e-06,
+      "loss": 1.2207,
+      "step": 10260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2586184223502018e-06,
+      "loss": 1.2246,
+      "step": 10261
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2573257905250535e-06,
+      "loss": 1.1914,
+      "step": 10262
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2560337782927812e-06,
+      "loss": 1.2617,
+      "step": 10263
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2547423857449525e-06,
+      "loss": 1.1797,
+      "step": 10264
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2534516129730867e-06,
+      "loss": 1.2246,
+      "step": 10265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2521614600686616e-06,
+      "loss": 1.1924,
+      "step": 10266
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2508719271231106e-06,
+      "loss": 1.1709,
+      "step": 10267
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2495830142278231e-06,
+      "loss": 1.1855,
+      "step": 10268
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2482947214741436e-06,
+      "loss": 1.2422,
+      "step": 10269
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2470070489533748e-06,
+      "loss": 1.2227,
+      "step": 10270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2457199967567713e-06,
+      "loss": 1.1826,
+      "step": 10271
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2444335649755478e-06,
+      "loss": 1.2021,
+      "step": 10272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2431477537008718e-06,
+      "loss": 1.2012,
+      "step": 10273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2418625630238746e-06,
+      "loss": 1.2188,
+      "step": 10274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.240577993035631e-06,
+      "loss": 1.208,
+      "step": 10275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2392940438271806e-06,
+      "loss": 1.2041,
+      "step": 10276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2380107154895182e-06,
+      "loss": 1.1943,
+      "step": 10277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2367280081135936e-06,
+      "loss": 1.1851,
+      "step": 10278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2354459217903092e-06,
+      "loss": 1.1641,
+      "step": 10279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2341644566105259e-06,
+      "loss": 1.1982,
+      "step": 10280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2328836126650656e-06,
+      "loss": 1.1836,
+      "step": 10281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2316033900447e-06,
+      "loss": 1.1562,
+      "step": 10282
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2303237888401565e-06,
+      "loss": 1.2217,
+      "step": 10283
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2290448091421214e-06,
+      "loss": 1.1875,
+      "step": 10284
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2277664510412358e-06,
+      "loss": 1.2109,
+      "step": 10285
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2264887146281e-06,
+      "loss": 1.2598,
+      "step": 10286
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.225211599993259e-06,
+      "loss": 1.2168,
+      "step": 10287
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.22393510722723e-06,
+      "loss": 1.2285,
+      "step": 10288
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2226592364204749e-06,
+      "loss": 1.1914,
+      "step": 10289
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.221383987663416e-06,
+      "loss": 1.2109,
+      "step": 10290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2201093610464277e-06,
+      "loss": 1.1543,
+      "step": 10291
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2188353566598431e-06,
+      "loss": 1.208,
+      "step": 10292
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2175619745939515e-06,
+      "loss": 1.1885,
+      "step": 10293
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2162892149389982e-06,
+      "loss": 1.165,
+      "step": 10294
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2150170777851833e-06,
+      "loss": 1.2227,
+      "step": 10295
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2137455632226614e-06,
+      "loss": 1.1826,
+      "step": 10296
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.212474671341547e-06,
+      "loss": 1.1709,
+      "step": 10297
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2112044022319092e-06,
+      "loss": 1.2031,
+      "step": 10298
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2099347559837672e-06,
+      "loss": 1.2207,
+      "step": 10299
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2086657326871044e-06,
+      "loss": 1.1953,
+      "step": 10300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2073973324318544e-06,
+      "loss": 1.2188,
+      "step": 10301
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.20612955530791e-06,
+      "loss": 1.2109,
+      "step": 10302
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2048624014051191e-06,
+      "loss": 1.1914,
+      "step": 10303
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2035958708132834e-06,
+      "loss": 1.1924,
+      "step": 10304
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2023299636221619e-06,
+      "loss": 1.1885,
+      "step": 10305
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.201064679921472e-06,
+      "loss": 1.2227,
+      "step": 10306
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1998000198008808e-06,
+      "loss": 1.1816,
+      "step": 10307
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1985359833500154e-06,
+      "loss": 1.1855,
+      "step": 10308
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1972725706584565e-06,
+      "loss": 1.2363,
+      "step": 10309
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1960097818157468e-06,
+      "loss": 1.1807,
+      "step": 10310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1947476169113792e-06,
+      "loss": 1.1562,
+      "step": 10311
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1934860760348e-06,
+      "loss": 1.21,
+      "step": 10312
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1922251592754165e-06,
+      "loss": 1.1914,
+      "step": 10313
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190964866722588e-06,
+      "loss": 1.2451,
+      "step": 10314
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189705198465636e-06,
+      "loss": 1.2129,
+      "step": 10315
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1884461545938252e-06,
+      "loss": 1.1807,
+      "step": 10316
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1871877351963913e-06,
+      "loss": 1.2061,
+      "step": 10317
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1859299403625158e-06,
+      "loss": 1.1699,
+      "step": 10318
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184672770181341e-06,
+      "loss": 1.209,
+      "step": 10319
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1834162247419568e-06,
+      "loss": 1.2119,
+      "step": 10320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1821603041334184e-06,
+      "loss": 1.2021,
+      "step": 10321
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1809050084447327e-06,
+      "loss": 1.2285,
+      "step": 10322
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1796503377648615e-06,
+      "loss": 1.167,
+      "step": 10323
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1783962921827242e-06,
+      "loss": 1.2246,
+      "step": 10324
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1771428717871936e-06,
+      "loss": 1.2061,
+      "step": 10325
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1758900766671012e-06,
+      "loss": 1.1768,
+      "step": 10326
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1746379069112334e-06,
+      "loss": 1.2227,
+      "step": 10327
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1733863626083276e-06,
+      "loss": 1.1982,
+      "step": 10328
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1721354438470832e-06,
+      "loss": 1.1768,
+      "step": 10329
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1708851507161522e-06,
+      "loss": 1.2158,
+      "step": 10330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1696354833041423e-06,
+      "loss": 1.2285,
+      "step": 10331
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1683864416996193e-06,
+      "loss": 1.1738,
+      "step": 10332
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1671380259911003e-06,
+      "loss": 1.2051,
+      "step": 10333
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1658902362670622e-06,
+      "loss": 1.1719,
+      "step": 10334
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1646430726159374e-06,
+      "loss": 1.2109,
+      "step": 10335
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.163396535126108e-06,
+      "loss": 1.1846,
+      "step": 10336
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.162150623885917e-06,
+      "loss": 1.2432,
+      "step": 10337
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1609053389836645e-06,
+      "loss": 1.1851,
+      "step": 10338
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1596606805076027e-06,
+      "loss": 1.2051,
+      "step": 10339
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1584166485459392e-06,
+      "loss": 1.1299,
+      "step": 10340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.157173243186841e-06,
+      "loss": 1.1455,
+      "step": 10341
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1559304645184254e-06,
+      "loss": 1.2373,
+      "step": 10342
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1546883126287721e-06,
+      "loss": 1.2178,
+      "step": 10343
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1534467876059075e-06,
+      "loss": 1.1934,
+      "step": 10344
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1522058895378185e-06,
+      "loss": 1.2451,
+      "step": 10345
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1509656185124518e-06,
+      "loss": 1.2139,
+      "step": 10346
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1497259746177047e-06,
+      "loss": 1.2314,
+      "step": 10347
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1484869579414281e-06,
+      "loss": 1.1875,
+      "step": 10348
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1472485685714307e-06,
+      "loss": 1.167,
+      "step": 10349
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.14601080659548e-06,
+      "loss": 1.1865,
+      "step": 10350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1447736721012959e-06,
+      "loss": 1.2139,
+      "step": 10351
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143537165176548e-06,
+      "loss": 1.2432,
+      "step": 10352
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1423012859088744e-06,
+      "loss": 1.2002,
+      "step": 10353
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1410660343858604e-06,
+      "loss": 1.2021,
+      "step": 10354
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1398314106950493e-06,
+      "loss": 1.1787,
+      "step": 10355
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1385974149239344e-06,
+      "loss": 1.25,
+      "step": 10356
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1373640471599724e-06,
+      "loss": 1.2119,
+      "step": 10357
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.13613130749057e-06,
+      "loss": 1.1865,
+      "step": 10358
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1348991960030931e-06,
+      "loss": 1.208,
+      "step": 10359
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1336677127848605e-06,
+      "loss": 1.2617,
+      "step": 10360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.132436857923147e-06,
+      "loss": 1.1572,
+      "step": 10361
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1312066315051852e-06,
+      "loss": 1.2266,
+      "step": 10362
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1299770336181604e-06,
+      "loss": 1.1729,
+      "step": 10363
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1287480643492122e-06,
+      "loss": 1.2549,
+      "step": 10364
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1275197237854395e-06,
+      "loss": 1.1992,
+      "step": 10365
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1262920120138932e-06,
+      "loss": 1.2119,
+      "step": 10366
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.125064929121583e-06,
+      "loss": 1.1562,
+      "step": 10367
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1238384751954711e-06,
+      "loss": 1.1875,
+      "step": 10368
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1226126503224767e-06,
+      "loss": 1.1816,
+      "step": 10369
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.121387454589473e-06,
+      "loss": 1.2441,
+      "step": 10370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1201628880832927e-06,
+      "loss": 1.2012,
+      "step": 10371
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1189389508907166e-06,
+      "loss": 1.1621,
+      "step": 10372
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.117715643098486e-06,
+      "loss": 1.1504,
+      "step": 10373
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1164929647932988e-06,
+      "loss": 1.1602,
+      "step": 10374
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1152709160618036e-06,
+      "loss": 1.1885,
+      "step": 10375
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.114049496990609e-06,
+      "loss": 1.2207,
+      "step": 10376
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.112828707666276e-06,
+      "loss": 1.2031,
+      "step": 10377
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1116085481753214e-06,
+      "loss": 1.2031,
+      "step": 10378
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.11038901860422e-06,
+      "loss": 1.2178,
+      "step": 10379
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1091701190393966e-06,
+      "loss": 1.166,
+      "step": 10380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1079518495672348e-06,
+      "loss": 1.1777,
+      "step": 10381
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1067342102740763e-06,
+      "loss": 1.2109,
+      "step": 10382
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1055172012462156e-06,
+      "loss": 1.2178,
+      "step": 10383
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.104300822569897e-06,
+      "loss": 1.1865,
+      "step": 10384
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1030850743313294e-06,
+      "loss": 1.2158,
+      "step": 10385
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1018699566166714e-06,
+      "loss": 1.1738,
+      "step": 10386
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1006554695120397e-06,
+      "loss": 1.2031,
+      "step": 10387
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0994416131035e-06,
+      "loss": 1.2324,
+      "step": 10388
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0982283874770838e-06,
+      "loss": 1.1816,
+      "step": 10389
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0970157927187707e-06,
+      "loss": 1.2363,
+      "step": 10390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.095803828914499e-06,
+      "loss": 1.2217,
+      "step": 10391
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0945924961501565e-06,
+      "loss": 1.1719,
+      "step": 10392
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093381794511592e-06,
+      "loss": 1.1777,
+      "step": 10393
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0921717240846097e-06,
+      "loss": 1.2402,
+      "step": 10394
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.090962284954964e-06,
+      "loss": 1.2627,
+      "step": 10395
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0897534772083706e-06,
+      "loss": 1.2275,
+      "step": 10396
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0885453009304958e-06,
+      "loss": 1.2344,
+      "step": 10397
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0873377562069641e-06,
+      "loss": 1.2256,
+      "step": 10398
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0861308431233554e-06,
+      "loss": 1.2129,
+      "step": 10399
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0849245617652005e-06,
+      "loss": 1.1992,
+      "step": 10400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0837189122179904e-06,
+      "loss": 1.2236,
+      "step": 10401
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0825138945671676e-06,
+      "loss": 1.2168,
+      "step": 10402
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.081309508898134e-06,
+      "loss": 1.1953,
+      "step": 10403
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0801057552962435e-06,
+      "loss": 1.1924,
+      "step": 10404
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0789026338468056e-06,
+      "loss": 1.209,
+      "step": 10405
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0777001446350854e-06,
+      "loss": 1.1865,
+      "step": 10406
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.076498287746306e-06,
+      "loss": 1.2334,
+      "step": 10407
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.075297063265638e-06,
+      "loss": 1.1797,
+      "step": 10408
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0740964712782154e-06,
+      "loss": 1.2539,
+      "step": 10409
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0728965118691236e-06,
+      "loss": 1.2227,
+      "step": 10410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0716971851234026e-06,
+      "loss": 1.208,
+      "step": 10411
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0704984911260496e-06,
+      "loss": 1.2363,
+      "step": 10412
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0693004299620158e-06,
+      "loss": 1.2324,
+      "step": 10413
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0681030017162087e-06,
+      "loss": 1.1611,
+      "step": 10414
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0669062064734903e-06,
+      "loss": 1.207,
+      "step": 10415
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0657100443186742e-06,
+      "loss": 1.2217,
+      "step": 10416
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.064514515336532e-06,
+      "loss": 1.2139,
+      "step": 10417
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0633196196117956e-06,
+      "loss": 1.1865,
+      "step": 10418
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0621253572291467e-06,
+      "loss": 1.2031,
+      "step": 10419
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.060931728273219e-06,
+      "loss": 1.1719,
+      "step": 10420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0597387328286056e-06,
+      "loss": 1.1885,
+      "step": 10421
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.058546370979856e-06,
+      "loss": 1.166,
+      "step": 10422
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0573546428114735e-06,
+      "loss": 1.2188,
+      "step": 10423
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0561635484079102e-06,
+      "loss": 1.1816,
+      "step": 10424
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0549730878535847e-06,
+      "loss": 1.2363,
+      "step": 10425
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0537832612328624e-06,
+      "loss": 1.25,
+      "step": 10426
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0525940686300694e-06,
+      "loss": 1.1836,
+      "step": 10427
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0514055101294784e-06,
+      "loss": 1.2422,
+      "step": 10428
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0502175858153262e-06,
+      "loss": 1.1846,
+      "step": 10429
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.049030295771799e-06,
+      "loss": 1.2344,
+      "step": 10430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0478436400830417e-06,
+      "loss": 1.209,
+      "step": 10431
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0466576188331523e-06,
+      "loss": 1.1836,
+      "step": 10432
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0454722321061828e-06,
+      "loss": 1.2461,
+      "step": 10433
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0442874799861423e-06,
+      "loss": 1.2188,
+      "step": 10434
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0431033625569963e-06,
+      "loss": 1.2061,
+      "step": 10435
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0419198799026586e-06,
+      "loss": 1.2285,
+      "step": 10436
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0407370321070054e-06,
+      "loss": 1.167,
+      "step": 10437
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0395548192538641e-06,
+      "loss": 1.1885,
+      "step": 10438
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0383732414270186e-06,
+      "loss": 1.2295,
+      "step": 10439
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0371922987102069e-06,
+      "loss": 1.1943,
+      "step": 10440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0360119911871224e-06,
+      "loss": 1.1826,
+      "step": 10441
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0348323189414144e-06,
+      "loss": 1.2061,
+      "step": 10442
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0336532820566859e-06,
+      "loss": 1.2031,
+      "step": 10443
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0324748806164954e-06,
+      "loss": 1.1816,
+      "step": 10444
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0312971147043548e-06,
+      "loss": 1.1787,
+      "step": 10445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0301199844037302e-06,
+      "loss": 1.1953,
+      "step": 10446
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.02894348979805e-06,
+      "loss": 1.1934,
+      "step": 10447
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0277676309706907e-06,
+      "loss": 1.1934,
+      "step": 10448
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0265924080049838e-06,
+      "loss": 1.1914,
+      "step": 10449
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.025417820984218e-06,
+      "loss": 1.2246,
+      "step": 10450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0242438699916357e-06,
+      "loss": 1.2285,
+      "step": 10451
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.023070555110437e-06,
+      "loss": 1.1641,
+      "step": 10452
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0218978764237698e-06,
+      "loss": 1.1602,
+      "step": 10453
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0207258340147464e-06,
+      "loss": 1.1738,
+      "step": 10454
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0195544279664283e-06,
+      "loss": 1.2178,
+      "step": 10455
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0183836583618346e-06,
+      "loss": 1.249,
+      "step": 10456
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0172135252839333e-06,
+      "loss": 1.1865,
+      "step": 10457
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0160440288156548e-06,
+      "loss": 1.2168,
+      "step": 10458
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0148751690398794e-06,
+      "loss": 1.1992,
+      "step": 10459
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0137069460394467e-06,
+      "loss": 1.1582,
+      "step": 10460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0125393598971468e-06,
+      "loss": 1.1963,
+      "step": 10461
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0113724106957256e-06,
+      "loss": 1.2002,
+      "step": 10462
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0102060985178874e-06,
+      "loss": 1.1924,
+      "step": 10463
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.009040423446288e-06,
+      "loss": 1.1816,
+      "step": 10464
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007875385563537e-06,
+      "loss": 1.1992,
+      "step": 10465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0067109849522017e-06,
+      "loss": 1.2207,
+      "step": 10466
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055472216948025e-06,
+      "loss": 1.2471,
+      "step": 10467
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004384095873815e-06,
+      "loss": 1.2148,
+      "step": 10468
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0032216075716715e-06,
+      "loss": 1.2422,
+      "step": 10469
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0020597568707569e-06,
+      "loss": 1.2139,
+      "step": 10470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0008985438534102e-06,
+      "loss": 1.2031,
+      "step": 10471
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.997379686019304e-07,
+      "loss": 1.2168,
+      "step": 10472
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.985780311985616e-07,
+      "loss": 1.1992,
+      "step": 10473
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.974187317255124e-07,
+      "loss": 1.2158,
+      "step": 10474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.962600702649406e-07,
+      "loss": 1.2227,
+      "step": 10475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.951020468989614e-07,
+      "loss": 1.1826,
+      "step": 10476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.939446617096428e-07,
+      "loss": 1.2168,
+      "step": 10477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.92787914779011e-07,
+      "loss": 1.1904,
+      "step": 10478
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.916318061890418e-07,
+      "loss": 1.2188,
+      "step": 10479
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.904763360216718e-07,
+      "loss": 1.1875,
+      "step": 10480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.893215043587857e-07,
+      "loss": 1.2637,
+      "step": 10481
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.881673112822255e-07,
+      "loss": 1.1797,
+      "step": 10482
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.870137568737936e-07,
+      "loss": 1.1729,
+      "step": 10483
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.858608412152415e-07,
+      "loss": 1.2051,
+      "step": 10484
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.847085643882725e-07,
+      "loss": 1.1836,
+      "step": 10485
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.835569264745504e-07,
+      "loss": 1.1758,
+      "step": 10486
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.824059275556919e-07,
+      "loss": 1.2129,
+      "step": 10487
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.812555677132707e-07,
+      "loss": 1.1641,
+      "step": 10488
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.80105847028805e-07,
+      "loss": 1.2344,
+      "step": 10489
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.78956765583784e-07,
+      "loss": 1.2383,
+      "step": 10490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.778083234596392e-07,
+      "loss": 1.1816,
+      "step": 10491
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.766605207377621e-07,
+      "loss": 1.1992,
+      "step": 10492
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.755133574994956e-07,
+      "loss": 1.1836,
+      "step": 10493
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.743668338261403e-07,
+      "loss": 1.1748,
+      "step": 10494
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.732209497989487e-07,
+      "loss": 1.2344,
+      "step": 10495
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.720757054991314e-07,
+      "loss": 1.2363,
+      "step": 10496
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.709311010078515e-07,
+      "loss": 1.167,
+      "step": 10497
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.697871364062272e-07,
+      "loss": 1.2002,
+      "step": 10498
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.686438117753294e-07,
+      "loss": 1.2266,
+      "step": 10499
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.675011271961898e-07,
+      "loss": 1.2197,
+      "step": 10500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.663590827497849e-07,
+      "loss": 1.2227,
+      "step": 10501
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.652176785170542e-07,
+      "loss": 1.1738,
+      "step": 10502
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.640769145788874e-07,
+      "loss": 1.2402,
+      "step": 10503
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.629367910161325e-07,
+      "loss": 1.1582,
+      "step": 10504
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.617973079095898e-07,
+      "loss": 1.2295,
+      "step": 10505
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.60658465340013e-07,
+      "loss": 1.2832,
+      "step": 10506
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.595202633881118e-07,
+      "loss": 1.1924,
+      "step": 10507
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.583827021345537e-07,
+      "loss": 1.166,
+      "step": 10508
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.572457816599535e-07,
+      "loss": 1.1816,
+      "step": 10509
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.561095020448862e-07,
+      "loss": 1.1602,
+      "step": 10510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.549738633698801e-07,
+      "loss": 1.1973,
+      "step": 10511
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.538388657154173e-07,
+      "loss": 1.1592,
+      "step": 10512
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.527045091619369e-07,
+      "loss": 1.2227,
+      "step": 10513
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.515707937898278e-07,
+      "loss": 1.2393,
+      "step": 10514
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.504377196794379e-07,
+      "loss": 1.2139,
+      "step": 10515
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.493052869110708e-07,
+      "loss": 1.2373,
+      "step": 10516
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.481734955649769e-07,
+      "loss": 1.1787,
+      "step": 10517
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.470423457213663e-07,
+      "loss": 1.2002,
+      "step": 10518
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.459118374604071e-07,
+      "loss": 1.2197,
+      "step": 10519
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.447819708622197e-07,
+      "loss": 1.1816,
+      "step": 10520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.436527460068734e-07,
+      "loss": 1.2305,
+      "step": 10521
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.425241629743975e-07,
+      "loss": 1.21,
+      "step": 10522
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.413962218447747e-07,
+      "loss": 1.2178,
+      "step": 10523
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.402689226979445e-07,
+      "loss": 1.2539,
+      "step": 10524
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.391422656137927e-07,
+      "loss": 1.1729,
+      "step": 10525
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.380162506721701e-07,
+      "loss": 1.1846,
+      "step": 10526
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.368908779528774e-07,
+      "loss": 1.1924,
+      "step": 10527
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.357661475356705e-07,
+      "loss": 1.1602,
+      "step": 10528
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.346420595002559e-07,
+      "loss": 1.1709,
+      "step": 10529
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.335186139262986e-07,
+      "loss": 1.167,
+      "step": 10530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.323958108934172e-07,
+      "loss": 1.2217,
+      "step": 10531
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.312736504811848e-07,
+      "loss": 1.2139,
+      "step": 10532
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.301521327691299e-07,
+      "loss": 1.167,
+      "step": 10533
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.290312578367322e-07,
+      "loss": 1.1846,
+      "step": 10534
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.279110257634305e-07,
+      "loss": 1.1455,
+      "step": 10535
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.267914366286168e-07,
+      "loss": 1.1807,
+      "step": 10536
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.25672490511631e-07,
+      "loss": 1.165,
+      "step": 10537
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.245541874917763e-07,
+      "loss": 1.1875,
+      "step": 10538
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.23436527648306e-07,
+      "loss": 1.21,
+      "step": 10539
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.223195110604288e-07,
+      "loss": 1.2715,
+      "step": 10540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.212031378073083e-07,
+      "loss": 1.1455,
+      "step": 10541
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.200874079680599e-07,
+      "loss": 1.2109,
+      "step": 10542
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.189723216217583e-07,
+      "loss": 1.2305,
+      "step": 10543
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.178578788474279e-07,
+      "loss": 1.1826,
+      "step": 10544
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.16744079724049e-07,
+      "loss": 1.2266,
+      "step": 10545
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.156309243305561e-07,
+      "loss": 1.1279,
+      "step": 10546
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.145184127458384e-07,
+      "loss": 1.248,
+      "step": 10547
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.134065450487406e-07,
+      "loss": 1.1758,
+      "step": 10548
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.122953213180608e-07,
+      "loss": 1.1934,
+      "step": 10549
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.111847416325503e-07,
+      "loss": 1.1855,
+      "step": 10550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.100748060709163e-07,
+      "loss": 1.21,
+      "step": 10551
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.089655147118226e-07,
+      "loss": 1.1982,
+      "step": 10552
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.078568676338795e-07,
+      "loss": 1.2119,
+      "step": 10553
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.067488649156575e-07,
+      "loss": 1.1777,
+      "step": 10554
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.056415066356839e-07,
+      "loss": 1.2109,
+      "step": 10555
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.04534792872438e-07,
+      "loss": 1.2295,
+      "step": 10556
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.034287237043482e-07,
+      "loss": 1.2344,
+      "step": 10557
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.023232992098041e-07,
+      "loss": 1.2002,
+      "step": 10558
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.012185194671463e-07,
+      "loss": 1.2383,
+      "step": 10559
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.001143845546723e-07,
+      "loss": 1.1895,
+      "step": 10560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.99010894550627e-07,
+      "loss": 1.21,
+      "step": 10561
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.979080495332205e-07,
+      "loss": 1.1963,
+      "step": 10562
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.968058495806087e-07,
+      "loss": 1.2441,
+      "step": 10563
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.957042947709071e-07,
+      "loss": 1.209,
+      "step": 10564
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.946033851821778e-07,
+      "loss": 1.2021,
+      "step": 10565
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.935031208924461e-07,
+      "loss": 1.1953,
+      "step": 10566
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.924035019796862e-07,
+      "loss": 1.2061,
+      "step": 10567
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.91304528521828e-07,
+      "loss": 1.1846,
+      "step": 10568
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.90206200596756e-07,
+      "loss": 1.1758,
+      "step": 10569
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.89108518282309e-07,
+      "loss": 1.2012,
+      "step": 10570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.880114816562801e-07,
+      "loss": 1.1748,
+      "step": 10571
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.869150907964163e-07,
+      "loss": 1.1719,
+      "step": 10572
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.858193457804165e-07,
+      "loss": 1.208,
+      "step": 10573
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.847242466859384e-07,
+      "loss": 1.1611,
+      "step": 10574
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.836297935905902e-07,
+      "loss": 1.1855,
+      "step": 10575
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.825359865719362e-07,
+      "loss": 1.2129,
+      "step": 10576
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.814428257074947e-07,
+      "loss": 1.1963,
+      "step": 10577
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.80350311074738e-07,
+      "loss": 1.167,
+      "step": 10578
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.79258442751092e-07,
+      "loss": 1.1904,
+      "step": 10579
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.78167220813938e-07,
+      "loss": 1.248,
+      "step": 10580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.770766453406133e-07,
+      "loss": 1.248,
+      "step": 10581
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.759867164084012e-07,
+      "loss": 1.2236,
+      "step": 10582
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.748974340945482e-07,
+      "loss": 1.2061,
+      "step": 10583
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.7380879847625e-07,
+      "loss": 1.2129,
+      "step": 10584
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.727208096306639e-07,
+      "loss": 1.207,
+      "step": 10585
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.71633467634888e-07,
+      "loss": 1.251,
+      "step": 10586
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.705467725659866e-07,
+      "loss": 1.2002,
+      "step": 10587
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.694607245009723e-07,
+      "loss": 1.2334,
+      "step": 10588
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.683753235168157e-07,
+      "loss": 1.2197,
+      "step": 10589
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.672905696904322e-07,
+      "loss": 1.1895,
+      "step": 10590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.662064630987055e-07,
+      "loss": 1.2217,
+      "step": 10591
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.651230038184643e-07,
+      "loss": 1.2041,
+      "step": 10592
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.640401919264941e-07,
+      "loss": 1.1543,
+      "step": 10593
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.629580274995308e-07,
+      "loss": 1.1333,
+      "step": 10594
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.61876510614268e-07,
+      "loss": 1.1836,
+      "step": 10595
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.60795641347354e-07,
+      "loss": 1.2139,
+      "step": 10596
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.597154197753899e-07,
+      "loss": 1.1934,
+      "step": 10597
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.586358459749311e-07,
+      "loss": 1.1465,
+      "step": 10598
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.575569200224854e-07,
+      "loss": 1.2217,
+      "step": 10599
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.564786419945181e-07,
+      "loss": 1.1973,
+      "step": 10600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.554010119674472e-07,
+      "loss": 1.1924,
+      "step": 10601
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.543240300176414e-07,
+      "loss": 1.1816,
+      "step": 10602
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.532476962214276e-07,
+      "loss": 1.2051,
+      "step": 10603
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.521720106550868e-07,
+      "loss": 1.2061,
+      "step": 10604
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.510969733948515e-07,
+      "loss": 1.1094,
+      "step": 10605
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.500225845169097e-07,
+      "loss": 1.2051,
+      "step": 10606
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.489488440974025e-07,
+      "loss": 1.1665,
+      "step": 10607
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.478757522124281e-07,
+      "loss": 1.1689,
+      "step": 10608
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.468033089380367e-07,
+      "loss": 1.2236,
+      "step": 10609
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.457315143502287e-07,
+      "loss": 1.1836,
+      "step": 10610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.446603685249644e-07,
+      "loss": 1.1846,
+      "step": 10611
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.435898715381552e-07,
+      "loss": 1.2178,
+      "step": 10612
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425200234656683e-07,
+      "loss": 1.1797,
+      "step": 10613
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.41450824383323e-07,
+      "loss": 1.1816,
+      "step": 10614
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.403822743668933e-07,
+      "loss": 1.1885,
+      "step": 10615
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.393143734921072e-07,
+      "loss": 1.2305,
+      "step": 10616
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.382471218346488e-07,
+      "loss": 1.2012,
+      "step": 10617
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.37180519470151e-07,
+      "loss": 1.2041,
+      "step": 10618
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.361145664742032e-07,
+      "loss": 1.1963,
+      "step": 10619
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.350492629223539e-07,
+      "loss": 1.2422,
+      "step": 10620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.339846088900993e-07,
+      "loss": 1.2383,
+      "step": 10621
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329206044528904e-07,
+      "loss": 1.2295,
+      "step": 10622
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.318572496861321e-07,
+      "loss": 1.2002,
+      "step": 10623
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.307945446651866e-07,
+      "loss": 1.1855,
+      "step": 10624
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.297324894653691e-07,
+      "loss": 1.2241,
+      "step": 10625
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.286710841619406e-07,
+      "loss": 1.1494,
+      "step": 10626
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.276103288301296e-07,
+      "loss": 1.2441,
+      "step": 10627
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.265502235451106e-07,
+      "loss": 1.1797,
+      "step": 10628
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.254907683820123e-07,
+      "loss": 1.1748,
+      "step": 10629
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.24431963415917e-07,
+      "loss": 1.2236,
+      "step": 10630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.233738087218635e-07,
+      "loss": 1.2275,
+      "step": 10631
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.223163043748428e-07,
+      "loss": 1.208,
+      "step": 10632
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.212594504498006e-07,
+      "loss": 1.2246,
+      "step": 10633
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.202032470216359e-07,
+      "loss": 1.1924,
+      "step": 10634
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.19147694165201e-07,
+      "loss": 1.1807,
+      "step": 10635
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.18092791955305e-07,
+      "loss": 1.208,
+      "step": 10636
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.170385404667081e-07,
+      "loss": 1.2285,
+      "step": 10637
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.159849397741227e-07,
+      "loss": 1.1504,
+      "step": 10638
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.149319899522179e-07,
+      "loss": 1.1982,
+      "step": 10639
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.138796910756186e-07,
+      "loss": 1.1973,
+      "step": 10640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.128280432188995e-07,
+      "loss": 1.1504,
+      "step": 10641
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.117770464565921e-07,
+      "loss": 1.209,
+      "step": 10642
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.10726700863178e-07,
+      "loss": 1.1582,
+      "step": 10643
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.096770065130977e-07,
+      "loss": 1.2139,
+      "step": 10644
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.086279634807437e-07,
+      "loss": 1.2041,
+      "step": 10645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.075795718404589e-07,
+      "loss": 1.1904,
+      "step": 10646
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.065318316665427e-07,
+      "loss": 1.1758,
+      "step": 10647
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.054847430332502e-07,
+      "loss": 1.2178,
+      "step": 10648
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.044383060147875e-07,
+      "loss": 1.2031,
+      "step": 10649
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.033925206853165e-07,
+      "loss": 1.1924,
+      "step": 10650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.023473871189502e-07,
+      "loss": 1.2109,
+      "step": 10651
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.013029053897592e-07,
+      "loss": 1.2305,
+      "step": 10652
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.002590755717665e-07,
+      "loss": 1.2041,
+      "step": 10653
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.992158977389453e-07,
+      "loss": 1.1523,
+      "step": 10654
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.98173371965224e-07,
+      "loss": 1.1953,
+      "step": 10655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.971314983244927e-07,
+      "loss": 1.1924,
+      "step": 10656
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.960902768905865e-07,
+      "loss": 1.2568,
+      "step": 10657
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.950497077372932e-07,
+      "loss": 1.167,
+      "step": 10658
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.940097909383615e-07,
+      "loss": 1.2461,
+      "step": 10659
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.929705265674892e-07,
+      "loss": 1.1904,
+      "step": 10660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.919319146983306e-07,
+      "loss": 1.1846,
+      "step": 10661
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.90893955404487e-07,
+      "loss": 1.2021,
+      "step": 10662
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.898566487595239e-07,
+      "loss": 1.2686,
+      "step": 10663
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.888199948369524e-07,
+      "loss": 1.1787,
+      "step": 10664
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.877839937102427e-07,
+      "loss": 1.1709,
+      "step": 10665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.867486454528128e-07,
+      "loss": 1.1816,
+      "step": 10666
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.857139501380395e-07,
+      "loss": 1.2246,
+      "step": 10667
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.846799078392519e-07,
+      "loss": 1.1787,
+      "step": 10668
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.836465186297315e-07,
+      "loss": 1.2373,
+      "step": 10669
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.826137825827151e-07,
+      "loss": 1.2148,
+      "step": 10670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.815816997713921e-07,
+      "loss": 1.1768,
+      "step": 10671
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.805502702689071e-07,
+      "loss": 1.2031,
+      "step": 10672
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.795194941483597e-07,
+      "loss": 1.2314,
+      "step": 10673
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.784893714827957e-07,
+      "loss": 1.1953,
+      "step": 10674
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.774599023452222e-07,
+      "loss": 1.2295,
+      "step": 10675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.764310868085989e-07,
+      "loss": 1.1875,
+      "step": 10676
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.754029249458361e-07,
+      "loss": 1.1865,
+      "step": 10677
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.743754168298001e-07,
+      "loss": 1.1699,
+      "step": 10678
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.733485625333115e-07,
+      "loss": 1.1641,
+      "step": 10679
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.72322362129142e-07,
+      "loss": 1.2139,
+      "step": 10680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.712968156900203e-07,
+      "loss": 1.2061,
+      "step": 10681
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.702719232886246e-07,
+      "loss": 1.1777,
+      "step": 10682
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.692476849975894e-07,
+      "loss": 1.2031,
+      "step": 10683
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.68224100889502e-07,
+      "loss": 1.1885,
+      "step": 10684
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.672011710369054e-07,
+      "loss": 1.2461,
+      "step": 10685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.66178895512294e-07,
+      "loss": 1.2139,
+      "step": 10686
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.651572743881153e-07,
+      "loss": 1.1768,
+      "step": 10687
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.641363077367725e-07,
+      "loss": 1.2119,
+      "step": 10688
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.631159956306233e-07,
+      "loss": 1.1875,
+      "step": 10689
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.620963381419744e-07,
+      "loss": 1.2725,
+      "step": 10690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.610773353430867e-07,
+      "loss": 1.2168,
+      "step": 10691
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.600589873061815e-07,
+      "loss": 1.1963,
+      "step": 10692
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.590412941034297e-07,
+      "loss": 1.2246,
+      "step": 10693
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.580242558069518e-07,
+      "loss": 1.2227,
+      "step": 10694
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.570078724888252e-07,
+      "loss": 1.209,
+      "step": 10695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.559921442210827e-07,
+      "loss": 1.2354,
+      "step": 10696
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.549770710757098e-07,
+      "loss": 1.2314,
+      "step": 10697
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.539626531246402e-07,
+      "loss": 1.1934,
+      "step": 10698
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.529488904397708e-07,
+      "loss": 1.248,
+      "step": 10699
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.519357830929441e-07,
+      "loss": 1.2324,
+      "step": 10700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.509233311559605e-07,
+      "loss": 1.2275,
+      "step": 10701
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.499115347005715e-07,
+      "loss": 1.1777,
+      "step": 10702
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.489003937984817e-07,
+      "loss": 1.1621,
+      "step": 10703
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.478899085213521e-07,
+      "loss": 1.1846,
+      "step": 10704
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.468800789407959e-07,
+      "loss": 1.2617,
+      "step": 10705
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.458709051283796e-07,
+      "loss": 1.1934,
+      "step": 10706
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.448623871556238e-07,
+      "loss": 1.2266,
+      "step": 10707
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.438545250939999e-07,
+      "loss": 1.1924,
+      "step": 10708
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.428473190149377e-07,
+      "loss": 1.2285,
+      "step": 10709
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.418407689898177e-07,
+      "loss": 1.167,
+      "step": 10710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.408348750899719e-07,
+      "loss": 1.1826,
+      "step": 10711
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.398296373866887e-07,
+      "loss": 1.1699,
+      "step": 10712
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.3882505595121e-07,
+      "loss": 1.1875,
+      "step": 10713
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.378211308547301e-07,
+      "loss": 1.2188,
+      "step": 10714
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.368178621683974e-07,
+      "loss": 1.1826,
+      "step": 10715
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.35815249963312e-07,
+      "loss": 1.2588,
+      "step": 10716
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.348132943105313e-07,
+      "loss": 1.1807,
+      "step": 10717
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.33811995281064e-07,
+      "loss": 1.1436,
+      "step": 10718
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.328113529458692e-07,
+      "loss": 1.1602,
+      "step": 10719
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.318113673758643e-07,
+      "loss": 1.1895,
+      "step": 10720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.30812038641916e-07,
+      "loss": 1.2441,
+      "step": 10721
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.298133668148511e-07,
+      "loss": 1.1865,
+      "step": 10722
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.288153519654429e-07,
+      "loss": 1.2246,
+      "step": 10723
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.278179941644193e-07,
+      "loss": 1.208,
+      "step": 10724
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.268212934824648e-07,
+      "loss": 1.1846,
+      "step": 10725
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.258252499902174e-07,
+      "loss": 1.207,
+      "step": 10726
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.248298637582607e-07,
+      "loss": 1.1934,
+      "step": 10727
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.238351348571426e-07,
+      "loss": 1.21,
+      "step": 10728
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.22841063357359e-07,
+      "loss": 1.2363,
+      "step": 10729
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.218476493293591e-07,
+      "loss": 1.1836,
+      "step": 10730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.208548928435455e-07,
+      "loss": 1.1973,
+      "step": 10731
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.198627939702752e-07,
+      "loss": 1.1777,
+      "step": 10732
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.188713527798585e-07,
+      "loss": 1.2393,
+      "step": 10733
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.178805693425583e-07,
+      "loss": 1.1748,
+      "step": 10734
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.168904437285907e-07,
+      "loss": 1.2012,
+      "step": 10735
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.15900976008127e-07,
+      "loss": 1.2119,
+      "step": 10736
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.149121662512915e-07,
+      "loss": 1.2451,
+      "step": 10737
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.139240145281612e-07,
+      "loss": 1.1865,
+      "step": 10738
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.129365209087635e-07,
+      "loss": 1.2002,
+      "step": 10739
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.119496854630836e-07,
+      "loss": 1.2021,
+      "step": 10740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.109635082610589e-07,
+      "loss": 1.2129,
+      "step": 10741
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.099779893725788e-07,
+      "loss": 1.2188,
+      "step": 10742
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.089931288674878e-07,
+      "loss": 1.1973,
+      "step": 10743
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.080089268155832e-07,
+      "loss": 1.1934,
+      "step": 10744
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.070253832866147e-07,
+      "loss": 1.2363,
+      "step": 10745
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.060424983502878e-07,
+      "loss": 1.2148,
+      "step": 10746
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.050602720762567e-07,
+      "loss": 1.2158,
+      "step": 10747
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.040787045341324e-07,
+      "loss": 1.2178,
+      "step": 10748
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.030977957934792e-07,
+      "loss": 1.208,
+      "step": 10749
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.021175459238139e-07,
+      "loss": 1.1782,
+      "step": 10750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.011379549946063e-07,
+      "loss": 1.1855,
+      "step": 10751
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001590230752808e-07,
+      "loss": 1.1699,
+      "step": 10752
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.991807502352143e-07,
+      "loss": 1.165,
+      "step": 10753
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.982031365437381e-07,
+      "loss": 1.1973,
+      "step": 10754
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.972261820701321e-07,
+      "loss": 1.248,
+      "step": 10755
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.962498868836343e-07,
+      "loss": 1.1914,
+      "step": 10756
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.952742510534349e-07,
+      "loss": 1.1748,
+      "step": 10757
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.942992746486798e-07,
+      "loss": 1.1816,
+      "step": 10758
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.933249577384626e-07,
+      "loss": 1.1982,
+      "step": 10759
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.923513003918336e-07,
+      "loss": 1.1748,
+      "step": 10760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.913783026777954e-07,
+      "loss": 1.1895,
+      "step": 10761
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.904059646653071e-07,
+      "loss": 1.2422,
+      "step": 10762
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.894342864232728e-07,
+      "loss": 1.2383,
+      "step": 10763
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.884632680205606e-07,
+      "loss": 1.1953,
+      "step": 10764
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.874929095259842e-07,
+      "loss": 1.2324,
+      "step": 10765
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.865232110083154e-07,
+      "loss": 1.2783,
+      "step": 10766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.855541725362725e-07,
+      "loss": 1.1924,
+      "step": 10767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.845857941785327e-07,
+      "loss": 1.1641,
+      "step": 10768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.836180760037259e-07,
+      "loss": 1.2002,
+      "step": 10769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.826510180804335e-07,
+      "loss": 1.2021,
+      "step": 10770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.816846204771921e-07,
+      "loss": 1.1475,
+      "step": 10771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.807188832624878e-07,
+      "loss": 1.1797,
+      "step": 10772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.797538065047649e-07,
+      "loss": 1.2012,
+      "step": 10773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.787893902724185e-07,
+      "loss": 1.1963,
+      "step": 10774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.778256346337931e-07,
+      "loss": 1.2051,
+      "step": 10775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.768625396571937e-07,
+      "loss": 1.1582,
+      "step": 10776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.759001054108727e-07,
+      "loss": 1.1904,
+      "step": 10777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.749383319630387e-07,
+      "loss": 1.2207,
+      "step": 10778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.739772193818516e-07,
+      "loss": 1.1211,
+      "step": 10779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.730167677354271e-07,
+      "loss": 1.1768,
+      "step": 10780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.720569770918306e-07,
+      "loss": 1.1768,
+      "step": 10781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.710978475190854e-07,
+      "loss": 1.2207,
+      "step": 10782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.701393790851618e-07,
+      "loss": 1.2178,
+      "step": 10783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.691815718579864e-07,
+      "loss": 1.2305,
+      "step": 10784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.682244259054404e-07,
+      "loss": 1.1875,
+      "step": 10785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.672679412953565e-07,
+      "loss": 1.2607,
+      "step": 10786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.663121180955201e-07,
+      "loss": 1.1973,
+      "step": 10787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.653569563736717e-07,
+      "loss": 1.2197,
+      "step": 10788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.644024561975016e-07,
+      "loss": 1.1592,
+      "step": 10789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.634486176346577e-07,
+      "loss": 1.2266,
+      "step": 10790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.624954407527351e-07,
+      "loss": 1.2227,
+      "step": 10791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.615429256192884e-07,
+      "loss": 1.2129,
+      "step": 10792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.605910723018183e-07,
+      "loss": 1.1934,
+      "step": 10793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.596398808677895e-07,
+      "loss": 1.1865,
+      "step": 10794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.586893513846071e-07,
+      "loss": 1.2012,
+      "step": 10795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.57739483919636e-07,
+      "loss": 1.1675,
+      "step": 10796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.567902785401947e-07,
+      "loss": 1.2334,
+      "step": 10797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.558417353135549e-07,
+      "loss": 1.2314,
+      "step": 10798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.548938543069338e-07,
+      "loss": 1.208,
+      "step": 10799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.539466355875124e-07,
+      "loss": 1.209,
+      "step": 10800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.530000792224201e-07,
+      "loss": 1.2041,
+      "step": 10801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.520541852787399e-07,
+      "loss": 1.1357,
+      "step": 10802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.511089538235038e-07,
+      "loss": 1.2168,
+      "step": 10803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.501643849237016e-07,
+      "loss": 1.2119,
+      "step": 10804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.492204786462763e-07,
+      "loss": 1.2461,
+      "step": 10805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.482772350581212e-07,
+      "loss": 1.2295,
+      "step": 10806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.47334654226085e-07,
+      "loss": 1.1943,
+      "step": 10807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.463927362169675e-07,
+      "loss": 1.1807,
+      "step": 10808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.45451481097522e-07,
+      "loss": 1.2295,
+      "step": 10809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.445108889344576e-07,
+      "loss": 1.2344,
+      "step": 10810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.435709597944306e-07,
+      "loss": 1.124,
+      "step": 10811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.426316937440557e-07,
+      "loss": 1.2021,
+      "step": 10812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.416930908498986e-07,
+      "loss": 1.2197,
+      "step": 10813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.407551511784771e-07,
+      "loss": 1.2373,
+      "step": 10814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.398178747962635e-07,
+      "loss": 1.1895,
+      "step": 10815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.388812617696827e-07,
+      "loss": 1.1738,
+      "step": 10816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.379453121651125e-07,
+      "loss": 1.2031,
+      "step": 10817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.370100260488843e-07,
+      "loss": 1.251,
+      "step": 10818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.360754034872807e-07,
+      "loss": 1.1836,
+      "step": 10819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.351414445465376e-07,
+      "loss": 1.2197,
+      "step": 10820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.342081492928454e-07,
+      "loss": 1.1709,
+      "step": 10821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.332755177923467e-07,
+      "loss": 1.1992,
+      "step": 10822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.323435501111385e-07,
+      "loss": 1.2217,
+      "step": 10823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.314122463152672e-07,
+      "loss": 1.2021,
+      "step": 10824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.304816064707342e-07,
+      "loss": 1.1836,
+      "step": 10825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.295516306434979e-07,
+      "loss": 1.2168,
+      "step": 10826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.2862231889946e-07,
+      "loss": 1.2246,
+      "step": 10827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.276936713044812e-07,
+      "loss": 1.2041,
+      "step": 10828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.2676568792438e-07,
+      "loss": 1.1992,
+      "step": 10829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.258383688249192e-07,
+      "loss": 1.1816,
+      "step": 10830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.249117140718175e-07,
+      "loss": 1.2109,
+      "step": 10831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.239857237307467e-07,
+      "loss": 1.2188,
+      "step": 10832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230603978673322e-07,
+      "loss": 1.2598,
+      "step": 10833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.221357365471548e-07,
+      "loss": 1.2422,
+      "step": 10834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.212117398357375e-07,
+      "loss": 1.208,
+      "step": 10835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.202884077985715e-07,
+      "loss": 1.1826,
+      "step": 10836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.193657405010899e-07,
+      "loss": 1.1787,
+      "step": 10837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.18443738008685e-07,
+      "loss": 1.1846,
+      "step": 10838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.175224003866942e-07,
+      "loss": 1.1875,
+      "step": 10839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.166017277004155e-07,
+      "loss": 1.2236,
+      "step": 10840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.156817200150978e-07,
+      "loss": 1.2305,
+      "step": 10841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.147623773959399e-07,
+      "loss": 1.2334,
+      "step": 10842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.138436999080966e-07,
+      "loss": 1.1758,
+      "step": 10843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.129256876166745e-07,
+      "loss": 1.1758,
+      "step": 10844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.120083405867327e-07,
+      "loss": 1.2119,
+      "step": 10845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.110916588832838e-07,
+      "loss": 1.2314,
+      "step": 10846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.101756425712945e-07,
+      "loss": 1.2275,
+      "step": 10847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.092602917156809e-07,
+      "loss": 1.2275,
+      "step": 10848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.083456063813142e-07,
+      "loss": 1.2031,
+      "step": 10849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.074315866330172e-07,
+      "loss": 1.2139,
+      "step": 10850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.065182325355678e-07,
+      "loss": 1.1992,
+      "step": 10851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.056055441536957e-07,
+      "loss": 1.1836,
+      "step": 10852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.046935215520822e-07,
+      "loss": 1.1699,
+      "step": 10853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.037821647953624e-07,
+      "loss": 1.207,
+      "step": 10854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.028714739481256e-07,
+      "loss": 1.2051,
+      "step": 10855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.019614490749093e-07,
+      "loss": 1.165,
+      "step": 10856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.010520902402094e-07,
+      "loss": 1.1909,
+      "step": 10857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.001433975084692e-07,
+      "loss": 1.2178,
+      "step": 10858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.992353709440935e-07,
+      "loss": 1.208,
+      "step": 10859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.983280106114286e-07,
+      "loss": 1.1797,
+      "step": 10860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.974213165747811e-07,
+      "loss": 1.1885,
+      "step": 10861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.965152888984083e-07,
+      "loss": 1.2158,
+      "step": 10862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.956099276465221e-07,
+      "loss": 1.249,
+      "step": 10863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.947052328832792e-07,
+      "loss": 1.2178,
+      "step": 10864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.938012046728026e-07,
+      "loss": 1.2129,
+      "step": 10865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.928978430791576e-07,
+      "loss": 1.2129,
+      "step": 10866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.919951481663666e-07,
+      "loss": 1.2432,
+      "step": 10867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.910931199984005e-07,
+      "loss": 1.2061,
+      "step": 10868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.901917586391892e-07,
+      "loss": 1.1924,
+      "step": 10869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.892910641526095e-07,
+      "loss": 1.1885,
+      "step": 10870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.883910366024958e-07,
+      "loss": 1.2139,
+      "step": 10871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.874916760526328e-07,
+      "loss": 1.2275,
+      "step": 10872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.865929825667571e-07,
+      "loss": 1.2031,
+      "step": 10873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.85694956208559e-07,
+      "loss": 1.2295,
+      "step": 10874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.847975970416841e-07,
+      "loss": 1.2236,
+      "step": 10875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.83900905129724e-07,
+      "loss": 1.2061,
+      "step": 10876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.830048805362309e-07,
+      "loss": 1.2061,
+      "step": 10877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.821095233247031e-07,
+      "loss": 1.2275,
+      "step": 10878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.812148335585966e-07,
+      "loss": 1.1846,
+      "step": 10879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.80320811301317e-07,
+      "loss": 1.2002,
+      "step": 10880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.794274566162239e-07,
+      "loss": 1.1992,
+      "step": 10881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.785347695666299e-07,
+      "loss": 1.2285,
+      "step": 10882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.776427502158e-07,
+      "loss": 1.1748,
+      "step": 10883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.76751398626949e-07,
+      "loss": 1.1934,
+      "step": 10884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.7586071486325e-07,
+      "loss": 1.2119,
+      "step": 10885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.749706989878234e-07,
+      "loss": 1.207,
+      "step": 10886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.740813510637455e-07,
+      "loss": 1.1768,
+      "step": 10887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.731926711540448e-07,
+      "loss": 1.1729,
+      "step": 10888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.723046593217019e-07,
+      "loss": 1.1953,
+      "step": 10889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.714173156296488e-07,
+      "loss": 1.1953,
+      "step": 10890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.705306401407751e-07,
+      "loss": 1.1963,
+      "step": 10891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.69644632917915e-07,
+      "loss": 1.1865,
+      "step": 10892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.687592940238618e-07,
+      "loss": 1.2041,
+      "step": 10893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.678746235213572e-07,
+      "loss": 1.2061,
+      "step": 10894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.669906214731036e-07,
+      "loss": 1.2012,
+      "step": 10895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.66107287941744e-07,
+      "loss": 1.2227,
+      "step": 10896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.652246229898838e-07,
+      "loss": 1.1855,
+      "step": 10897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.643426266800755e-07,
+      "loss": 1.1914,
+      "step": 10898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.634612990748289e-07,
+      "loss": 1.1465,
+      "step": 10899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.625806402365985e-07,
+      "loss": 1.1797,
+      "step": 10900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.61700650227801e-07,
+      "loss": 1.1826,
+      "step": 10901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.608213291108e-07,
+      "loss": 1.2158,
+      "step": 10902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.599426769479144e-07,
+      "loss": 1.209,
+      "step": 10903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.59064693801411e-07,
+      "loss": 1.2285,
+      "step": 10904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.581873797335146e-07,
+      "loss": 1.1904,
+      "step": 10905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.573107348063989e-07,
+      "loss": 1.207,
+      "step": 10906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.564347590821939e-07,
+      "loss": 1.1719,
+      "step": 10907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.555594526229779e-07,
+      "loss": 1.2285,
+      "step": 10908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.546848154907857e-07,
+      "loss": 1.2354,
+      "step": 10909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.538108477476012e-07,
+      "loss": 1.1973,
+      "step": 10910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.529375494553646e-07,
+      "loss": 1.1738,
+      "step": 10911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.520649206759632e-07,
+      "loss": 1.2207,
+      "step": 10912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.511929614712419e-07,
+      "loss": 1.167,
+      "step": 10913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.503216719029969e-07,
+      "loss": 1.2246,
+      "step": 10914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.494510520329754e-07,
+      "loss": 1.2109,
+      "step": 10915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.485811019228793e-07,
+      "loss": 1.252,
+      "step": 10916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.477118216343614e-07,
+      "loss": 1.1924,
+      "step": 10917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.468432112290267e-07,
+      "loss": 1.25,
+      "step": 10918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.459752707684362e-07,
+      "loss": 1.1943,
+      "step": 10919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.451080003140985e-07,
+      "loss": 1.1631,
+      "step": 10920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.442413999274765e-07,
+      "loss": 1.2168,
+      "step": 10921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.433754696699878e-07,
+      "loss": 1.2402,
+      "step": 10922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.425102096029988e-07,
+      "loss": 1.2012,
+      "step": 10923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.416456197878328e-07,
+      "loss": 1.1982,
+      "step": 10924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.407817002857618e-07,
+      "loss": 1.1875,
+      "step": 10925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.399184511580124e-07,
+      "loss": 1.2197,
+      "step": 10926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.390558724657646e-07,
+      "loss": 1.1729,
+      "step": 10927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.381939642701461e-07,
+      "loss": 1.2373,
+      "step": 10928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.373327266322414e-07,
+      "loss": 1.1646,
+      "step": 10929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.364721596130851e-07,
+      "loss": 1.1982,
+      "step": 10930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.356122632736694e-07,
+      "loss": 1.25,
+      "step": 10931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.347530376749321e-07,
+      "loss": 1.2158,
+      "step": 10932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.33894482877767e-07,
+      "loss": 1.166,
+      "step": 10933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.330365989430197e-07,
+      "loss": 1.2275,
+      "step": 10934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.321793859314905e-07,
+      "loss": 1.2363,
+      "step": 10935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.313228439039253e-07,
+      "loss": 1.2031,
+      "step": 10936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.304669729210321e-07,
+      "loss": 1.2705,
+      "step": 10937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.296117730434636e-07,
+      "loss": 1.1748,
+      "step": 10938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.287572443318301e-07,
+      "loss": 1.2012,
+      "step": 10939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.279033868466899e-07,
+      "loss": 1.2031,
+      "step": 10940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.270502006485556e-07,
+      "loss": 1.1543,
+      "step": 10941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.261976857978946e-07,
+      "loss": 1.2188,
+      "step": 10942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253458423551217e-07,
+      "loss": 1.1953,
+      "step": 10943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.244946703806098e-07,
+      "loss": 1.208,
+      "step": 10944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.236441699346794e-07,
+      "loss": 1.2617,
+      "step": 10945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.227943410776071e-07,
+      "loss": 1.2041,
+      "step": 10946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.219451838696199e-07,
+      "loss": 1.2197,
+      "step": 10947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.210966983708965e-07,
+      "loss": 1.208,
+      "step": 10948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.202488846415688e-07,
+      "loss": 1.2607,
+      "step": 10949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.194017427417231e-07,
+      "loss": 1.2158,
+      "step": 10950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.185552727313947e-07,
+      "loss": 1.2109,
+      "step": 10951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.177094746705725e-07,
+      "loss": 1.1885,
+      "step": 10952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.168643486192004e-07,
+      "loss": 1.2119,
+      "step": 10953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.160198946371709e-07,
+      "loss": 1.1963,
+      "step": 10954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.151761127843313e-07,
+      "loss": 1.2041,
+      "step": 10955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.143330031204796e-07,
+      "loss": 1.1973,
+      "step": 10956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.134905657053657e-07,
+      "loss": 1.2158,
+      "step": 10957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.126488005986951e-07,
+      "loss": 1.1738,
+      "step": 10958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.118077078601235e-07,
+      "loss": 1.1562,
+      "step": 10959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.109672875492588e-07,
+      "loss": 1.2012,
+      "step": 10960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.101275397256611e-07,
+      "loss": 1.1865,
+      "step": 10961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.092884644488428e-07,
+      "loss": 1.2227,
+      "step": 10962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.084500617782718e-07,
+      "loss": 1.2373,
+      "step": 10963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.076123317733627e-07,
+      "loss": 1.1982,
+      "step": 10964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.067752744934862e-07,
+      "loss": 1.2158,
+      "step": 10965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.059388899979623e-07,
+      "loss": 1.1807,
+      "step": 10966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051031783460725e-07,
+      "loss": 1.2139,
+      "step": 10967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.042681395970362e-07,
+      "loss": 1.2354,
+      "step": 10968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.034337738100358e-07,
+      "loss": 1.1787,
+      "step": 10969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.026000810442033e-07,
+      "loss": 1.1934,
+      "step": 10970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.017670613586223e-07,
+      "loss": 1.1924,
+      "step": 10971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.009347148123255e-07,
+      "loss": 1.1855,
+      "step": 10972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.001030414643049e-07,
+      "loss": 1.2168,
+      "step": 10973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.992720413735008e-07,
+      "loss": 1.1465,
+      "step": 10974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.984417145988052e-07,
+      "loss": 1.2373,
+      "step": 10975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.976120611990654e-07,
+      "loss": 1.1846,
+      "step": 10976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.967830812330754e-07,
+      "loss": 1.1738,
+      "step": 10977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.959547747595861e-07,
+      "loss": 1.1729,
+      "step": 10978
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.951271418373005e-07,
+      "loss": 1.2002,
+      "step": 10979
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.943001825248728e-07,
+      "loss": 1.2285,
+      "step": 10980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.934738968809083e-07,
+      "loss": 1.208,
+      "step": 10981
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.92648284963968e-07,
+      "loss": 1.1846,
+      "step": 10982
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.918233468325606e-07,
+      "loss": 1.2324,
+      "step": 10983
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.909990825451527e-07,
+      "loss": 1.1982,
+      "step": 10984
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.901754921601565e-07,
+      "loss": 1.2197,
+      "step": 10985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.893525757359407e-07,
+      "loss": 1.2334,
+      "step": 10986
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.885303333308267e-07,
+      "loss": 1.2031,
+      "step": 10987
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.877087650030843e-07,
+      "loss": 1.248,
+      "step": 10988
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.868878708109403e-07,
+      "loss": 1.2295,
+      "step": 10989
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.860676508125705e-07,
+      "loss": 1.2168,
+      "step": 10990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.85248105066104e-07,
+      "loss": 1.2266,
+      "step": 10991
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.84429233629623e-07,
+      "loss": 1.2256,
+      "step": 10992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.83611036561159e-07,
+      "loss": 1.1963,
+      "step": 10993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.827935139186979e-07,
+      "loss": 1.1855,
+      "step": 10994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.819766657601766e-07,
+      "loss": 1.2031,
+      "step": 10995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.811604921434898e-07,
+      "loss": 1.1924,
+      "step": 10996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.803449931264748e-07,
+      "loss": 1.2139,
+      "step": 10997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.795301687669273e-07,
+      "loss": 1.1904,
+      "step": 10998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.787160191225937e-07,
+      "loss": 1.2471,
+      "step": 10999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.779025442511754e-07,
+      "loss": 1.2148,
+      "step": 11000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.770897442103172e-07,
+      "loss": 1.1992,
+      "step": 11001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.762776190576279e-07,
+      "loss": 1.2275,
+      "step": 11002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.754661688506612e-07,
+      "loss": 1.1465,
+      "step": 11003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.746553936469256e-07,
+      "loss": 1.21,
+      "step": 11004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.738452935038784e-07,
+      "loss": 1.1631,
+      "step": 11005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.730358684789316e-07,
+      "loss": 1.2217,
+      "step": 11006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7222711862945026e-07,
+      "loss": 1.1719,
+      "step": 11007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7141904401274975e-07,
+      "loss": 1.2295,
+      "step": 11008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7061164468609863e-07,
+      "loss": 1.2363,
+      "step": 11009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.698049207067168e-07,
+      "loss": 1.2021,
+      "step": 11010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.689988721317773e-07,
+      "loss": 1.165,
+      "step": 11011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.681934990184056e-07,
+      "loss": 1.2061,
+      "step": 11012
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6738880142367714e-07,
+      "loss": 1.1768,
+      "step": 11013
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6658477940462077e-07,
+      "loss": 1.1636,
+      "step": 11014
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6578143301821644e-07,
+      "loss": 1.2393,
+      "step": 11015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.649787623213997e-07,
+      "loss": 1.2373,
+      "step": 11016
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.641767673710551e-07,
+      "loss": 1.1709,
+      "step": 11017
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6337544822401826e-07,
+      "loss": 1.1377,
+      "step": 11018
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.625748049370815e-07,
+      "loss": 1.1904,
+      "step": 11019
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6177483756698507e-07,
+      "loss": 1.2266,
+      "step": 11020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6097554617042126e-07,
+      "loss": 1.1982,
+      "step": 11021
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6017693080403715e-07,
+      "loss": 1.2256,
+      "step": 11022
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5937899152442956e-07,
+      "loss": 1.2568,
+      "step": 11023
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5858172838815e-07,
+      "loss": 1.2031,
+      "step": 11024
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5778514145169983e-07,
+      "loss": 1.1738,
+      "step": 11025
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.569892307715318e-07,
+      "loss": 1.1729,
+      "step": 11026
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5619399640405404e-07,
+      "loss": 1.1631,
+      "step": 11027
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5539943840562485e-07,
+      "loss": 1.1768,
+      "step": 11028
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.546055568325525e-07,
+      "loss": 1.1904,
+      "step": 11029
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5381235174110085e-07,
+      "loss": 1.1992,
+      "step": 11030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.530198231874816e-07,
+      "loss": 1.2031,
+      "step": 11031
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5222797122786655e-07,
+      "loss": 1.2314,
+      "step": 11032
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5143679591837074e-07,
+      "loss": 1.1826,
+      "step": 11033
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.506462973150638e-07,
+      "loss": 1.2129,
+      "step": 11034
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4985647547397095e-07,
+      "loss": 1.1963,
+      "step": 11035
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.490673304510651e-07,
+      "loss": 1.1748,
+      "step": 11036
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.482788623022727e-07,
+      "loss": 1.1934,
+      "step": 11037
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4749107108347236e-07,
+      "loss": 1.1719,
+      "step": 11038
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4670395685049715e-07,
+      "loss": 1.166,
+      "step": 11039
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4591751965912923e-07,
+      "loss": 1.1875,
+      "step": 11040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4513175956510056e-07,
+      "loss": 1.1709,
+      "step": 11041
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.443466766241e-07,
+      "loss": 1.1572,
+      "step": 11042
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.435622708917653e-07,
+      "loss": 1.2812,
+      "step": 11043
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.427785424236886e-07,
+      "loss": 1.1826,
+      "step": 11044
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.419954912754121e-07,
+      "loss": 1.2178,
+      "step": 11045
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4121311750243146e-07,
+      "loss": 1.2119,
+      "step": 11046
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4043142116019124e-07,
+      "loss": 1.2314,
+      "step": 11047
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3965040230409375e-07,
+      "loss": 1.2363,
+      "step": 11048
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.388700609894858e-07,
+      "loss": 1.2021,
+      "step": 11049
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3809039727167214e-07,
+      "loss": 1.1865,
+      "step": 11050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.373114112059074e-07,
+      "loss": 1.1562,
+      "step": 11051
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3653310284739736e-07,
+      "loss": 1.1836,
+      "step": 11052
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.357554722513024e-07,
+      "loss": 1.2334,
+      "step": 11053
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3497851947273186e-07,
+      "loss": 1.2227,
+      "step": 11054
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.342022445667482e-07,
+      "loss": 1.2158,
+      "step": 11055
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.334266475883686e-07,
+      "loss": 1.1992,
+      "step": 11056
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.326517285925558e-07,
+      "loss": 1.2197,
+      "step": 11057
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3187748763423023e-07,
+      "loss": 1.248,
+      "step": 11058
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3110392476826247e-07,
+      "loss": 1.1582,
+      "step": 11059
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.303310400494742e-07,
+      "loss": 1.2168,
+      "step": 11060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.295588335326395e-07,
+      "loss": 1.166,
+      "step": 11061
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2878730527248556e-07,
+      "loss": 1.1768,
+      "step": 11062
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.280164553236898e-07,
+      "loss": 1.1885,
+      "step": 11063
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.272462837408842e-07,
+      "loss": 1.1846,
+      "step": 11064
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.264767905786471e-07,
+      "loss": 1.1465,
+      "step": 11065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.25707975891515e-07,
+      "loss": 1.2041,
+      "step": 11066
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2493983973397214e-07,
+      "loss": 1.1982,
+      "step": 11067
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.241723821604604e-07,
+      "loss": 1.1553,
+      "step": 11068
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2340560322536417e-07,
+      "loss": 1.1768,
+      "step": 11069
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.226395029830277e-07,
+      "loss": 1.1895,
+      "step": 11070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.218740814877442e-07,
+      "loss": 1.1777,
+      "step": 11071
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2110933879376037e-07,
+      "loss": 1.1699,
+      "step": 11072
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2034527495526945e-07,
+      "loss": 1.1738,
+      "step": 11073
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.195818900264248e-07,
+      "loss": 1.2158,
+      "step": 11074
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.188191840613254e-07,
+      "loss": 1.1826,
+      "step": 11075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.180571571140268e-07,
+      "loss": 1.1611,
+      "step": 11076
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.172958092385304e-07,
+      "loss": 1.1895,
+      "step": 11077
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1653514048879296e-07,
+      "loss": 1.1836,
+      "step": 11078
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.157751509187258e-07,
+      "loss": 1.1934,
+      "step": 11079
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1501584058218694e-07,
+      "loss": 1.1689,
+      "step": 11080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1425720953298886e-07,
+      "loss": 1.2881,
+      "step": 11081
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1349925782489754e-07,
+      "loss": 1.1895,
+      "step": 11082
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.127419855116277e-07,
+      "loss": 1.2168,
+      "step": 11083
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1198539264684755e-07,
+      "loss": 1.1855,
+      "step": 11084
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1122947928417646e-07,
+      "loss": 1.207,
+      "step": 11085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1047424547718595e-07,
+      "loss": 1.2148,
+      "step": 11086
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0971969127939883e-07,
+      "loss": 1.1807,
+      "step": 11087
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0896581674429114e-07,
+      "loss": 1.21,
+      "step": 11088
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.082126219252902e-07,
+      "loss": 1.251,
+      "step": 11089
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.074601068757744e-07,
+      "loss": 1.1904,
+      "step": 11090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0670827164907446e-07,
+      "loss": 1.2129,
+      "step": 11091
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0595711629847433e-07,
+      "loss": 1.1572,
+      "step": 11092
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0520664087720596e-07,
+      "loss": 1.1387,
+      "step": 11093
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0445684543845565e-07,
+      "loss": 1.1963,
+      "step": 11094
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0370773003536313e-07,
+      "loss": 1.1904,
+      "step": 11095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0295929472101703e-07,
+      "loss": 1.208,
+      "step": 11096
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0221153954845936e-07,
+      "loss": 1.1914,
+      "step": 11097
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0146446457068336e-07,
+      "loss": 1.207,
+      "step": 11098
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0071806984063323e-07,
+      "loss": 1.1504,
+      "step": 11099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9997235541120895e-07,
+      "loss": 1.2129,
+      "step": 11100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.99227321335256e-07,
+      "loss": 1.1924,
+      "step": 11101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.984829676655755e-07,
+      "loss": 1.1865,
+      "step": 11102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9773929445491963e-07,
+      "loss": 1.2334,
+      "step": 11103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.969963017559952e-07,
+      "loss": 1.1631,
+      "step": 11104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9625398962145456e-07,
+      "loss": 1.1719,
+      "step": 11105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9551235810390776e-07,
+      "loss": 1.1982,
+      "step": 11106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.947714072559128e-07,
+      "loss": 1.2266,
+      "step": 11107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.94031137129981e-07,
+      "loss": 1.2188,
+      "step": 11108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.932915477785759e-07,
+      "loss": 1.1904,
+      "step": 11109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9255263925411126e-07,
+      "loss": 1.1689,
+      "step": 11110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9181441160895393e-07,
+      "loss": 1.1787,
+      "step": 11111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.910768648954233e-07,
+      "loss": 1.1768,
+      "step": 11112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.903399991657886e-07,
+      "loss": 1.2188,
+      "step": 11113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8960381447226916e-07,
+      "loss": 1.1992,
+      "step": 11114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.88868310867041e-07,
+      "loss": 1.2197,
+      "step": 11115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8813348840222807e-07,
+      "loss": 1.2178,
+      "step": 11116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.873993471299076e-07,
+      "loss": 1.2344,
+      "step": 11117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8666588710210894e-07,
+      "loss": 1.1978,
+      "step": 11118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8593310837081066e-07,
+      "loss": 1.1982,
+      "step": 11119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.852010109879456e-07,
+      "loss": 1.1611,
+      "step": 11120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8446959500539895e-07,
+      "loss": 1.21,
+      "step": 11121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8373886047500254e-07,
+      "loss": 1.1797,
+      "step": 11122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.830088074485461e-07,
+      "loss": 1.208,
+      "step": 11123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8227943597776816e-07,
+      "loss": 1.1611,
+      "step": 11124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.815507461143597e-07,
+      "loss": 1.1797,
+      "step": 11125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.808227379099605e-07,
+      "loss": 1.2061,
+      "step": 11126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8009541141616814e-07,
+      "loss": 1.2129,
+      "step": 11127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.793687666845247e-07,
+      "loss": 1.2031,
+      "step": 11128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7864280376653126e-07,
+      "loss": 1.1875,
+      "step": 11129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7791752271363335e-07,
+      "loss": 1.167,
+      "step": 11130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.77192923577232e-07,
+      "loss": 1.1904,
+      "step": 11131
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7646900640867956e-07,
+      "loss": 1.2012,
+      "step": 11132
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.757457712592827e-07,
+      "loss": 1.2559,
+      "step": 11133
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7502321818029483e-07,
+      "loss": 1.2148,
+      "step": 11134
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.743013472229229e-07,
+      "loss": 1.1816,
+      "step": 11135
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.73580158438327e-07,
+      "loss": 1.2354,
+      "step": 11136
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7285965187761843e-07,
+      "loss": 1.1689,
+      "step": 11137
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.721398275918564e-07,
+      "loss": 1.1602,
+      "step": 11138
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.714206856320557e-07,
+      "loss": 1.1963,
+      "step": 11139
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.707022260491844e-07,
+      "loss": 1.1729,
+      "step": 11140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.699844488941595e-07,
+      "loss": 1.2168,
+      "step": 11141
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.692673542178471e-07,
+      "loss": 1.1689,
+      "step": 11142
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.685509420710687e-07,
+      "loss": 1.168,
+      "step": 11143
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.67835212504597e-07,
+      "loss": 1.1729,
+      "step": 11144
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.671201655691559e-07,
+      "loss": 1.1738,
+      "step": 11145
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.664058013154204e-07,
+      "loss": 1.1914,
+      "step": 11146
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.656921197940166e-07,
+      "loss": 1.208,
+      "step": 11147
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6497912105552424e-07,
+      "loss": 1.2422,
+      "step": 11148
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6426680515047384e-07,
+      "loss": 1.1826,
+      "step": 11149
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6355517212934624e-07,
+      "loss": 1.1494,
+      "step": 11150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.628442220425743e-07,
+      "loss": 1.1445,
+      "step": 11151
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.621339549405445e-07,
+      "loss": 1.1631,
+      "step": 11152
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.614243708735921e-07,
+      "loss": 1.209,
+      "step": 11153
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.607154698920057e-07,
+      "loss": 1.2324,
+      "step": 11154
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.600072520460263e-07,
+      "loss": 1.1689,
+      "step": 11155
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.592997173858437e-07,
+      "loss": 1.2002,
+      "step": 11156
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5859286596160225e-07,
+      "loss": 1.1826,
+      "step": 11157
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5788669782339413e-07,
+      "loss": 1.1768,
+      "step": 11158
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.571812130212682e-07,
+      "loss": 1.2441,
+      "step": 11159
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5647641160521995e-07,
+      "loss": 1.1709,
+      "step": 11160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.557722936251995e-07,
+      "loss": 1.1904,
+      "step": 11161
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.55068859131108e-07,
+      "loss": 1.1777,
+      "step": 11162
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.543661081727967e-07,
+      "loss": 1.1914,
+      "step": 11163
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.536640408000713e-07,
+      "loss": 1.2158,
+      "step": 11164
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.529626570626876e-07,
+      "loss": 1.1963,
+      "step": 11165
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5226195701034917e-07,
+      "loss": 1.1885,
+      "step": 11166
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.515619406927173e-07,
+      "loss": 1.2432,
+      "step": 11167
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.508626081594002e-07,
+      "loss": 1.1758,
+      "step": 11168
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5016395945996374e-07,
+      "loss": 1.1729,
+      "step": 11169
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.494659946439161e-07,
+      "loss": 1.1514,
+      "step": 11170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4876871376072543e-07,
+      "loss": 1.2168,
+      "step": 11171
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.480721168598056e-07,
+      "loss": 1.2002,
+      "step": 11172
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4737620399052707e-07,
+      "loss": 1.1992,
+      "step": 11173
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.46680975202206e-07,
+      "loss": 1.2227,
+      "step": 11174
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.459864305441152e-07,
+      "loss": 1.2012,
+      "step": 11175
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4529257006547744e-07,
+      "loss": 1.1494,
+      "step": 11176
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.445993938154657e-07,
+      "loss": 1.21,
+      "step": 11177
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.439069018432051e-07,
+      "loss": 1.2354,
+      "step": 11178
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.43215094197773e-07,
+      "loss": 1.1895,
+      "step": 11179
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.425239709281969e-07,
+      "loss": 1.1875,
+      "step": 11180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.418335320834587e-07,
+      "loss": 1.1934,
+      "step": 11181
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.411437777124871e-07,
+      "loss": 1.2061,
+      "step": 11182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4045470786416735e-07,
+      "loss": 1.2119,
+      "step": 11183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3976632258733156e-07,
+      "loss": 1.1914,
+      "step": 11184
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3907862193076847e-07,
+      "loss": 1.2256,
+      "step": 11185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3839160594321243e-07,
+      "loss": 1.1973,
+      "step": 11186
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3770527467335443e-07,
+      "loss": 1.2441,
+      "step": 11187
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3701962816983234e-07,
+      "loss": 1.2002,
+      "step": 11188
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.363346664812406e-07,
+      "loss": 1.1758,
+      "step": 11189
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.356503896561203e-07,
+      "loss": 1.1377,
+      "step": 11190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3496679774296713e-07,
+      "loss": 1.208,
+      "step": 11191
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.342838907902268e-07,
+      "loss": 1.1729,
+      "step": 11192
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.336016688462984e-07,
+      "loss": 1.2236,
+      "step": 11193
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3292013195952876e-07,
+      "loss": 1.208,
+      "step": 11194
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.322392801782193e-07,
+      "loss": 1.2061,
+      "step": 11195
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3155911355062245e-07,
+      "loss": 1.2051,
+      "step": 11196
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.308796321249397e-07,
+      "loss": 1.2119,
+      "step": 11197
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3020083594932805e-07,
+      "loss": 1.2314,
+      "step": 11198
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2952272507189355e-07,
+      "loss": 1.2041,
+      "step": 11199
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2884529954069213e-07,
+      "loss": 1.1797,
+      "step": 11200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2816855940373536e-07,
+      "loss": 1.2188,
+      "step": 11201
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.274925047089816e-07,
+      "loss": 1.1729,
+      "step": 11202
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.268171355043437e-07,
+      "loss": 1.2119,
+      "step": 11203
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2614245183768433e-07,
+      "loss": 1.2139,
+      "step": 11204
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.254684537568209e-07,
+      "loss": 1.2207,
+      "step": 11205
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2479514130951627e-07,
+      "loss": 1.1797,
+      "step": 11206
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.241225145434901e-07,
+      "loss": 1.1924,
+      "step": 11207
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2345057350640973e-07,
+      "loss": 1.208,
+      "step": 11208
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2277931824589827e-07,
+      "loss": 1.1758,
+      "step": 11209
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2210874880952424e-07,
+      "loss": 1.1768,
+      "step": 11210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2143886524481307e-07,
+      "loss": 1.1533,
+      "step": 11211
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.207696675992389e-07,
+      "loss": 1.2275,
+      "step": 11212
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.201011559202294e-07,
+      "loss": 1.2354,
+      "step": 11213
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.194333302551589e-07,
+      "loss": 1.2393,
+      "step": 11214
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1876619065135726e-07,
+      "loss": 1.2061,
+      "step": 11215
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1809973715610454e-07,
+      "loss": 1.1758,
+      "step": 11216
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.174339698166329e-07,
+      "loss": 1.1816,
+      "step": 11217
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.167688886801257e-07,
+      "loss": 1.1875,
+      "step": 11218
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.161044937937152e-07,
+      "loss": 1.2178,
+      "step": 11219
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.154407852044894e-07,
+      "loss": 1.168,
+      "step": 11220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1477776295948504e-07,
+      "loss": 1.2012,
+      "step": 11221
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.141154271056879e-07,
+      "loss": 1.1738,
+      "step": 11222
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1345377769004035e-07,
+      "loss": 1.1855,
+      "step": 11223
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1279281475943166e-07,
+      "loss": 1.2588,
+      "step": 11224
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1213253836070546e-07,
+      "loss": 1.1826,
+      "step": 11225
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1147294854065533e-07,
+      "loss": 1.2393,
+      "step": 11226
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.108140453460251e-07,
+      "loss": 1.165,
+      "step": 11227
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1015582882351404e-07,
+      "loss": 1.1826,
+      "step": 11228
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0949829901976814e-07,
+      "loss": 1.1875,
+      "step": 11229
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.088414559813857e-07,
+      "loss": 1.2334,
+      "step": 11230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0818529975491727e-07,
+      "loss": 1.1982,
+      "step": 11231
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.075298303868657e-07,
+      "loss": 1.2021,
+      "step": 11232
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.068750479236848e-07,
+      "loss": 1.2197,
+      "step": 11233
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.062209524117765e-07,
+      "loss": 1.2207,
+      "step": 11234
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.055675438974981e-07,
+      "loss": 1.2119,
+      "step": 11235
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.04914822427157e-07,
+      "loss": 1.2432,
+      "step": 11236
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.042627880470117e-07,
+      "loss": 1.2188,
+      "step": 11237
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0361144080326976e-07,
+      "loss": 1.2031,
+      "step": 11238
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.029607807420931e-07,
+      "loss": 1.1855,
+      "step": 11239
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0231080790959377e-07,
+      "loss": 1.1934,
+      "step": 11240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.016615223518371e-07,
+      "loss": 1.1436,
+      "step": 11241
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.010129241148374e-07,
+      "loss": 1.1953,
+      "step": 11242
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.003650132445601e-07,
+      "loss": 1.2178,
+      "step": 11243
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.997177897869208e-07,
+      "loss": 1.1553,
+      "step": 11244
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9907125378779155e-07,
+      "loss": 1.1875,
+      "step": 11245
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9842540529299025e-07,
+      "loss": 1.1865,
+      "step": 11246
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9778024434828803e-07,
+      "loss": 1.1836,
+      "step": 11247
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.971357709994083e-07,
+      "loss": 1.1924,
+      "step": 11248
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.964919852920245e-07,
+      "loss": 1.1924,
+      "step": 11249
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.958488872717635e-07,
+      "loss": 1.2275,
+      "step": 11250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.952064769841978e-07,
+      "loss": 1.208,
+      "step": 11251
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.945647544748576e-07,
+      "loss": 1.2354,
+      "step": 11252
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9392371978922196e-07,
+      "loss": 1.1797,
+      "step": 11253
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9328337297271913e-07,
+      "loss": 1.168,
+      "step": 11254
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.926437140707317e-07,
+      "loss": 1.2412,
+      "step": 11255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.920047431285933e-07,
+      "loss": 1.2061,
+      "step": 11256
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9136646019158553e-07,
+      "loss": 1.2031,
+      "step": 11257
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.907288653049456e-07,
+      "loss": 1.1611,
+      "step": 11258
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.900919585138584e-07,
+      "loss": 1.2412,
+      "step": 11259
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.894557398634612e-07,
+      "loss": 1.2266,
+      "step": 11260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.888202093988435e-07,
+      "loss": 1.1836,
+      "step": 11261
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8818536716504476e-07,
+      "loss": 1.2686,
+      "step": 11262
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.875512132070579e-07,
+      "loss": 1.1699,
+      "step": 11263
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.869177475698237e-07,
+      "loss": 1.1836,
+      "step": 11264
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.862849702982362e-07,
+      "loss": 1.2363,
+      "step": 11265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8565288143714067e-07,
+      "loss": 1.2285,
+      "step": 11266
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.850214810313334e-07,
+      "loss": 1.2314,
+      "step": 11267
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8439076912555983e-07,
+      "loss": 1.21,
+      "step": 11268
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.837607457645197e-07,
+      "loss": 1.1719,
+      "step": 11269
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8313141099286513e-07,
+      "loss": 1.2139,
+      "step": 11270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.825027648551937e-07,
+      "loss": 1.2119,
+      "step": 11271
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8187480739605867e-07,
+      "loss": 1.1826,
+      "step": 11272
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8124753865996336e-07,
+      "loss": 1.1465,
+      "step": 11273
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.806209586913633e-07,
+      "loss": 1.1699,
+      "step": 11274
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.799950675346619e-07,
+      "loss": 1.2012,
+      "step": 11275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.793698652342169e-07,
+      "loss": 1.2012,
+      "step": 11276
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7874535183433746e-07,
+      "loss": 1.1777,
+      "step": 11277
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7812152737928255e-07,
+      "loss": 1.1719,
+      "step": 11278
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.774983919132612e-07,
+      "loss": 1.1826,
+      "step": 11279
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7687594548043707e-07,
+      "loss": 1.1973,
+      "step": 11280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.762541881249203e-07,
+      "loss": 1.1719,
+      "step": 11281
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.756331198907769e-07,
+      "loss": 1.2002,
+      "step": 11282
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7501274082202156e-07,
+      "loss": 1.1895,
+      "step": 11283
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7439305096262024e-07,
+      "loss": 1.2354,
+      "step": 11284
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.737740503564901e-07,
+      "loss": 1.2207,
+      "step": 11285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.731557390475015e-07,
+      "loss": 1.207,
+      "step": 11286
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.725381170794705e-07,
+      "loss": 1.2471,
+      "step": 11287
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.719211844961711e-07,
+      "loss": 1.2539,
+      "step": 11288
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.713049413413238e-07,
+      "loss": 1.2158,
+      "step": 11289
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.706893876586014e-07,
+      "loss": 1.1855,
+      "step": 11290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7007452349163024e-07,
+      "loss": 1.2246,
+      "step": 11291
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.694603488839831e-07,
+      "loss": 1.1641,
+      "step": 11292
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6884686387918857e-07,
+      "loss": 1.2002,
+      "step": 11293
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.682340685207241e-07,
+      "loss": 1.2109,
+      "step": 11294
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.67621962852016e-07,
+      "loss": 1.1816,
+      "step": 11295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6701054691644743e-07,
+      "loss": 1.1787,
+      "step": 11296
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6639982075734707e-07,
+      "loss": 1.1855,
+      "step": 11297
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6578978441799796e-07,
+      "loss": 1.2217,
+      "step": 11298
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.651804379416345e-07,
+      "loss": 1.2295,
+      "step": 11299
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.645717813714388e-07,
+      "loss": 1.1611,
+      "step": 11300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6396381475054743e-07,
+      "loss": 1.209,
+      "step": 11301
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.633565381220482e-07,
+      "loss": 1.1895,
+      "step": 11302
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6274995152897666e-07,
+      "loss": 1.1943,
+      "step": 11303
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.621440550143217e-07,
+      "loss": 1.2334,
+      "step": 11304
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6153884862102354e-07,
+      "loss": 1.209,
+      "step": 11305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.609343323919755e-07,
+      "loss": 1.2354,
+      "step": 11306
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.603305063700168e-07,
+      "loss": 1.2119,
+      "step": 11307
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5972737059794083e-07,
+      "loss": 1.1807,
+      "step": 11308
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.591249251184924e-07,
+      "loss": 1.1826,
+      "step": 11309
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5852316997436733e-07,
+      "loss": 1.2129,
+      "step": 11310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.579221052082115e-07,
+      "loss": 1.2061,
+      "step": 11311
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.573217308626197e-07,
+      "loss": 1.2246,
+      "step": 11312
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.567220469801446e-07,
+      "loss": 1.207,
+      "step": 11313
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.561230536032855e-07,
+      "loss": 1.1885,
+      "step": 11314
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.555247507744907e-07,
+      "loss": 1.2061,
+      "step": 11315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.549271385361618e-07,
+      "loss": 1.2158,
+      "step": 11316
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5433021693065386e-07,
+      "loss": 1.1982,
+      "step": 11317
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5373398600026855e-07,
+      "loss": 1.1973,
+      "step": 11318
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.531384457872621e-07,
+      "loss": 1.1924,
+      "step": 11319
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5254359633383964e-07,
+      "loss": 1.1514,
+      "step": 11320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.519494376821574e-07,
+      "loss": 1.1455,
+      "step": 11321
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.513559698743262e-07,
+      "loss": 1.2383,
+      "step": 11322
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5076319295240235e-07,
+      "loss": 1.2441,
+      "step": 11323
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.501711069583968e-07,
+      "loss": 1.208,
+      "step": 11324
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.495797119342702e-07,
+      "loss": 1.1611,
+      "step": 11325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.489890079219348e-07,
+      "loss": 1.208,
+      "step": 11326
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.483989949632548e-07,
+      "loss": 1.1992,
+      "step": 11327
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4780967310004346e-07,
+      "loss": 1.2012,
+      "step": 11328
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.472210423740662e-07,
+      "loss": 1.251,
+      "step": 11329
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.466331028270408e-07,
+      "loss": 1.1807,
+      "step": 11330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4604585450063057e-07,
+      "loss": 1.1826,
+      "step": 11331
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4545929743645667e-07,
+      "loss": 1.2656,
+      "step": 11332
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4487343167608926e-07,
+      "loss": 1.2344,
+      "step": 11333
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4428825726104497e-07,
+      "loss": 1.2021,
+      "step": 11334
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4370377423279856e-07,
+      "loss": 1.2139,
+      "step": 11335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4311998263277126e-07,
+      "loss": 1.1631,
+      "step": 11336
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4253688250233555e-07,
+      "loss": 1.2168,
+      "step": 11337
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4195447388281833e-07,
+      "loss": 1.2178,
+      "step": 11338
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.413727568154911e-07,
+      "loss": 1.2178,
+      "step": 11339
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4079173134158087e-07,
+      "loss": 1.2236,
+      "step": 11340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4021139750226686e-07,
+      "loss": 1.2383,
+      "step": 11341
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3963175533867733e-07,
+      "loss": 1.2002,
+      "step": 11342
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.390528048918894e-07,
+      "loss": 1.209,
+      "step": 11343
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3847454620293365e-07,
+      "loss": 1.168,
+      "step": 11344
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3789697931279276e-07,
+      "loss": 1.207,
+      "step": 11345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.373201042623996e-07,
+      "loss": 1.2217,
+      "step": 11346
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3674392109263366e-07,
+      "loss": 1.1982,
+      "step": 11347
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.361684298443312e-07,
+      "loss": 1.21,
+      "step": 11348
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3559363055827734e-07,
+      "loss": 1.1475,
+      "step": 11349
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.350195232752095e-07,
+      "loss": 1.2334,
+      "step": 11350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.344461080358107e-07,
+      "loss": 1.1777,
+      "step": 11351
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3387338488072287e-07,
+      "loss": 1.208,
+      "step": 11352
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3330135385053243e-07,
+      "loss": 1.2207,
+      "step": 11353
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3273001498577919e-07,
+      "loss": 1.2412,
+      "step": 11354
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.321593683269552e-07,
+      "loss": 1.2109,
+      "step": 11355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.315894139145014e-07,
+      "loss": 1.1963,
+      "step": 11356
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.310201517888111e-07,
+      "loss": 1.207,
+      "step": 11357
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3045158199022754e-07,
+      "loss": 1.168,
+      "step": 11358
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.298837045590452e-07,
+      "loss": 1.1934,
+      "step": 11359
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2931651953550849e-07,
+      "loss": 1.168,
+      "step": 11360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.287500269598153e-07,
+      "loss": 1.2012,
+      "step": 11361
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.281842268721124e-07,
+      "loss": 1.2002,
+      "step": 11362
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.276191193124977e-07,
+      "loss": 1.2314,
+      "step": 11363
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2705470432102027e-07,
+      "loss": 1.1592,
+      "step": 11364
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2649098193768148e-07,
+      "loss": 1.1787,
+      "step": 11365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2592795220243158e-07,
+      "loss": 1.1826,
+      "step": 11366
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2536561515517086e-07,
+      "loss": 1.1699,
+      "step": 11367
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2480397083575412e-07,
+      "loss": 1.2168,
+      "step": 11368
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2424301928398396e-07,
+      "loss": 1.2188,
+      "step": 11369
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2368276053961524e-07,
+      "loss": 1.1982,
+      "step": 11370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.231231946423551e-07,
+      "loss": 1.1904,
+      "step": 11371
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2256432163185736e-07,
+      "loss": 1.25,
+      "step": 11372
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2200614154773038e-07,
+      "loss": 1.2402,
+      "step": 11373
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2144865442953245e-07,
+      "loss": 1.1646,
+      "step": 11374
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2089186031677424e-07,
+      "loss": 1.2139,
+      "step": 11375
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2033575924891193e-07,
+      "loss": 1.1523,
+      "step": 11376
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1978035126535846e-07,
+      "loss": 1.1748,
+      "step": 11377
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1922563640547677e-07,
+      "loss": 1.2373,
+      "step": 11378
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.186716147085799e-07,
+      "loss": 1.2666,
+      "step": 11379
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.181182862139275e-07,
+      "loss": 1.1758,
+      "step": 11380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1756565096073712e-07,
+      "loss": 1.21,
+      "step": 11381
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1701370898817298e-07,
+      "loss": 1.1709,
+      "step": 11382
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1646246033535156e-07,
+      "loss": 1.1875,
+      "step": 11383
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.159119050413394e-07,
+      "loss": 1.2012,
+      "step": 11384
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.153620431451542e-07,
+      "loss": 1.2236,
+      "step": 11385
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1481287468576473e-07,
+      "loss": 1.1807,
+      "step": 11386
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1426439970209212e-07,
+      "loss": 1.1621,
+      "step": 11387
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1371661823300526e-07,
+      "loss": 1.1904,
+      "step": 11388
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1316953031732423e-07,
+      "loss": 1.2061,
+      "step": 11389
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1262313599382357e-07,
+      "loss": 1.2197,
+      "step": 11390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.120774353012245e-07,
+      "loss": 1.1611,
+      "step": 11391
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1153242827820054e-07,
+      "loss": 1.1807,
+      "step": 11392
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1098811496337856e-07,
+      "loss": 1.21,
+      "step": 11393
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.104444953953322e-07,
+      "loss": 1.2295,
+      "step": 11394
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0990156961258833e-07,
+      "loss": 1.2129,
+      "step": 11395
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0935933765362403e-07,
+      "loss": 1.2012,
+      "step": 11396
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0881779955686632e-07,
+      "loss": 1.2178,
+      "step": 11397
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.082769553606956e-07,
+      "loss": 1.252,
+      "step": 11398
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0773680510344007e-07,
+      "loss": 1.2217,
+      "step": 11399
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.071973488233814e-07,
+      "loss": 1.209,
+      "step": 11400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0665858655874894e-07,
+      "loss": 1.208,
+      "step": 11401
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.061205183477266e-07,
+      "loss": 1.2188,
+      "step": 11402
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0558314422844837e-07,
+      "loss": 1.1729,
+      "step": 11403
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0504646423899377e-07,
+      "loss": 1.2109,
+      "step": 11404
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0451047841740125e-07,
+      "loss": 1.2334,
+      "step": 11405
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.039751868016526e-07,
+      "loss": 1.2803,
+      "step": 11406
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0344058942968758e-07,
+      "loss": 1.2158,
+      "step": 11407
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0290668633939025e-07,
+      "loss": 1.2227,
+      "step": 11408
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0237347756860037e-07,
+      "loss": 1.1738,
+      "step": 11409
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0184096315510437e-07,
+      "loss": 1.1816,
+      "step": 11410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.013091431366443e-07,
+      "loss": 1.2188,
+      "step": 11411
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0077801755090775e-07,
+      "loss": 1.2031,
+      "step": 11412
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0024758643553467e-07,
+      "loss": 1.168,
+      "step": 11413
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9971784982811936e-07,
+      "loss": 1.1709,
+      "step": 11414
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9918880776620518e-07,
+      "loss": 1.1699,
+      "step": 11415
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9866046028728103e-07,
+      "loss": 1.2324,
+      "step": 11416
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.981328074287936e-07,
+      "loss": 1.2139,
+      "step": 11417
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.976058492281374e-07,
+      "loss": 1.2021,
+      "step": 11418
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9707958572265817e-07,
+      "loss": 1.2197,
+      "step": 11419
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.965540169496505e-07,
+      "loss": 1.2158,
+      "step": 11420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9602914294636344e-07,
+      "loss": 1.1846,
+      "step": 11421
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9550496374999394e-07,
+      "loss": 1.1924,
+      "step": 11422
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.949814793976912e-07,
+      "loss": 1.168,
+      "step": 11423
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9445868992655327e-07,
+      "loss": 1.2217,
+      "step": 11424
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9393659537363162e-07,
+      "loss": 1.1729,
+      "step": 11425
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.934151957759256e-07,
+      "loss": 1.2363,
+      "step": 11426
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9289449117038784e-07,
+      "loss": 1.2383,
+      "step": 11427
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.923744815939199e-07,
+      "loss": 1.25,
+      "step": 11428
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9185516708337458e-07,
+      "loss": 1.2197,
+      "step": 11429
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.91336547675558e-07,
+      "loss": 1.1797,
+      "step": 11430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9081862340722291e-07,
+      "loss": 1.2002,
+      "step": 11431
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9030139431507334e-07,
+      "loss": 1.2031,
+      "step": 11432
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8978486043576772e-07,
+      "loss": 1.1855,
+      "step": 11433
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8926902180591121e-07,
+      "loss": 1.1953,
+      "step": 11434
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8875387846206127e-07,
+      "loss": 1.1904,
+      "step": 11435
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8823943044072757e-07,
+      "loss": 1.2627,
+      "step": 11436
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8772567777836758e-07,
+      "loss": 1.249,
+      "step": 11437
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.872126205113911e-07,
+      "loss": 1.2002,
+      "step": 11438
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8670025867616015e-07,
+      "loss": 1.2168,
+      "step": 11439
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8618859230898346e-07,
+      "loss": 1.251,
+      "step": 11440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8567762144612313e-07,
+      "loss": 1.1768,
+      "step": 11441
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.851673461237913e-07,
+      "loss": 1.1855,
+      "step": 11442
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8465776637815459e-07,
+      "loss": 1.1611,
+      "step": 11443
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8414888224532301e-07,
+      "loss": 1.1602,
+      "step": 11444
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8364069376136218e-07,
+      "loss": 1.2236,
+      "step": 11445
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8313320096228881e-07,
+      "loss": 1.1846,
+      "step": 11446
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.826264038840675e-07,
+      "loss": 1.1514,
+      "step": 11447
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8212030256261504e-07,
+      "loss": 1.1846,
+      "step": 11448
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.816148970337983e-07,
+      "loss": 1.1719,
+      "step": 11449
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8111018733343643e-07,
+      "loss": 1.1934,
+      "step": 11450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.806061734972997e-07,
+      "loss": 1.1416,
+      "step": 11451
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8010285556110396e-07,
+      "loss": 1.1982,
+      "step": 11452
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.796002335605207e-07,
+      "loss": 1.2412,
+      "step": 11453
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.790983075311714e-07,
+      "loss": 1.1611,
+      "step": 11454
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7859707750862765e-07,
+      "loss": 1.1758,
+      "step": 11455
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7809654352840988e-07,
+      "loss": 1.165,
+      "step": 11456
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7759670562599307e-07,
+      "loss": 1.1787,
+      "step": 11457
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.770975638368e-07,
+      "loss": 1.1602,
+      "step": 11458
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7659911819620458e-07,
+      "loss": 1.2012,
+      "step": 11459
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7610136873953187e-07,
+      "loss": 1.1826,
+      "step": 11460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7560431550205593e-07,
+      "loss": 1.1963,
+      "step": 11461
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7510795851900519e-07,
+      "loss": 1.1631,
+      "step": 11462
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7461229782555489e-07,
+      "loss": 1.1758,
+      "step": 11463
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7411733345683246e-07,
+      "loss": 1.2197,
+      "step": 11464
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7362306544791652e-07,
+      "loss": 1.1333,
+      "step": 11465
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.731294938338368e-07,
+      "loss": 1.2256,
+      "step": 11466
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7263661864957205e-07,
+      "loss": 1.2012,
+      "step": 11467
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7214443993005092e-07,
+      "loss": 1.1885,
+      "step": 11468
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7165295771015445e-07,
+      "loss": 1.1807,
+      "step": 11469
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7116217202471586e-07,
+      "loss": 1.2852,
+      "step": 11470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.706720829085151e-07,
+      "loss": 1.1553,
+      "step": 11471
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7018269039628555e-07,
+      "loss": 1.1523,
+      "step": 11472
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.696939945227105e-07,
+      "loss": 1.207,
+      "step": 11473
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6920599532242343e-07,
+      "loss": 1.1826,
+      "step": 11474
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.687186928300111e-07,
+      "loss": 1.2041,
+      "step": 11475
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6823208708000472e-07,
+      "loss": 1.2256,
+      "step": 11476
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.677461781068923e-07,
+      "loss": 1.2178,
+      "step": 11477
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6726096594510855e-07,
+      "loss": 1.1816,
+      "step": 11478
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6677645062904367e-07,
+      "loss": 1.1855,
+      "step": 11479
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6629263219303138e-07,
+      "loss": 1.1875,
+      "step": 11480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6580951067136308e-07,
+      "loss": 1.2021,
+      "step": 11481
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6532708609827586e-07,
+      "loss": 1.1387,
+      "step": 11482
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6484535850796012e-07,
+      "loss": 1.1621,
+      "step": 11483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6436432793455416e-07,
+      "loss": 1.1797,
+      "step": 11484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6388399441214952e-07,
+      "loss": 1.1719,
+      "step": 11485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6340435797478793e-07,
+      "loss": 1.1533,
+      "step": 11486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6292541865646106e-07,
+      "loss": 1.1748,
+      "step": 11487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6244717649111174e-07,
+      "loss": 1.2256,
+      "step": 11488
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6196963151263066e-07,
+      "loss": 1.2227,
+      "step": 11489
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.614927837548641e-07,
+      "loss": 1.2119,
+      "step": 11490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6101663325160498e-07,
+      "loss": 1.2227,
+      "step": 11491
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6054118003659857e-07,
+      "loss": 1.1885,
+      "step": 11492
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.60066424143539e-07,
+      "loss": 1.2119,
+      "step": 11493
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5959236560607382e-07,
+      "loss": 1.2012,
+      "step": 11494
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.591190044577995e-07,
+      "loss": 1.1528,
+      "step": 11495
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5864634073226138e-07,
+      "loss": 1.166,
+      "step": 11496
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5817437446295825e-07,
+      "loss": 1.1904,
+      "step": 11497
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.577031056833389e-07,
+      "loss": 1.1748,
+      "step": 11498
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5723253442680107e-07,
+      "loss": 1.2197,
+      "step": 11499
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5676266072669365e-07,
+      "loss": 1.1455,
+      "step": 11500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5629348461631887e-07,
+      "loss": 1.1865,
+      "step": 11501
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5582500612892458e-07,
+      "loss": 1.2344,
+      "step": 11502
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5535722529771424e-07,
+      "loss": 1.1436,
+      "step": 11503
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5489014215583799e-07,
+      "loss": 1.1875,
+      "step": 11504
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.544237567363982e-07,
+      "loss": 1.2148,
+      "step": 11505
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5395806907244736e-07,
+      "loss": 1.207,
+      "step": 11506
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5349307919698797e-07,
+      "loss": 1.1963,
+      "step": 11507
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5302878714297697e-07,
+      "loss": 1.1895,
+      "step": 11508
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5256519294331584e-07,
+      "loss": 1.1963,
+      "step": 11509
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.521022966308594e-07,
+      "loss": 1.2715,
+      "step": 11510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5164009823841475e-07,
+      "loss": 1.2988,
+      "step": 11511
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5117859779873788e-07,
+      "loss": 1.207,
+      "step": 11512
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5071779534453379e-07,
+      "loss": 1.2686,
+      "step": 11513
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5025769090845854e-07,
+      "loss": 1.1816,
+      "step": 11514
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4979828452312273e-07,
+      "loss": 1.2383,
+      "step": 11515
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4933957622108474e-07,
+      "loss": 1.1504,
+      "step": 11516
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4888156603484972e-07,
+      "loss": 1.208,
+      "step": 11517
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4842425399687943e-07,
+      "loss": 1.1758,
+      "step": 11518
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4796764013958244e-07,
+      "loss": 1.2451,
+      "step": 11519
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4751172449531947e-07,
+      "loss": 1.1719,
+      "step": 11520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.470565070964003e-07,
+      "loss": 1.1787,
+      "step": 11521
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4660198797508795e-07,
+      "loss": 1.2285,
+      "step": 11522
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4614816716359337e-07,
+      "loss": 1.2324,
+      "step": 11523
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4569504469407858e-07,
+      "loss": 1.2119,
+      "step": 11524
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4524262059865567e-07,
+      "loss": 1.1855,
+      "step": 11525
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4479089490938903e-07,
+      "loss": 1.1572,
+      "step": 11526
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4433986765829188e-07,
+      "loss": 1.2539,
+      "step": 11527
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4388953887732871e-07,
+      "loss": 1.1992,
+      "step": 11528
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4343990859841396e-07,
+      "loss": 1.2129,
+      "step": 11529
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4299097685341434e-07,
+      "loss": 1.2158,
+      "step": 11530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4254274367414333e-07,
+      "loss": 1.2236,
+      "step": 11531
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4209520909236997e-07,
+      "loss": 1.1099,
+      "step": 11532
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4164837313980774e-07,
+      "loss": 1.2002,
+      "step": 11533
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4120223584812575e-07,
+      "loss": 1.2305,
+      "step": 11534
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4075679724894208e-07,
+      "loss": 1.2168,
+      "step": 11535
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4031205737382369e-07,
+      "loss": 1.21,
+      "step": 11536
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.398680162542898e-07,
+      "loss": 1.2666,
+      "step": 11537
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3942467392181081e-07,
+      "loss": 1.1748,
+      "step": 11538
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3898203040780488e-07,
+      "loss": 1.2012,
+      "step": 11539
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3854008574364253e-07,
+      "loss": 1.1982,
+      "step": 11540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.380988399606442e-07,
+      "loss": 1.1904,
+      "step": 11541
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.376582930900805e-07,
+      "loss": 1.1738,
+      "step": 11542
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3721844516317418e-07,
+      "loss": 1.2148,
+      "step": 11543
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677929621109697e-07,
+      "loss": 1.2295,
+      "step": 11544
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3634084626497069e-07,
+      "loss": 1.167,
+      "step": 11545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.359030953558682e-07,
+      "loss": 1.1924,
+      "step": 11546
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3546604351481363e-07,
+      "loss": 1.2178,
+      "step": 11547
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3502969077278217e-07,
+      "loss": 1.2109,
+      "step": 11548
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3459403716069464e-07,
+      "loss": 1.1445,
+      "step": 11549
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3415908270942634e-07,
+      "loss": 1.2578,
+      "step": 11550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3372482744980485e-07,
+      "loss": 1.1758,
+      "step": 11551
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3329127141260555e-07,
+      "loss": 1.2178,
+      "step": 11552
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3285841462855277e-07,
+      "loss": 1.1748,
+      "step": 11553
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3242625712832414e-07,
+      "loss": 1.2363,
+      "step": 11554
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.319947989425463e-07,
+      "loss": 1.2314,
+      "step": 11555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3156404010179701e-07,
+      "loss": 1.1846,
+      "step": 11556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.311339806366041e-07,
+      "loss": 1.1699,
+      "step": 11557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3070462057744537e-07,
+      "loss": 1.1934,
+      "step": 11558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3027595995474984e-07,
+      "loss": 1.1973,
+      "step": 11559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2984799879889655e-07,
+      "loss": 1.1758,
+      "step": 11560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2942073714021564e-07,
+      "loss": 1.2217,
+      "step": 11561
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2899417500898625e-07,
+      "loss": 1.1924,
+      "step": 11562
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2856831243543977e-07,
+      "loss": 1.1768,
+      "step": 11563
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2814314944975535e-07,
+      "loss": 1.1738,
+      "step": 11564
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2771868608206673e-07,
+      "loss": 1.209,
+      "step": 11565
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2729492236245422e-07,
+      "loss": 1.2822,
+      "step": 11566
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.268718583209494e-07,
+      "loss": 1.2314,
+      "step": 11567
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2644949398753603e-07,
+      "loss": 1.1924,
+      "step": 11568
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2602782939214686e-07,
+      "loss": 1.1982,
+      "step": 11569
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2560686456466353e-07,
+      "loss": 1.2285,
+      "step": 11570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2518659953492219e-07,
+      "loss": 1.2119,
+      "step": 11571
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2476703433270566e-07,
+      "loss": 1.1914,
+      "step": 11572
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2434816898774905e-07,
+      "loss": 1.2363,
+      "step": 11573
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.239300035297364e-07,
+      "loss": 1.2412,
+      "step": 11574
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2351253798830398e-07,
+      "loss": 1.2139,
+      "step": 11575
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2309577239303816e-07,
+      "loss": 1.2129,
+      "step": 11576
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2267970677347306e-07,
+      "loss": 1.2158,
+      "step": 11577
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2226434115909625e-07,
+      "loss": 1.2158,
+      "step": 11578
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2184967557934524e-07,
+      "loss": 1.1885,
+      "step": 11579
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2143571006360766e-07,
+      "loss": 1.1953,
+      "step": 11580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2102244464121893e-07,
+      "loss": 1.1777,
+      "step": 11581
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2060987934147006e-07,
+      "loss": 1.1797,
+      "step": 11582
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2019801419359767e-07,
+      "loss": 1.1514,
+      "step": 11583
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1978684922679174e-07,
+      "loss": 1.1992,
+      "step": 11584
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1937638447019007e-07,
+      "loss": 1.1963,
+      "step": 11585
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1896661995288384e-07,
+      "loss": 1.1816,
+      "step": 11586
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1855755570391092e-07,
+      "loss": 1.1816,
+      "step": 11587
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.181491917522648e-07,
+      "loss": 1.2637,
+      "step": 11588
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1774152812688455e-07,
+      "loss": 1.2031,
+      "step": 11589
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.173345648566604e-07,
+      "loss": 1.1611,
+      "step": 11590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1692830197043481e-07,
+      "loss": 1.2266,
+      "step": 11591
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1652273949699921e-07,
+      "loss": 1.1855,
+      "step": 11592
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1611787746509617e-07,
+      "loss": 1.2256,
+      "step": 11593
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1571371590341829e-07,
+      "loss": 1.1699,
+      "step": 11594
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1531025484060709e-07,
+      "loss": 1.1523,
+      "step": 11595
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1490749430525861e-07,
+      "loss": 1.1729,
+      "step": 11596
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1450543432591443e-07,
+      "loss": 1.2168,
+      "step": 11597
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1410407493106957e-07,
+      "loss": 1.1836,
+      "step": 11598
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.137034161491668e-07,
+      "loss": 1.2178,
+      "step": 11599
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1330345800860121e-07,
+      "loss": 1.2275,
+      "step": 11600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.12904200537719e-07,
+      "loss": 1.2246,
+      "step": 11601
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1250564376481531e-07,
+      "loss": 1.1709,
+      "step": 11602
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1210778771813425e-07,
+      "loss": 1.2441,
+      "step": 11603
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1171063242587432e-07,
+      "loss": 1.1807,
+      "step": 11604
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1131417791618082e-07,
+      "loss": 1.1982,
+      "step": 11605
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1091842421714905e-07,
+      "loss": 1.2217,
+      "step": 11606
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1052337135682766e-07,
+      "loss": 1.1875,
+      "step": 11607
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1012901936321319e-07,
+      "loss": 1.2344,
+      "step": 11608
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0973536826425436e-07,
+      "loss": 1.1475,
+      "step": 11609
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0934241808784774e-07,
+      "loss": 1.1855,
+      "step": 11610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0895016886184329e-07,
+      "loss": 1.1924,
+      "step": 11611
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0855862061403877e-07,
+      "loss": 1.1885,
+      "step": 11612
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.081677733721842e-07,
+      "loss": 1.1807,
+      "step": 11613
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0777762716397632e-07,
+      "loss": 1.1758,
+      "step": 11614
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0738818201706747e-07,
+      "loss": 1.2256,
+      "step": 11615
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0699943795905666e-07,
+      "loss": 1.2236,
+      "step": 11616
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0661139501749407e-07,
+      "loss": 1.2324,
+      "step": 11617
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0622405321987994e-07,
+      "loss": 1.2422,
+      "step": 11618
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0583741259366564e-07,
+      "loss": 1.1504,
+      "step": 11619
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.054514731662537e-07,
+      "loss": 1.2256,
+      "step": 11620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0506623496499335e-07,
+      "loss": 1.2139,
+      "step": 11621
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0468169801718609e-07,
+      "loss": 1.2109,
+      "step": 11622
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0429786235008676e-07,
+      "loss": 1.2051,
+      "step": 11623
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0391472799089586e-07,
+      "loss": 1.2227,
+      "step": 11624
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.035322949667661e-07,
+      "loss": 1.2744,
+      "step": 11625
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0315056330480134e-07,
+      "loss": 1.2246,
+      "step": 11626
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0276953303205439e-07,
+      "loss": 1.21,
+      "step": 11627
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0238920417552922e-07,
+      "loss": 1.2227,
+      "step": 11628
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.020095767621787e-07,
+      "loss": 1.21,
+      "step": 11629
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0163065081890799e-07,
+      "loss": 1.2109,
+      "step": 11630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0125242637257116e-07,
+      "loss": 1.1777,
+      "step": 11631
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0087490344997342e-07,
+      "loss": 1.2031,
+      "step": 11632
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0049808207786893e-07,
+      "loss": 1.2256,
+      "step": 11633
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.00121962282963e-07,
+      "loss": 1.2085,
+      "step": 11634
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.97465440919132e-08,
+      "loss": 1.1611,
+      "step": 11635
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.937182753132269e-08,
+      "loss": 1.2188,
+      "step": 11636
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.89978126277491e-08,
+      "loss": 1.2158,
+      "step": 11637
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.862449940769903e-08,
+      "loss": 1.1875,
+      "step": 11638
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.825188789762796e-08,
+      "loss": 1.1855,
+      "step": 11639
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.787997812394478e-08,
+      "loss": 1.2021,
+      "step": 11640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.750877011300508e-08,
+      "loss": 1.1572,
+      "step": 11641
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.713826389111669e-08,
+      "loss": 1.2012,
+      "step": 11642
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.67684594845375e-08,
+      "loss": 1.2158,
+      "step": 11643
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.639935691947432e-08,
+      "loss": 1.1982,
+      "step": 11644
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.603095622208847e-08,
+      "loss": 1.1924,
+      "step": 11645
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.56632574184857e-08,
+      "loss": 1.2842,
+      "step": 11646
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.529626053472629e-08,
+      "loss": 1.2031,
+      "step": 11647
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.492996559681833e-08,
+      "loss": 1.1914,
+      "step": 11648
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.456437263072216e-08,
+      "loss": 1.1738,
+      "step": 11649
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.419948166234704e-08,
+      "loss": 1.2363,
+      "step": 11650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.383529271755121e-08,
+      "loss": 1.1914,
+      "step": 11651
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.34718058221462e-08,
+      "loss": 1.1816,
+      "step": 11652
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.310902100189256e-08,
+      "loss": 1.209,
+      "step": 11653
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.27469382825008e-08,
+      "loss": 1.1875,
+      "step": 11654
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.238555768963043e-08,
+      "loss": 1.2266,
+      "step": 11655
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.202487924889426e-08,
+      "loss": 1.209,
+      "step": 11656
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.166490298585185e-08,
+      "loss": 1.1826,
+      "step": 11657
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.130562892601502e-08,
+      "loss": 1.2295,
+      "step": 11658
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.094705709484675e-08,
+      "loss": 1.1689,
+      "step": 11659
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.05891875177578e-08,
+      "loss": 1.2256,
+      "step": 11660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.023202022011124e-08,
+      "loss": 1.2012,
+      "step": 11661
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.987555522721903e-08,
+      "loss": 1.1865,
+      "step": 11662
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.951979256434318e-08,
+      "loss": 1.1963,
+      "step": 11663
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.916473225669798e-08,
+      "loss": 1.2266,
+      "step": 11664
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.881037432944439e-08,
+      "loss": 1.1641,
+      "step": 11665
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.845671880769791e-08,
+      "loss": 1.166,
+      "step": 11666
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.81037657165218e-08,
+      "loss": 1.1758,
+      "step": 11667
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.775151508092938e-08,
+      "loss": 1.2109,
+      "step": 11668
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.739996692588404e-08,
+      "loss": 1.2637,
+      "step": 11669
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.704912127630027e-08,
+      "loss": 1.1265,
+      "step": 11670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.669897815704376e-08,
+      "loss": 1.1953,
+      "step": 11671
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.634953759292685e-08,
+      "loss": 1.1992,
+      "step": 11672
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.600079960871643e-08,
+      "loss": 1.21,
+      "step": 11673
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.565276422912605e-08,
+      "loss": 1.1836,
+      "step": 11674
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.530543147882264e-08,
+      "loss": 1.2129,
+      "step": 11675
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.495880138241986e-08,
+      "loss": 1.2129,
+      "step": 11676
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.46128739644847e-08,
+      "loss": 1.1855,
+      "step": 11677
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.426764924953201e-08,
+      "loss": 1.1416,
+      "step": 11678
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.392312726202889e-08,
+      "loss": 1.1572,
+      "step": 11679
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.357930802639025e-08,
+      "loss": 1.1396,
+      "step": 11680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.323619156698437e-08,
+      "loss": 1.165,
+      "step": 11681
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.289377790812624e-08,
+      "loss": 1.1641,
+      "step": 11682
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.255206707408314e-08,
+      "loss": 1.1729,
+      "step": 11683
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.221105908907234e-08,
+      "loss": 1.207,
+      "step": 11684
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.18707539772623e-08,
+      "loss": 1.1699,
+      "step": 11685
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.153115176276705e-08,
+      "loss": 1.2119,
+      "step": 11686
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.119225246965734e-08,
+      "loss": 1.1943,
+      "step": 11687
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.085405612194953e-08,
+      "loss": 1.2114,
+      "step": 11688
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.051656274361219e-08,
+      "loss": 1.2344,
+      "step": 11689
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.017977235856289e-08,
+      "loss": 1.2114,
+      "step": 11690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.98436849906703e-08,
+      "loss": 1.2285,
+      "step": 11691
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950830066375315e-08,
+      "loss": 1.2021,
+      "step": 11692
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.91736194015802e-08,
+      "loss": 1.2021,
+      "step": 11693
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.883964122786914e-08,
+      "loss": 1.2598,
+      "step": 11694
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.850636616628992e-08,
+      "loss": 1.1455,
+      "step": 11695
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.817379424046146e-08,
+      "loss": 1.166,
+      "step": 11696
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.784192547395375e-08,
+      "loss": 1.2441,
+      "step": 11697
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.751075989028467e-08,
+      "loss": 1.1787,
+      "step": 11698
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.718029751292545e-08,
+      "loss": 1.2031,
+      "step": 11699
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.685053836529622e-08,
+      "loss": 1.1562,
+      "step": 11700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.652148247076497e-08,
+      "loss": 1.1309,
+      "step": 11701
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.619312985265414e-08,
+      "loss": 1.1855,
+      "step": 11702
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.586548053423292e-08,
+      "loss": 1.2119,
+      "step": 11703
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.553853453872163e-08,
+      "loss": 1.2666,
+      "step": 11704
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.521229188929279e-08,
+      "loss": 1.208,
+      "step": 11705
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.488675260906464e-08,
+      "loss": 1.2539,
+      "step": 11706
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.456191672110869e-08,
+      "loss": 1.2061,
+      "step": 11707
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.423778424844652e-08,
+      "loss": 1.2246,
+      "step": 11708
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.391435521404977e-08,
+      "loss": 1.209,
+      "step": 11709
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.35916296408401e-08,
+      "loss": 1.1709,
+      "step": 11710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.3269607551687e-08,
+      "loss": 1.167,
+      "step": 11711
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.294828896941441e-08,
+      "loss": 1.248,
+      "step": 11712
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.262767391679416e-08,
+      "loss": 1.1973,
+      "step": 11713
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.23077624165458e-08,
+      "loss": 1.2031,
+      "step": 11714
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.198855449134345e-08,
+      "loss": 1.2324,
+      "step": 11715
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.167005016380901e-08,
+      "loss": 1.1963,
+      "step": 11716
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.135224945651553e-08,
+      "loss": 1.1562,
+      "step": 11717
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.103515239198388e-08,
+      "loss": 1.2061,
+      "step": 11718
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.071875899268832e-08,
+      "loss": 1.207,
+      "step": 11719
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.040306928104979e-08,
+      "loss": 1.1758,
+      "step": 11720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.008808327944372e-08,
+      "loss": 1.1572,
+      "step": 11721
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.977380101019115e-08,
+      "loss": 1.207,
+      "step": 11722
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.946022249556539e-08,
+      "loss": 1.2275,
+      "step": 11723
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.914734775779086e-08,
+      "loss": 1.2432,
+      "step": 11724
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.883517681904095e-08,
+      "loss": 1.1631,
+      "step": 11725
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.852370970143795e-08,
+      "loss": 1.2334,
+      "step": 11726
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.821294642705534e-08,
+      "loss": 1.166,
+      "step": 11727
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.79028870179188e-08,
+      "loss": 1.2197,
+      "step": 11728
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.759353149600078e-08,
+      "loss": 1.1816,
+      "step": 11729
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.728487988322596e-08,
+      "loss": 1.2598,
+      "step": 11730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.697693220146795e-08,
+      "loss": 1.1689,
+      "step": 11731
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.66696884725504e-08,
+      "loss": 1.1807,
+      "step": 11732
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.636314871825034e-08,
+      "loss": 1.1641,
+      "step": 11733
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.605731296028927e-08,
+      "loss": 1.2246,
+      "step": 11734
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.575218122034211e-08,
+      "loss": 1.2031,
+      "step": 11735
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.544775352003485e-08,
+      "loss": 1.2051,
+      "step": 11736
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.514402988094137e-08,
+      "loss": 1.1973,
+      "step": 11737
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.484101032458778e-08,
+      "loss": 1.1572,
+      "step": 11738
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.453869487244691e-08,
+      "loss": 1.2207,
+      "step": 11739
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.423708354594604e-08,
+      "loss": 1.1924,
+      "step": 11740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.393617636645921e-08,
+      "loss": 1.1494,
+      "step": 11741
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.363597335531158e-08,
+      "loss": 1.252,
+      "step": 11742
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.333647453377833e-08,
+      "loss": 1.1875,
+      "step": 11743
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.303767992308585e-08,
+      "loss": 1.1953,
+      "step": 11744
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.273958954440939e-08,
+      "loss": 1.1719,
+      "step": 11745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.24422034188743e-08,
+      "loss": 1.1621,
+      "step": 11746
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.214552156755594e-08,
+      "loss": 1.2412,
+      "step": 11747
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.18495440114808e-08,
+      "loss": 1.2158,
+      "step": 11748
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.155427077162545e-08,
+      "loss": 1.165,
+      "step": 11749
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.125970186891539e-08,
+      "loss": 1.1748,
+      "step": 11750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.0965837324225e-08,
+      "loss": 1.208,
+      "step": 11751
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.06726771583821e-08,
+      "loss": 1.1641,
+      "step": 11752
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.038022139216337e-08,
+      "loss": 1.2305,
+      "step": 11753
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.008847004629448e-08,
+      "loss": 1.1689,
+      "step": 11754
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.979742314145109e-08,
+      "loss": 1.2061,
+      "step": 11755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.950708069826116e-08,
+      "loss": 1.2471,
+      "step": 11756
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.921744273730046e-08,
+      "loss": 1.1484,
+      "step": 11757
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.892850927909588e-08,
+      "loss": 1.1738,
+      "step": 11758
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.864028034412217e-08,
+      "loss": 1.1982,
+      "step": 11759
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.835275595280854e-08,
+      "loss": 1.1865,
+      "step": 11760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.806593612553202e-08,
+      "loss": 1.207,
+      "step": 11761
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7779820882618576e-08,
+      "loss": 1.1968,
+      "step": 11762
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.749441024434421e-08,
+      "loss": 1.2197,
+      "step": 11763
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.720970423093719e-08,
+      "loss": 1.1943,
+      "step": 11764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6925702862574705e-08,
+      "loss": 1.1973,
+      "step": 11765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.664240615938399e-08,
+      "loss": 1.2217,
+      "step": 11766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6359814141441204e-08,
+      "loss": 1.1699,
+      "step": 11767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.607792682877478e-08,
+      "loss": 1.1826,
+      "step": 11768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5796744241360944e-08,
+      "loss": 1.1914,
+      "step": 11769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5516266399128216e-08,
+      "loss": 1.2109,
+      "step": 11770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.523649332195402e-08,
+      "loss": 1.1816,
+      "step": 11771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.495742502966362e-08,
+      "loss": 1.1875,
+      "step": 11772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4679061542037836e-08,
+      "loss": 1.167,
+      "step": 11773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4401402878802024e-08,
+      "loss": 1.2217,
+      "step": 11774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4124449059634875e-08,
+      "loss": 1.2363,
+      "step": 11775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3848200104164027e-08,
+      "loss": 1.2461,
+      "step": 11776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.357265603196715e-08,
+      "loss": 1.25,
+      "step": 11777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.329781686257196e-08,
+      "loss": 1.208,
+      "step": 11778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.302368261545621e-08,
+      "loss": 1.2344,
+      "step": 11779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2750253310046574e-08,
+      "loss": 1.2314,
+      "step": 11780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.247752896572311e-08,
+      "loss": 1.2012,
+      "step": 11781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2205509601813696e-08,
+      "loss": 1.1641,
+      "step": 11782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.193419523759402e-08,
+      "loss": 1.1943,
+      "step": 11783
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1663585892294254e-08,
+      "loss": 1.2158,
+      "step": 11784
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1393681585092393e-08,
+      "loss": 1.2119,
+      "step": 11785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1124482335115354e-08,
+      "loss": 1.1924,
+      "step": 11786
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.085598816144233e-08,
+      "loss": 1.1553,
+      "step": 11787
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0588199083099196e-08,
+      "loss": 1.2217,
+      "step": 11788
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.032111511906745e-08,
+      "loss": 1.1953,
+      "step": 11789
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.005473628827417e-08,
+      "loss": 1.1797,
+      "step": 11790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.978906260959759e-08,
+      "loss": 1.1895,
+      "step": 11791
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9524094101864875e-08,
+      "loss": 1.1836,
+      "step": 11792
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.925983078385543e-08,
+      "loss": 1.1982,
+      "step": 11793
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.899627267429763e-08,
+      "loss": 1.2227,
+      "step": 11794
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8733419791868744e-08,
+      "loss": 1.1953,
+      "step": 11795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.847127215519831e-08,
+      "loss": 1.1426,
+      "step": 11796
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8209829782864816e-08,
+      "loss": 1.1797,
+      "step": 11797
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.794909269339565e-08,
+      "loss": 1.2041,
+      "step": 11798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.768906090526937e-08,
+      "loss": 1.1787,
+      "step": 11799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.742973443691568e-08,
+      "loss": 1.2295,
+      "step": 11800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.717111330671209e-08,
+      "loss": 1.1992,
+      "step": 11801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6913197532986176e-08,
+      "loss": 1.1865,
+      "step": 11802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6655987134017755e-08,
+      "loss": 1.1602,
+      "step": 11803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.639948212803558e-08,
+      "loss": 1.166,
+      "step": 11804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.614368253321733e-08,
+      "loss": 1.1641,
+      "step": 11805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.588858836769183e-08,
+      "loss": 1.2305,
+      "step": 11806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.563419964953686e-08,
+      "loss": 1.1641,
+      "step": 11807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5380516396781314e-08,
+      "loss": 1.2373,
+      "step": 11808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.512753862740416e-08,
+      "loss": 1.1934,
+      "step": 11809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.487526635933437e-08,
+      "loss": 1.2666,
+      "step": 11810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.462369961044877e-08,
+      "loss": 1.2109,
+      "step": 11811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.437283839857753e-08,
+      "loss": 1.2451,
+      "step": 11812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.412268274149756e-08,
+      "loss": 1.2168,
+      "step": 11813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.387323265694021e-08,
+      "loss": 1.2373,
+      "step": 11814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.362448816258025e-08,
+      "loss": 1.2412,
+      "step": 11815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3376449276049114e-08,
+      "loss": 1.2842,
+      "step": 11816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3129116014923866e-08,
+      "loss": 1.2217,
+      "step": 11817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.288248839673381e-08,
+      "loss": 1.1982,
+      "step": 11818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2636566438956085e-08,
+      "loss": 1.2188,
+      "step": 11819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.239135015902118e-08,
+      "loss": 1.2041,
+      "step": 11820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.214683957430632e-08,
+      "loss": 1.1846,
+      "step": 11821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1903034702139854e-08,
+      "loss": 1.1631,
+      "step": 11822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.165993555980019e-08,
+      "loss": 1.207,
+      "step": 11823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.141754216451577e-08,
+      "loss": 1.1631,
+      "step": 11824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.117585453346618e-08,
+      "loss": 1.209,
+      "step": 11825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.093487268377882e-08,
+      "loss": 1.1709,
+      "step": 11826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.069459663253228e-08,
+      "loss": 1.1973,
+      "step": 11827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0455026396754027e-08,
+      "loss": 1.1982,
+      "step": 11828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.021616199342382e-08,
+      "loss": 1.1797,
+      "step": 11829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.997800343946923e-08,
+      "loss": 1.2344,
+      "step": 11830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.974055075176786e-08,
+      "loss": 1.209,
+      "step": 11831
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.950380394714959e-08,
+      "loss": 1.1709,
+      "step": 11832
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.926776304239099e-08,
+      "loss": 1.1909,
+      "step": 11833
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9032428054222026e-08,
+      "loss": 1.1543,
+      "step": 11834
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.879779899931935e-08,
+      "loss": 1.2256,
+      "step": 11835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8563875894311875e-08,
+      "loss": 1.2012,
+      "step": 11836
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.833065875577635e-08,
+      "loss": 1.2061,
+      "step": 11837
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8098147600243994e-08,
+      "loss": 1.2354,
+      "step": 11838
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.786634244418941e-08,
+      "loss": 1.2422,
+      "step": 11839
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.763524330404278e-08,
+      "loss": 1.2422,
+      "step": 11840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.740485019618101e-08,
+      "loss": 1.2109,
+      "step": 11841
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7175163136933255e-08,
+      "loss": 1.2314,
+      "step": 11842
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6946182142575395e-08,
+      "loss": 1.2197,
+      "step": 11843
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.671790722933777e-08,
+      "loss": 1.2061,
+      "step": 11844
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.649033841339633e-08,
+      "loss": 1.2148,
+      "step": 11845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.626347571087929e-08,
+      "loss": 1.2314,
+      "step": 11846
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.60373191378649e-08,
+      "loss": 1.1846,
+      "step": 11847
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5811868710380335e-08,
+      "loss": 1.2236,
+      "step": 11848
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.558712444440282e-08,
+      "loss": 1.2109,
+      "step": 11849
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.536308635586183e-08,
+      "loss": 1.1953,
+      "step": 11850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.513975446063356e-08,
+      "loss": 1.1943,
+      "step": 11851
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.491712877454534e-08,
+      "loss": 1.1758,
+      "step": 11852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4695209313375666e-08,
+      "loss": 1.21,
+      "step": 11853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4473996092850845e-08,
+      "loss": 1.1807,
+      "step": 11854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4253489128648346e-08,
+      "loss": 1.1943,
+      "step": 11855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.403368843639676e-08,
+      "loss": 1.1367,
+      "step": 11856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.381459403167142e-08,
+      "loss": 1.2041,
+      "step": 11857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.359620593000212e-08,
+      "loss": 1.1943,
+      "step": 11858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3378524146863156e-08,
+      "loss": 1.2031,
+      "step": 11859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3161548697682177e-08,
+      "loss": 1.1548,
+      "step": 11860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2945279597838e-08,
+      "loss": 1.2109,
+      "step": 11861
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.272971686265725e-08,
+      "loss": 1.209,
+      "step": 11862
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.251486050741548e-08,
+      "loss": 1.1797,
+      "step": 11863
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.230071054733941e-08,
+      "loss": 1.1768,
+      "step": 11864
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2087266997608e-08,
+      "loss": 1.1934,
+      "step": 11865
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1874529873345825e-08,
+      "loss": 1.1748,
+      "step": 11866
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1662499189630824e-08,
+      "loss": 1.2725,
+      "step": 11867
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.145117496148875e-08,
+      "loss": 1.2119,
+      "step": 11868
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1240557203896516e-08,
+      "loss": 1.1592,
+      "step": 11869
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1030645931781066e-08,
+      "loss": 1.2305,
+      "step": 11870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.082144116001828e-08,
+      "loss": 1.2031,
+      "step": 11871
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0612942903434084e-08,
+      "loss": 1.1191,
+      "step": 11872
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.040515117680553e-08,
+      "loss": 1.1865,
+      "step": 11873
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.019806599485864e-08,
+      "loss": 1.1816,
+      "step": 11874
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.999168737226943e-08,
+      "loss": 1.1982,
+      "step": 11875
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9786015323663986e-08,
+      "loss": 1.21,
+      "step": 11876
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9581049863618426e-08,
+      "loss": 1.249,
+      "step": 11877
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.93767910066578e-08,
+      "loss": 1.1924,
+      "step": 11878
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9173238767258305e-08,
+      "loss": 1.1475,
+      "step": 11879
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.897039315984618e-08,
+      "loss": 1.2734,
+      "step": 11880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8768254198797695e-08,
+      "loss": 1.21,
+      "step": 11881
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.856682189843585e-08,
+      "loss": 1.1836,
+      "step": 11882
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8366096273039213e-08,
+      "loss": 1.1748,
+      "step": 11883
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8166077336830854e-08,
+      "loss": 1.2305,
+      "step": 11884
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7966765103987213e-08,
+      "loss": 1.1914,
+      "step": 11885
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.776815958863477e-08,
+      "loss": 1.2197,
+      "step": 11886
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7570260804845596e-08,
+      "loss": 1.1465,
+      "step": 11887
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7373068766647358e-08,
+      "loss": 1.1533,
+      "step": 11888
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.717658348801333e-08,
+      "loss": 1.1777,
+      "step": 11889
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6980804982870144e-08,
+      "loss": 1.2021,
+      "step": 11890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.678573326509115e-08,
+      "loss": 1.168,
+      "step": 11891
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6591368348501957e-08,
+      "loss": 1.1523,
+      "step": 11892
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6397710246877096e-08,
+      "loss": 1.1924,
+      "step": 11893
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6204758973941145e-08,
+      "loss": 1.165,
+      "step": 11894
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.601251454336762e-08,
+      "loss": 1.2197,
+      "step": 11895
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5820976968781164e-08,
+      "loss": 1.208,
+      "step": 11896
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5630146263757595e-08,
+      "loss": 1.252,
+      "step": 11897
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.544002244181942e-08,
+      "loss": 1.2031,
+      "step": 11898
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5250605516440318e-08,
+      "loss": 1.1924,
+      "step": 11899
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5061895501046206e-08,
+      "loss": 1.2412,
+      "step": 11900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4873892409009724e-08,
+      "loss": 1.1636,
+      "step": 11901
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.468659625365466e-08,
+      "loss": 1.2148,
+      "step": 11902
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4500007048254838e-08,
+      "loss": 1.1895,
+      "step": 11903
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4314124806033014e-08,
+      "loss": 1.1963,
+      "step": 11904
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4128949540164204e-08,
+      "loss": 1.1567,
+      "step": 11905
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3944481263771245e-08,
+      "loss": 1.1973,
+      "step": 11906
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3760719989927018e-08,
+      "loss": 1.1982,
+      "step": 11907
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.357766573165443e-08,
+      "loss": 1.1797,
+      "step": 11908
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3395318501927555e-08,
+      "loss": 1.1729,
+      "step": 11909
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3213678313668276e-08,
+      "loss": 1.2559,
+      "step": 11910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3032745179749626e-08,
+      "loss": 1.25,
+      "step": 11911
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2852519112994687e-08,
+      "loss": 1.2373,
+      "step": 11912
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.267300012617546e-08,
+      "loss": 1.2266,
+      "step": 11913
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2494188232015102e-08,
+      "loss": 1.2119,
+      "step": 11914
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2316083443186808e-08,
+      "loss": 1.1426,
+      "step": 11915
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2138685772310486e-08,
+      "loss": 1.2334,
+      "step": 11916
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.196199523196052e-08,
+      "loss": 1.2451,
+      "step": 11917
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1786011834658005e-08,
+      "loss": 1.2002,
+      "step": 11918
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1610735592875188e-08,
+      "loss": 1.2598,
+      "step": 11919
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.143616651903324e-08,
+      "loss": 1.2568,
+      "step": 11920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1262304625504493e-08,
+      "loss": 1.2305,
+      "step": 11921
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1089149924610196e-08,
+      "loss": 1.1787,
+      "step": 11922
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0916702428622758e-08,
+      "loss": 1.167,
+      "step": 11923
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0744962149761294e-08,
+      "loss": 1.2119,
+      "step": 11924
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0573929100199395e-08,
+      "loss": 1.1904,
+      "step": 11925
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0403603292056263e-08,
+      "loss": 1.1758,
+      "step": 11926
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0233984737405565e-08,
+      "loss": 1.2021,
+      "step": 11927
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.006507344826547e-08,
+      "loss": 1.1826,
+      "step": 11928
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9896869436607514e-08,
+      "loss": 1.2627,
+      "step": 11929
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9729372714352158e-08,
+      "loss": 1.1826,
+      "step": 11930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9562583293371017e-08,
+      "loss": 1.1748,
+      "step": 11931
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.939650118548353e-08,
+      "loss": 1.2246,
+      "step": 11932
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9231126402459167e-08,
+      "loss": 1.2168,
+      "step": 11933
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.906645895601966e-08,
+      "loss": 1.209,
+      "step": 11934
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8902498857834572e-08,
+      "loss": 1.1924,
+      "step": 11935
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.873924611952238e-08,
+      "loss": 1.2158,
+      "step": 11936
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.857670075265494e-08,
+      "loss": 1.207,
+      "step": 11937
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8414862768749708e-08,
+      "loss": 1.2139,
+      "step": 11938
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8253732179277504e-08,
+      "loss": 1.1934,
+      "step": 11939
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.809330899565698e-08,
+      "loss": 1.2285,
+      "step": 11940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7933593229257917e-08,
+      "loss": 1.1562,
+      "step": 11941
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.777458489139905e-08,
+      "loss": 1.2109,
+      "step": 11942
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.761628399334914e-08,
+      "loss": 1.1494,
+      "step": 11943
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7458690546326985e-08,
+      "loss": 1.2129,
+      "step": 11944
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7301804561502543e-08,
+      "loss": 1.1982,
+      "step": 11945
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7145626049991372e-08,
+      "loss": 1.1465,
+      "step": 11946
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6990155022864607e-08,
+      "loss": 1.2324,
+      "step": 11947
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6835391491139e-08,
+      "loss": 1.2129,
+      "step": 11948
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6681335465783543e-08,
+      "loss": 1.2236,
+      "step": 11949
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.652798695771618e-08,
+      "loss": 1.1875,
+      "step": 11950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.637534597780377e-08,
+      "loss": 1.2168,
+      "step": 11951
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6223412536864326e-08,
+      "loss": 1.1719,
+      "step": 11952
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.607218664566479e-08,
+      "loss": 1.2441,
+      "step": 11953
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5921668314924367e-08,
+      "loss": 1.1865,
+      "step": 11954
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5771857555308968e-08,
+      "loss": 1.2412,
+      "step": 11955
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5622754377435656e-08,
+      "loss": 1.1982,
+      "step": 11956
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5474358791871536e-08,
+      "loss": 1.208,
+      "step": 11957
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5326670809133748e-08,
+      "loss": 1.166,
+      "step": 11958
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5179690439688366e-08,
+      "loss": 1.1826,
+      "step": 11959
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5033417693951502e-08,
+      "loss": 1.1621,
+      "step": 11960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4887852582290419e-08,
+      "loss": 1.1943,
+      "step": 11961
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4742995115021307e-08,
+      "loss": 1.1943,
+      "step": 11962
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4598845302410402e-08,
+      "loss": 1.1914,
+      "step": 11963
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4455403154672865e-08,
+      "loss": 1.2539,
+      "step": 11964
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4312668681975006e-08,
+      "loss": 1.2178,
+      "step": 11965
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4170641894430959e-08,
+      "loss": 1.1641,
+      "step": 11966
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4029322802108225e-08,
+      "loss": 1.1738,
+      "step": 11967
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3888711415021016e-08,
+      "loss": 1.1914,
+      "step": 11968
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3748807743133586e-08,
+      "loss": 1.2178,
+      "step": 11969
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3609611796362443e-08,
+      "loss": 1.2178,
+      "step": 11970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471123584573032e-08,
+      "loss": 1.1934,
+      "step": 11971
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3333343117577502e-08,
+      "loss": 1.1211,
+      "step": 11972
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3196270405142486e-08,
+      "loss": 1.1875,
+      "step": 11973
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3059905456980215e-08,
+      "loss": 1.1562,
+      "step": 11974
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.29242482827574e-08,
+      "loss": 1.2061,
+      "step": 11975
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2789298892086355e-08,
+      "loss": 1.1616,
+      "step": 11976
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2655057294531648e-08,
+      "loss": 1.2305,
+      "step": 11977
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.252152349960567e-08,
+      "loss": 1.1846,
+      "step": 11978
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2388697516774184e-08,
+      "loss": 1.2002,
+      "step": 11979
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.225657935544966e-08,
+      "loss": 1.1719,
+      "step": 11980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2125169024993499e-08,
+      "loss": 1.1514,
+      "step": 11981
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1994466534721582e-08,
+      "loss": 1.1816,
+      "step": 11982
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.186447189389539e-08,
+      "loss": 1.2002,
+      "step": 11983
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1735185111726443e-08,
+      "loss": 1.2227,
+      "step": 11984
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1606606197379634e-08,
+      "loss": 1.2021,
+      "step": 11985
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1478735159966559e-08,
+      "loss": 1.1953,
+      "step": 11986
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1351572008548861e-08,
+      "loss": 1.2275,
+      "step": 11987
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.122511675213822e-08,
+      "loss": 1.2129,
+      "step": 11988
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1099369399697468e-08,
+      "loss": 1.1582,
+      "step": 11989
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0974329960138364e-08,
+      "loss": 1.1953,
+      "step": 11990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0849998442322707e-08,
+      "loss": 1.2012,
+      "step": 11991
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0726374855060118e-08,
+      "loss": 1.1963,
+      "step": 11992
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0603459207113586e-08,
+      "loss": 1.209,
+      "step": 11993
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0481251507192814e-08,
+      "loss": 1.1514,
+      "step": 11994
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0359751763959757e-08,
+      "loss": 1.1963,
+      "step": 11995
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.02389599860242e-08,
+      "loss": 1.1797,
+      "step": 11996
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0118876181948179e-08,
+      "loss": 1.1904,
+      "step": 11997
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.999500360239334e-09,
+      "loss": 1.165,
+      "step": 11998
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.880832529360895e-09,
+      "loss": 1.2451,
+      "step": 11999
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.762872697720582e-09,
+      "loss": 1.251,
+      "step": 12000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.645620873679484e-09,
+      "loss": 1.2041,
+      "step": 12001
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.52907706554651e-09,
+      "loss": 1.1719,
+      "step": 12002
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.413241281581719e-09,
+      "loss": 1.2188,
+      "step": 12003
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.298113529994101e-09,
+      "loss": 1.2441,
+      "step": 12004
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.183693818942686e-09,
+      "loss": 1.1689,
+      "step": 12005
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.069982156535428e-09,
+      "loss": 1.1982,
+      "step": 12006
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.956978550833662e-09,
+      "loss": 1.2051,
+      "step": 12007
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.844683009843202e-09,
+      "loss": 1.1885,
+      "step": 12008
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.73309554152324e-09,
+      "loss": 1.2109,
+      "step": 12009
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.622216153783003e-09,
+      "loss": 1.1738,
+      "step": 12010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.51204485447954e-09,
+      "loss": 1.207,
+      "step": 12011
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.402581651421049e-09,
+      "loss": 1.1855,
+      "step": 12012
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.29382655236466e-09,
+      "loss": 1.1602,
+      "step": 12013
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.185779565018648e-09,
+      "loss": 1.1855,
+      "step": 12014
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.078440697040225e-09,
+      "loss": 1.1699,
+      "step": 12015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.971809956035525e-09,
+      "loss": 1.2061,
+      "step": 12016
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.865887349561841e-09,
+      "loss": 1.2393,
+      "step": 12017
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.760672885126496e-09,
+      "loss": 1.1738,
+      "step": 12018
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.656166570185753e-09,
+      "loss": 1.1709,
+      "step": 12019
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.552368412145905e-09,
+      "loss": 1.1904,
+      "step": 12020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.449278418363293e-09,
+      "loss": 1.1826,
+      "step": 12021
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.3468965961442925e-09,
+      "loss": 1.1943,
+      "step": 12022
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.245222952743103e-09,
+      "loss": 1.2021,
+      "step": 12023
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.14425749536729e-09,
+      "loss": 1.1709,
+      "step": 12024
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.044000231171133e-09,
+      "loss": 1.2021,
+      "step": 12025
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.944451167258948e-09,
+      "loss": 1.2432,
+      "step": 12026
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.845610310688422e-09,
+      "loss": 1.208,
+      "step": 12027
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.7474776684617324e-09,
+      "loss": 1.1729,
+      "step": 12028
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.650053247535537e-09,
+      "loss": 1.2324,
+      "step": 12029
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.553337054813203e-09,
+      "loss": 1.1846,
+      "step": 12030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.4573290971492455e-09,
+      "loss": 1.2275,
+      "step": 12031
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.362029381347112e-09,
+      "loss": 1.1675,
+      "step": 12032
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.2674379141614005e-09,
+      "loss": 1.1826,
+      "step": 12033
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.173554702296747e-09,
+      "loss": 1.1865,
+      "step": 12034
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.080379752404497e-09,
+      "loss": 1.2119,
+      "step": 12035
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9879130710904785e-09,
+      "loss": 1.1982,
+      "step": 12036
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.896154664905007e-09,
+      "loss": 1.2451,
+      "step": 12037
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.8051045403528795e-09,
+      "loss": 1.2139,
+      "step": 12038
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7147627038867114e-09,
+      "loss": 1.249,
+      "step": 12039
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.62512916190805e-09,
+      "loss": 1.0996,
+      "step": 12040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.5362039207707e-09,
+      "loss": 1.1846,
+      "step": 12041
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.447986986775178e-09,
+      "loss": 1.2207,
+      "step": 12042
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3604783661742595e-09,
+      "loss": 1.1885,
+      "step": 12043
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.273678065169652e-09,
+      "loss": 1.2432,
+      "step": 12044
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.187586089911989e-09,
+      "loss": 1.21,
+      "step": 12045
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.102202446504168e-09,
+      "loss": 1.2275,
+      "step": 12046
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.017527140996903e-09,
+      "loss": 1.1396,
+      "step": 12047
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.933560179389841e-09,
+      "loss": 1.2266,
+      "step": 12048
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.850301567635996e-09,
+      "loss": 1.1807,
+      "step": 12049
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.767751311632874e-09,
+      "loss": 1.1895,
+      "step": 12050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.685909417233569e-09,
+      "loss": 1.2246,
+      "step": 12051
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.604775890237889e-09,
+      "loss": 1.1768,
+      "step": 12052
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.524350736393457e-09,
+      "loss": 1.1885,
+      "step": 12053
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.444633961402378e-09,
+      "loss": 1.208,
+      "step": 12054
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.365625570913467e-09,
+      "loss": 1.1729,
+      "step": 12055
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2873255705255805e-09,
+      "loss": 1.1973,
+      "step": 12056
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.20973396578761e-09,
+      "loss": 1.2363,
+      "step": 12057
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.132850762199603e-09,
+      "loss": 1.2197,
+      "step": 12058
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0566759652105326e-09,
+      "loss": 1.2202,
+      "step": 12059
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.981209580217193e-09,
+      "loss": 1.1865,
+      "step": 12060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.90645161256864e-09,
+      "loss": 1.2051,
+      "step": 12061
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.832402067563968e-09,
+      "loss": 1.2559,
+      "step": 12062
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.759060950448978e-09,
+      "loss": 1.2451,
+      "step": 12063
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.686428266423958e-09,
+      "loss": 1.1768,
+      "step": 12064
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.61450402063368e-09,
+      "loss": 1.1396,
+      "step": 12065
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5432882181773985e-09,
+      "loss": 1.2168,
+      "step": 12066
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4727808641021878e-09,
+      "loss": 1.2256,
+      "step": 12067
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.40298196340294e-09,
+      "loss": 1.1641,
+      "step": 12068
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.333891521027921e-09,
+      "loss": 1.2139,
+      "step": 12069
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.265509541873213e-09,
+      "loss": 1.2139,
+      "step": 12070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.19783603078605e-09,
+      "loss": 1.2158,
+      "step": 12071
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1308709925603755e-09,
+      "loss": 1.1807,
+      "step": 12072
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.064614431943502e-09,
+      "loss": 1.25,
+      "step": 12073
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.999066353630564e-09,
+      "loss": 1.2109,
+      "step": 12074
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9342267622667344e-09,
+      "loss": 1.1924,
+      "step": 12075
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8700956624472254e-09,
+      "loss": 1.2061,
+      "step": 12076
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8066730587172908e-09,
+      "loss": 1.1929,
+      "step": 12077
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7439589555722234e-09,
+      "loss": 1.1924,
+      "step": 12078
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.681953357455136e-09,
+      "loss": 1.2168,
+      "step": 12079
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6206562687625114e-09,
+      "loss": 1.1836,
+      "step": 12080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5600676938364323e-09,
+      "loss": 1.2012,
+      "step": 12081
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5001876369712406e-09,
+      "loss": 1.2207,
+      "step": 12082
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.44101610241132e-09,
+      "loss": 1.1797,
+      "step": 12083
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3825530943488716e-09,
+      "loss": 1.2578,
+      "step": 12084
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.324798616929469e-09,
+      "loss": 1.1855,
+      "step": 12085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2677526742431733e-09,
+      "loss": 1.1699,
+      "step": 12086
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2114152703356372e-09,
+      "loss": 1.1816,
+      "step": 12087
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.155786409198113e-09,
+      "loss": 1.2344,
+      "step": 12088
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.100866094773002e-09,
+      "loss": 1.1523,
+      "step": 12089
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0466543309527464e-09,
+      "loss": 1.1797,
+      "step": 12090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.993151121578718e-09,
+      "loss": 1.1963,
+      "step": 12091
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9403564704434386e-09,
+      "loss": 1.2148,
+      "step": 12092
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8882703812883596e-09,
+      "loss": 1.2363,
+      "step": 12093
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8368928578049728e-09,
+      "loss": 1.2217,
+      "step": 12094
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7862239036336992e-09,
+      "loss": 1.167,
+      "step": 12095
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7362635223661106e-09,
+      "loss": 1.2627,
+      "step": 12096
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6870117175427082e-09,
+      "loss": 1.1963,
+      "step": 12097
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6384684926529226e-09,
+      "loss": 1.249,
+      "step": 12098
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5906338511395558e-09,
+      "loss": 1.1826,
+      "step": 12099
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5435077963898981e-09,
+      "loss": 1.248,
+      "step": 12100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.497090331745721e-09,
+      "loss": 1.1826,
+      "step": 12101
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4513814604955044e-09,
+      "loss": 1.2139,
+      "step": 12102
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.40638118587888e-09,
+      "loss": 1.2441,
+      "step": 12103
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.362089511085518e-09,
+      "loss": 1.2217,
+      "step": 12104
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3185064392540193e-09,
+      "loss": 1.1851,
+      "step": 12105
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2756319734730238e-09,
+      "loss": 1.1221,
+      "step": 12106
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.233466116780102e-09,
+      "loss": 1.1953,
+      "step": 12107
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1920088721661948e-09,
+      "loss": 1.208,
+      "step": 12108
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1512602425667318e-09,
+      "loss": 1.2236,
+      "step": 12109
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1112202308705133e-09,
+      "loss": 1.1729,
+      "step": 12110
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0718888399152693e-09,
+      "loss": 1.2139,
+      "step": 12111
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.03326607248877e-09,
+      "loss": 1.1562,
+      "step": 12112
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.953519313277148e-10,
+      "loss": 1.2217,
+      "step": 12113
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.581464191188438e-10,
+      "loss": 1.1797,
+      "step": 12114
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.216495384989366e-10,
+      "loss": 1.166,
+      "step": 12115
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.858612920559228e-10,
+      "loss": 1.2021,
+      "step": 12116
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.507816823244419e-10,
+      "loss": 1.1973,
+      "step": 12117
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.164107117913933e-10,
+      "loss": 1.1943,
+      "step": 12118
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.827483828914961e-10,
+      "loss": 1.2197,
+      "step": 12119
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.497946980117298e-10,
+      "loss": 1.2354,
+      "step": 12120
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.175496594880038e-10,
+      "loss": 1.1787,
+      "step": 12121
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.860132696029364e-10,
+      "loss": 1.2266,
+      "step": 12122
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.551855305947374e-10,
+      "loss": 1.2295,
+      "step": 12123
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.250664446461052e-10,
+      "loss": 1.208,
+      "step": 12124
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.956560138919987e-10,
+      "loss": 1.2578,
+      "step": 12125
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.669542404163064e-10,
+      "loss": 1.1885,
+      "step": 12126
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.389611262540672e-10,
+      "loss": 1.208,
+      "step": 12127
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.116766733892497e-10,
+      "loss": 1.2236,
+      "step": 12128
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.851008837547522e-10,
+      "loss": 1.2207,
+      "step": 12129
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5923375923351276e-10,
+      "loss": 1.1709,
+      "step": 12130
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3407530165962e-10,
+      "loss": 1.1836,
+      "step": 12131
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0962551281609195e-10,
+      "loss": 1.1416,
+      "step": 12132
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.858843944359869e-10,
+      "loss": 1.1968,
+      "step": 12133
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.628519482001824e-10,
+      "loss": 1.2139,
+      "step": 12134
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4052817574292684e-10,
+      "loss": 1.2256,
+      "step": 12135
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1891307864628797e-10,
+      "loss": 1.2197,
+      "step": 12136
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.980066584401531e-10,
+      "loss": 1.1348,
+      "step": 12137
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.778089166077802e-10,
+      "loss": 1.2539,
+      "step": 12138
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5831985457913657e-10,
+      "loss": 1.2256,
+      "step": 12139
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3953947373756e-10,
+      "loss": 1.2227,
+      "step": 12140
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2146777541309782e-10,
+      "loss": 1.252,
+      "step": 12141
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0410476088583708e-10,
+      "loss": 1.1748,
+      "step": 12142
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8745043138590491e-10,
+      "loss": 1.2266,
+      "step": 12143
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7150478809568882e-10,
+      "loss": 1.2129,
+      "step": 12144
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5626783214317543e-10,
+      "loss": 1.2012,
+      "step": 12145
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4173956460972194e-10,
+      "loss": 1.2344,
+      "step": 12146
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2791998652339488e-10,
+      "loss": 1.2051,
+      "step": 12147
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1480909886452119e-10,
+      "loss": 1.2109,
+      "step": 12148
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0240690256235752e-10,
+      "loss": 1.2246,
+      "step": 12149
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.071339849620054e-11,
+      "loss": 1.2061,
+      "step": 12150
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.972858749427659e-11,
+      "loss": 1.2031,
+      "step": 12151
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.945247033374181e-11,
+      "loss": 1.1719,
+      "step": 12152
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.988504774512294e-11,
+      "loss": 1.2529,
+      "step": 12153
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.102632040565603e-11,
+      "loss": 1.2163,
+      "step": 12154
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2876288942617087e-11,
+      "loss": 1.1846,
+      "step": 12155
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5434953934432305e-11,
+      "loss": 1.2256,
+      "step": 12156
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.87023159073474e-11,
+      "loss": 1.208,
+      "step": 12157
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2678375339868498e-11,
+      "loss": 1.124,
+      "step": 12158
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7363132657211014e-11,
+      "loss": 1.1729,
+      "step": 12159
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2756588237960998e-11,
+      "loss": 1.1836,
+      "step": 12160
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.8587424074138e-12,
+      "loss": 1.2168,
+      "step": 12161
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.669595442014952e-12,
+      "loss": 1.2012,
+      "step": 12162
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1891475682499507e-12,
+      "loss": 1.1743,
+      "step": 12163
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4173989604238104e-12,
+      "loss": 1.1748,
+      "step": 12164
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5434974621217923e-13,
+      "loss": 1.2314,
+      "step": 12165
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 1.2266,
+      "step": 12166
+    },
+    {
+      "epoch": 1.0,
+      "step": 12166,
+      "total_flos": 2.4823622704163717e+19,
+      "train_loss": 1.2700025140504274,
+      "train_runtime": 143056.2736,
+      "train_samples_per_second": 10.885,
+      "train_steps_per_second": 0.085
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 12166,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 608,
+  "total_flos": 2.4823622704163717e+19,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}